mirror of
https://github.com/BerriAI/litellm.git
synced 2026-10-03 02:22:24 +00:00
refactor(rust): move tests.rs files inline or under tests/ and drop autotests = false (#43028)
* refactor(rust): move tests.rs files inline or under tests/ Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> * refactor(rust): move tests.rs files inline or under tests/ Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> * refactor(rust): inline path-included test files into their owning src files Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> * chore(rust): drop stray proptest regression file Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> * test(rust): cover lowercase, empty and non-authorization headers in bearer detection Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --------- Co-authored-by: Yujong Lee <yujong@berri.ai> Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
This commit is contained in:
parent
1fbd1e9ce9
commit
e2781c4713
39 changed files with 6486 additions and 6359 deletions
10
litellm-rust/AGENTS.md
Normal file
10
litellm-rust/AGENTS.md
Normal file
|
|
@ -0,0 +1,10 @@
|
|||
# Rust workspace rules
|
||||
|
||||
## Test placement
|
||||
|
||||
- Never create a `tests.rs` (or `test.rs`) file under `src/`, and never `#[path = "tests.rs"] mod tests;`
|
||||
- A test that reaches private items lives inline, in a `#[cfg(test)] mod tests { ... }` at the bottom of the file that owns those items
|
||||
- A test that only uses the crate's public API lives in `crates/<crate>/tests/<subject>.rs`, next to `src/`
|
||||
- Split a mixed test file along that line instead of widening visibility to move it
|
||||
- A test for another crate's item belongs in that crate, not in a downstream one
|
||||
- Never set `autotests = false` or hand-list `[[test]]` targets; every file directly under `tests/` is discovered by cargo, and a shared helper goes in `tests/<name>/mod.rs` or `tests/<subject>/support.rs` so it is not picked up as a test crate of its own
|
||||
|
|
@ -4,7 +4,6 @@ version = "0.1.0"
|
|||
edition.workspace = true
|
||||
license.workspace = true
|
||||
repository.workspace = true
|
||||
autotests = false
|
||||
|
||||
[dependencies]
|
||||
litellm-host.workspace = true
|
||||
|
|
|
|||
File diff suppressed because it is too large
Load diff
|
|
@ -63,5 +63,151 @@ impl PendingLogging {
|
|||
}
|
||||
|
||||
#[cfg(test)]
|
||||
#[path = "../tests/deferred.rs"]
|
||||
mod tests;
|
||||
mod tests {
|
||||
use std::ffi::CStr;
|
||||
|
||||
use pyo3::prelude::*;
|
||||
use pyo3::types::PyDict;
|
||||
use rstest::rstest;
|
||||
|
||||
use super::{PendingLogging, PendingSuccess};
|
||||
use crate::PythonLogger;
|
||||
use crate::test_support::{local, namespace, run};
|
||||
|
||||
/// A deferred success for the namespace's `logger` and `response`, bound as `pending`.
|
||||
fn defer<'py>(py: Python<'py>, script: &CStr) -> Bound<'py, PyDict> {
|
||||
let locals = namespace(py, c"response = object()");
|
||||
run(py, &locals, script);
|
||||
let pending = Py::new(
|
||||
py,
|
||||
PendingLogging {
|
||||
pending: Some(PendingSuccess {
|
||||
logger: PythonLogger::new(local(&locals, "logger").unbind()),
|
||||
response: Some(local(&locals, "response").unbind()),
|
||||
start: py.None(),
|
||||
end: Some(py.None()),
|
||||
}),
|
||||
},
|
||||
)
|
||||
.unwrap();
|
||||
locals.set_item("pending", pending).unwrap();
|
||||
locals
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn release_enqueues_the_success_once_in_the_releasing_context() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = defer(
|
||||
py,
|
||||
c"
|
||||
from contextvars import ContextVar
|
||||
|
||||
marker = ContextVar('marker', default='unset')
|
||||
observed = []
|
||||
|
||||
def on_enqueue(coroutine):
|
||||
observed.append(marker.get())
|
||||
pending.release(True)
|
||||
|
||||
logger.on_enqueue = on_enqueue
|
||||
",
|
||||
);
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
marker.set('release')
|
||||
pending.release(True)
|
||||
pending.release(True)
|
||||
assert observed == ['release'], observed
|
||||
assert logger.names() == ['async_success_handler', 'enqueued'], logger.calls
|
||||
assert logger.calls[0][1] is response
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_blocked_release_drops_the_success_for_good() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = defer(py, c"");
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
pending.release(False)
|
||||
pending.release(True)
|
||||
assert logger.calls == [], logger.calls
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::ordinary_error(c"RuntimeError('queue full')", false)]
|
||||
#[case::cancellation(c"asyncio.CancelledError()", true)]
|
||||
fn a_failed_enqueue_closes_the_coroutine_and_is_never_replayed(
|
||||
#[case] failure: &CStr,
|
||||
#[case] propagates: bool,
|
||||
) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = defer(
|
||||
py,
|
||||
c"
|
||||
import asyncio
|
||||
|
||||
def on_enqueue(coroutine):
|
||||
raise failure
|
||||
|
||||
logger.on_enqueue = on_enqueue
|
||||
",
|
||||
);
|
||||
locals
|
||||
.set_item("failure", py.eval(failure, None, Some(&locals)).unwrap())
|
||||
.unwrap();
|
||||
let released = local(&locals, "pending").call_method1("release", (true,));
|
||||
match released {
|
||||
Ok(_) => assert!(!propagates),
|
||||
Err(error) => {
|
||||
assert!(propagates);
|
||||
assert!(error.value(py).is(local(&locals, "failure")));
|
||||
}
|
||||
}
|
||||
locals.set_item("propagates", propagates).unwrap();
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
pending.release(True)
|
||||
assert logger.names() == ['async_success_handler', 'enqueued', 'closed'], logger.calls
|
||||
assert unraisable_from(logger) == ([] if propagates else [failure])
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn an_unreleased_success_does_not_keep_its_logger_alive() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = defer(py, c"");
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
import gc
|
||||
import weakref
|
||||
|
||||
logger.pending = pending
|
||||
reference = weakref.ref(logger)
|
||||
del logger, pending
|
||||
gc.collect()
|
||||
assert reference() is None
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -16,13 +16,218 @@ mod deferred;
|
|||
mod logger;
|
||||
mod preparation;
|
||||
mod python;
|
||||
#[cfg(test)]
|
||||
#[path = "../tests/support.rs"]
|
||||
mod test_support;
|
||||
|
||||
pub(crate) use adapter::LegacyLogging;
|
||||
pub use adapter::{LegacySurface, PassThroughStream};
|
||||
pub use call::{PublicCall, run_legacy_call};
|
||||
pub(crate) use callbacks::{LegacyCallbacks, is_internal_call};
|
||||
pub(crate) use logger::{DeploymentHooks, PythonLogger, finalize, setup};
|
||||
pub(crate) use preparation::prepare;
|
||||
|
||||
#[cfg(test)]
|
||||
mod test_support {
|
||||
use std::ffi::CStr;
|
||||
|
||||
use pyo3::prelude::*;
|
||||
use pyo3::types::{PyDict, PyTuple};
|
||||
|
||||
use crate::{LegacyLogging, LegacySurface, PublicCall};
|
||||
|
||||
/// The parameters of every `callbacks_legacy_python` function, as the real module declares them.
|
||||
/// `tests/test_litellm/rust_bridge/test_callbacks_legacy_python.py` pins this file to the Python
|
||||
/// signatures, and [`namespace`] binds every fake call against it.
|
||||
pub(crate) const PYTHON_CONTRACT: &str = include_str!("../python_contract.json");
|
||||
|
||||
/// Stand-ins for `callbacks_legacy_python`, the only Python module the crate calls. Tests
|
||||
/// share one interpreter and run concurrently, so each fake is installed idempotently and
|
||||
/// forwards to the per-test `StubLogger` it is handed (directly, or as `kwargs['logger']`).
|
||||
/// Every fake is bound against the contract first, so a call the real module would reject
|
||||
/// fails here too.
|
||||
const STUBS: &CStr = c"
|
||||
import contextvars
|
||||
import inspect
|
||||
import json
|
||||
import sys
|
||||
import traceback
|
||||
import types
|
||||
|
||||
for name in ('litellm', 'litellm.rust_bridge', 'litellm.rust_bridge.callbacks_legacy_python'):
|
||||
sys.modules.setdefault(name, types.ModuleType(name))
|
||||
|
||||
legacy = sys.modules['litellm.rust_bridge.callbacks_legacy_python']
|
||||
CONTRACT = json.loads(python_contract)
|
||||
|
||||
|
||||
def contracted(name, fake):
|
||||
signature = inspect.Signature(
|
||||
[inspect.Parameter(parameter, inspect.Parameter.POSITIONAL_OR_KEYWORD) for parameter in CONTRACT[name]]
|
||||
)
|
||||
|
||||
def checked(*args, **kwargs):
|
||||
signature.bind(*args, **kwargs)
|
||||
return fake(*args, **kwargs)
|
||||
|
||||
return checked
|
||||
|
||||
|
||||
if not hasattr(legacy, 'is_internal'):
|
||||
legacy.is_internal = contextvars.ContextVar('is_internal_call', default=False)
|
||||
|
||||
FAKES = {
|
||||
'setup': lambda call_type, args, kwargs, start, asynchronous: types.SimpleNamespace(
|
||||
logger=kwargs['logger_factory'](kwargs) if 'logger_factory' in kwargs else kwargs['logger'],
|
||||
kwargs=kwargs,
|
||||
),
|
||||
'check_limits': lambda arguments: arguments['logger'].check_limits(arguments),
|
||||
'finalize': lambda response, logger, kwargs, start, end: logger.record('finalize', response),
|
||||
'update_logging': lambda logger, kwargs, model, optional_params, litellm_params, provider: logger.update_from_kwargs(
|
||||
kwargs=kwargs,
|
||||
model=model,
|
||||
optional_params=optional_params,
|
||||
litellm_params=litellm_params,
|
||||
custom_llm_provider=provider,
|
||||
),
|
||||
'pre_call': lambda logger, input, api_key, additional_args: logger.pre_call(input, api_key, additional_args),
|
||||
'post_call': lambda logger, original_response, api_key, additional_args: logger.post_call(
|
||||
original_response, api_key, additional_args
|
||||
),
|
||||
'defers_async_logging': lambda logger: bool(getattr(logger, '_defer_async_logging', False)),
|
||||
'defer_success': lambda logger, pending: setattr(logger, '_native_pending_logging', pending),
|
||||
'sync_success_for_async_call': lambda logger, response, start, end: logger.handle_sync_success_callbacks_for_async_calls(
|
||||
response, start, end
|
||||
),
|
||||
'failure_handler': lambda logger, error, start, end, asynchronous: (
|
||||
logger.async_failure_handler if asynchronous else logger.failure_handler
|
||||
)(error, ''.join(traceback.format_exception(error)), start, end),
|
||||
'submit_success': lambda logger, response, start, end: logger.record('submit', (response, start, end)),
|
||||
'async_success_handler': lambda logger, response, start, end: logger.async_success_handler(response, start, end),
|
||||
'enqueue_logging': lambda coroutine: coroutine.enqueue(),
|
||||
'restore_context': lambda logger: logger.record('restore', None),
|
||||
'custom_pricing_fields': lambda: ('ocr_cost_per_page',),
|
||||
'is_internal_call': lambda: legacy.is_internal.get(),
|
||||
'credential_list': lambda: [],
|
||||
'warn_unknown_credential': lambda name, loaded: None,
|
||||
'before_deployment_call': lambda kwargs, call_type: kwargs['logger'].hook('pre', kwargs, call_type),
|
||||
'after_deployment_success': lambda kwargs, response, call_type: kwargs['logger'].hook(
|
||||
'success', response, call_type
|
||||
),
|
||||
'after_deployment_failure': lambda kwargs, error, call_type: kwargs['logger'].hook('failure', error, call_type),
|
||||
'stream_opened': lambda logger: logger.record('stream_opened', None),
|
||||
'stream_success': lambda logger, request_body, chunks, start, end, first_chunk: logger.record(
|
||||
'stream_success', list(chunks)
|
||||
),
|
||||
'stream_failure': lambda logger, request_body, chunks, error: logger.record('stream_failure', error),
|
||||
}
|
||||
assert FAKES.keys() == CONTRACT.keys(), sorted(FAKES.keys() ^ CONTRACT.keys())
|
||||
for name, fake in FAKES.items():
|
||||
setattr(legacy, name, contracted(name, fake))
|
||||
|
||||
|
||||
unraisable = sys.modules.setdefault(
|
||||
'litellm_test_unraisable', types.ModuleType('litellm_test_unraisable')
|
||||
)
|
||||
if not hasattr(unraisable, 'events'):
|
||||
unraisable.events = []
|
||||
sys.unraisablehook = lambda event: unraisable.events.append((event.object, event.exc_value))
|
||||
|
||||
|
||||
def unraisable_from(owner):
|
||||
return [error for source, error in unraisable.events if source is owner]
|
||||
|
||||
|
||||
class StubCoroutine:
|
||||
def __init__(self, logger):
|
||||
self.logger = logger
|
||||
|
||||
def enqueue(self):
|
||||
self.logger.record('enqueued', None)
|
||||
self.logger.on_enqueue(self)
|
||||
|
||||
def close(self):
|
||||
self.logger.record('closed', None)
|
||||
|
||||
|
||||
class StubLogger:
|
||||
def __init__(self):
|
||||
self.calls = []
|
||||
self.hooks = {}
|
||||
self.on_enqueue = lambda coroutine: None
|
||||
|
||||
def record(self, name, value):
|
||||
self.calls.append((name, value))
|
||||
|
||||
def names(self):
|
||||
return [name for name, _ in self.calls]
|
||||
|
||||
def hook(self, phase, value, call_type):
|
||||
self.record(phase + '_hook', call_type)
|
||||
return self.hooks.get(phase, lambda value: 'awaitable')(value)
|
||||
|
||||
def check_limits(self, arguments):
|
||||
self.record('check_limits', arguments)
|
||||
|
||||
def failure_handler(self, error, trace, start, end):
|
||||
self.record('failure_handler', error)
|
||||
|
||||
def async_failure_handler(self, error, trace, start, end):
|
||||
self.record('async_failure_handler', error)
|
||||
return 'awaitable'
|
||||
|
||||
def success_handler(self, response, start, end):
|
||||
self.record('success_handler', response)
|
||||
|
||||
def async_success_handler(self, response, start, end):
|
||||
self.record('async_success_handler', response)
|
||||
return StubCoroutine(self)
|
||||
|
||||
def handle_sync_success_callbacks_for_async_calls(self, response, start, end):
|
||||
self.record('sync_success_for_async_call', response)
|
||||
|
||||
|
||||
logger = StubLogger()
|
||||
";
|
||||
|
||||
/// A namespace with the stubs, `StubLogger` and a fresh `logger`, after `script` ran in it.
|
||||
pub(crate) fn namespace<'py>(py: Python<'py>, script: &CStr) -> Bound<'py, PyDict> {
|
||||
let locals = PyDict::new(py);
|
||||
locals.set_item("python_contract", PYTHON_CONTRACT).unwrap();
|
||||
py.run(STUBS, Some(&locals), Some(&locals)).unwrap();
|
||||
py.run(script, Some(&locals), Some(&locals)).unwrap();
|
||||
locals
|
||||
}
|
||||
|
||||
pub(crate) fn run(py: Python<'_>, locals: &Bound<'_, PyDict>, code: &CStr) {
|
||||
py.run(code, Some(locals), Some(locals)).unwrap();
|
||||
}
|
||||
|
||||
pub(crate) fn local<'py>(locals: &Bound<'py, PyDict>, name: &str) -> Bound<'py, PyAny> {
|
||||
locals.get_item(name).unwrap().unwrap()
|
||||
}
|
||||
|
||||
/// A legacy call over the namespace's `kwargs` (or none) and `request` (or `None`).
|
||||
pub(crate) fn legacy_call(
|
||||
py: Python<'_>,
|
||||
locals: &Bound<'_, PyDict>,
|
||||
asynchronous: bool,
|
||||
) -> LegacyLogging {
|
||||
let request = locals
|
||||
.get_item("request")
|
||||
.unwrap()
|
||||
.unwrap_or_else(|| py.None().into_bound(py));
|
||||
let kwargs = locals
|
||||
.get_item("kwargs")
|
||||
.unwrap()
|
||||
.map(|kwargs| kwargs.cast_into::<PyDict>().unwrap())
|
||||
.unwrap_or_else(|| PyDict::new(py));
|
||||
let call = PublicCall::capture(&request, &PyTuple::empty(py), &kwargs).unwrap();
|
||||
LegacyLogging::new(
|
||||
py,
|
||||
LegacySurface {
|
||||
call_type: "test",
|
||||
input_description: "test input",
|
||||
stream: None,
|
||||
},
|
||||
call,
|
||||
asynchronous,
|
||||
)
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -1,146 +0,0 @@
|
|||
use std::ffi::CStr;
|
||||
|
||||
use pyo3::prelude::*;
|
||||
use pyo3::types::PyDict;
|
||||
use rstest::rstest;
|
||||
|
||||
use super::{PendingLogging, PendingSuccess};
|
||||
use crate::PythonLogger;
|
||||
use crate::test_support::{local, namespace, run};
|
||||
|
||||
/// A deferred success for the namespace's `logger` and `response`, bound as `pending`.
|
||||
fn defer<'py>(py: Python<'py>, script: &CStr) -> Bound<'py, PyDict> {
|
||||
let locals = namespace(py, c"response = object()");
|
||||
run(py, &locals, script);
|
||||
let pending = Py::new(
|
||||
py,
|
||||
PendingLogging {
|
||||
pending: Some(PendingSuccess {
|
||||
logger: PythonLogger::new(local(&locals, "logger").unbind()),
|
||||
response: Some(local(&locals, "response").unbind()),
|
||||
start: py.None(),
|
||||
end: Some(py.None()),
|
||||
}),
|
||||
},
|
||||
)
|
||||
.unwrap();
|
||||
locals.set_item("pending", pending).unwrap();
|
||||
locals
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn release_enqueues_the_success_once_in_the_releasing_context() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = defer(
|
||||
py,
|
||||
c"
|
||||
from contextvars import ContextVar
|
||||
|
||||
marker = ContextVar('marker', default='unset')
|
||||
observed = []
|
||||
|
||||
def on_enqueue(coroutine):
|
||||
observed.append(marker.get())
|
||||
pending.release(True)
|
||||
|
||||
logger.on_enqueue = on_enqueue
|
||||
",
|
||||
);
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
marker.set('release')
|
||||
pending.release(True)
|
||||
pending.release(True)
|
||||
assert observed == ['release'], observed
|
||||
assert logger.names() == ['async_success_handler', 'enqueued'], logger.calls
|
||||
assert logger.calls[0][1] is response
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_blocked_release_drops_the_success_for_good() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = defer(py, c"");
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
pending.release(False)
|
||||
pending.release(True)
|
||||
assert logger.calls == [], logger.calls
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::ordinary_error(c"RuntimeError('queue full')", false)]
|
||||
#[case::cancellation(c"asyncio.CancelledError()", true)]
|
||||
fn a_failed_enqueue_closes_the_coroutine_and_is_never_replayed(
|
||||
#[case] failure: &CStr,
|
||||
#[case] propagates: bool,
|
||||
) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = defer(
|
||||
py,
|
||||
c"
|
||||
import asyncio
|
||||
|
||||
def on_enqueue(coroutine):
|
||||
raise failure
|
||||
|
||||
logger.on_enqueue = on_enqueue
|
||||
",
|
||||
);
|
||||
locals
|
||||
.set_item("failure", py.eval(failure, None, Some(&locals)).unwrap())
|
||||
.unwrap();
|
||||
let released = local(&locals, "pending").call_method1("release", (true,));
|
||||
match released {
|
||||
Ok(_) => assert!(!propagates),
|
||||
Err(error) => {
|
||||
assert!(propagates);
|
||||
assert!(error.value(py).is(local(&locals, "failure")));
|
||||
}
|
||||
}
|
||||
locals.set_item("propagates", propagates).unwrap();
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
pending.release(True)
|
||||
assert logger.names() == ['async_success_handler', 'enqueued', 'closed'], logger.calls
|
||||
assert unraisable_from(logger) == ([] if propagates else [failure])
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn an_unreleased_success_does_not_keep_its_logger_alive() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = defer(py, c"");
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
import gc
|
||||
import weakref
|
||||
|
||||
logger.pending = pending
|
||||
reference = weakref.ref(logger)
|
||||
del logger, pending
|
||||
gc.collect()
|
||||
assert reference() is None
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
|
@ -1,282 +0,0 @@
|
|||
use std::ffi::CStr;
|
||||
|
||||
use litellm_host::event::{FailureOrigin, Timing};
|
||||
use litellm_host_python::{LifecycleEvent, LifecycleStep, PythonLifecycle};
|
||||
use pyo3::exceptions::asyncio::CancelledError;
|
||||
use pyo3::prelude::*;
|
||||
use pyo3::types::PyDict;
|
||||
use rstest::rstest;
|
||||
|
||||
use super::LegacyLogging;
|
||||
use crate::test_support::{legacy_call, local, namespace, run};
|
||||
|
||||
const CALL: &CStr = c"
|
||||
document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
|
||||
kwargs = {'logger': logger, 'document': document}
|
||||
";
|
||||
|
||||
const TIMING: Timing = Timing {
|
||||
start_time: 0.0,
|
||||
end_time: 1.0,
|
||||
};
|
||||
|
||||
fn begin<'py>(
|
||||
py: Python<'py>,
|
||||
locals: &Bound<'py, PyDict>,
|
||||
asynchronous: bool,
|
||||
) -> (LegacyLogging, LifecycleStep) {
|
||||
let mut logging = legacy_call(py, locals, asynchronous);
|
||||
let kwargs = local(locals, "kwargs")
|
||||
.cast_into::<PyDict>()
|
||||
.unwrap()
|
||||
.unbind();
|
||||
let step = logging.begin(py, kwargs, 0.0).unwrap();
|
||||
(logging, step)
|
||||
}
|
||||
|
||||
fn arguments<'py>(py: Python<'py>, step: LifecycleStep) -> Bound<'py, PyDict> {
|
||||
let LifecycleStep::Arguments(arguments) = step else {
|
||||
panic!("expected the prepared arguments");
|
||||
};
|
||||
arguments.into_bound(py)
|
||||
}
|
||||
|
||||
fn awaits_deployment_hook(step: &LifecycleStep) -> bool {
|
||||
matches!(step, LifecycleStep::Await(_))
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::synchronous(false)]
|
||||
#[case::asynchronous(true)]
|
||||
fn deployment_pre_call_hook_runs_only_for_asynchronous_calls(#[case] asynchronous: bool) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(py, CALL);
|
||||
let (_, step) = begin(py, &locals, asynchronous);
|
||||
assert_eq!(awaits_deployment_hook(&step), asynchronous);
|
||||
let names: Vec<String> = local(&locals, "logger")
|
||||
.call_method0("names")
|
||||
.unwrap()
|
||||
.extract()
|
||||
.unwrap();
|
||||
assert_eq!(names.contains(&"pre_hook".to_string()), asynchronous);
|
||||
});
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn kwargs_returned_by_the_pre_call_hook_are_what_the_call_prepares() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(
|
||||
py,
|
||||
c"
|
||||
document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
|
||||
replacement = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,ZWRpdGVk'}
|
||||
kwargs = {'logger': logger, 'document': document}
|
||||
replaced_kwargs = {'logger': logger, 'document': replacement, 'pages': [0]}
|
||||
",
|
||||
);
|
||||
let (mut logging, step) = begin(py, &locals, true);
|
||||
assert!(awaits_deployment_hook(&step));
|
||||
let step = logging
|
||||
.resume(py, Ok(local(&locals, "replaced_kwargs").unbind()))
|
||||
.unwrap();
|
||||
locals.set_item("prepared", arguments(py, step)).unwrap();
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
assert prepared['document'] is replacement
|
||||
assert prepared['pages'] is replaced_kwargs['pages']
|
||||
assert prepared['litellm_logging_obj'] is logger
|
||||
assert 'litellm_logging_obj' not in replaced_kwargs
|
||||
[checked] = [value for name, value in logger.calls if name == 'check_limits']
|
||||
assert checked is prepared
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::synchronous(false)]
|
||||
#[case::asynchronous(true)]
|
||||
fn a_keyword_the_bridge_never_reads_reaches_every_reader_as_the_callers_object(
|
||||
#[case] asynchronous: bool,
|
||||
) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(
|
||||
py,
|
||||
c"
|
||||
opaque = object()
|
||||
hooked = []
|
||||
logger.hooks = {'pre': lambda kwargs: hooked.append(kwargs['vendor_extension']) or kwargs}
|
||||
kwargs = {'logger': logger, 'vendor_extension': opaque}
|
||||
",
|
||||
);
|
||||
let (mut logging, step) = begin(py, &locals, asynchronous);
|
||||
let step = match step {
|
||||
LifecycleStep::Await(hook_result) => logging.resume(py, Ok(hook_result)).unwrap(),
|
||||
step => step,
|
||||
};
|
||||
locals.set_item("prepared", arguments(py, step)).unwrap();
|
||||
locals.set_item("asynchronous", asynchronous).unwrap();
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
assert prepared['vendor_extension'] is opaque
|
||||
[checked] = [value for name, value in logger.calls if name == 'check_limits']
|
||||
assert checked['vendor_extension'] is opaque
|
||||
assert hooked == ([opaque] if asynchronous else []), hooked
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn response_returned_by_the_post_call_hook_is_finalized_and_returned() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(
|
||||
py,
|
||||
c"
|
||||
kwargs = {'logger': logger}
|
||||
response = object()
|
||||
replacement = object()
|
||||
logger.hooks = {'pre': lambda kwargs: kwargs}
|
||||
",
|
||||
);
|
||||
let (mut logging, _) = begin(py, &locals, true);
|
||||
logging
|
||||
.resume(py, Ok(local(&locals, "kwargs").unbind()))
|
||||
.unwrap();
|
||||
let step = logging
|
||||
.after_success(py, local(&locals, "response").unbind(), TIMING)
|
||||
.unwrap();
|
||||
assert!(awaits_deployment_hook(&step));
|
||||
let step = logging
|
||||
.resume(py, Ok(local(&locals, "replacement").unbind()))
|
||||
.unwrap();
|
||||
let LifecycleStep::Response(returned) = step else {
|
||||
panic!("expected the finalized response");
|
||||
};
|
||||
assert!(returned.bind(py).is(local(&locals, "replacement")));
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
[finalized] = [value for name, value in logger.calls if name == 'finalize']
|
||||
assert finalized is replacement
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::pre_call(false)]
|
||||
#[case::post_call(true)]
|
||||
fn cancelling_a_deployment_hook_ends_the_call_with_that_cancellation(#[case] post_call: bool) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(py, c"kwargs = {'logger': logger}\nresponse = object()");
|
||||
let (mut logging, _) = begin(py, &locals, true);
|
||||
if post_call {
|
||||
logging
|
||||
.resume(py, Ok(local(&locals, "kwargs").unbind()))
|
||||
.unwrap();
|
||||
logging
|
||||
.after_success(py, local(&locals, "response").unbind(), TIMING)
|
||||
.unwrap();
|
||||
}
|
||||
let cancellation = CancelledError::new_err("cancelled");
|
||||
let cancelled = cancellation.value(py).clone();
|
||||
let error = logging.resume(py, Err(cancellation)).err().unwrap();
|
||||
assert!(error.value(py).is(&cancelled));
|
||||
let names: Vec<String> = local(&locals, "logger")
|
||||
.call_method0("names")
|
||||
.unwrap()
|
||||
.extract()
|
||||
.unwrap();
|
||||
assert!(!names.iter().any(|name| name.contains("handler")));
|
||||
});
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::hook_completed(false)]
|
||||
#[case::hook_cancelled(true)]
|
||||
fn failure_callbacks_run_after_the_failure_hook_however_it_ends(#[case] cancelled: bool) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(
|
||||
py,
|
||||
c"kwargs = {'logger': logger}\nfailure = ValueError('provider')",
|
||||
);
|
||||
let (mut logging, _) = begin(py, &locals, true);
|
||||
logging
|
||||
.resume(py, Ok(local(&locals, "kwargs").unbind()))
|
||||
.unwrap();
|
||||
let failure = PyErr::from_value(local(&locals, "failure"));
|
||||
let failed = LifecycleEvent::Failed {
|
||||
timing: TIMING,
|
||||
origin: FailureOrigin::Call,
|
||||
error: &failure,
|
||||
};
|
||||
let step = logging.emit(py, failed).unwrap();
|
||||
assert!(awaits_deployment_hook(&step));
|
||||
let hook_result = if cancelled {
|
||||
Err(CancelledError::new_err("cancelled"))
|
||||
} else {
|
||||
Ok(py.None())
|
||||
};
|
||||
assert!(matches!(
|
||||
logging.resume(py, hook_result).unwrap(),
|
||||
LifecycleStep::Await(_)
|
||||
));
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
assert logger.names()[-3:] == ['failure_hook', 'failure_handler', 'async_failure_handler'], logger.calls
|
||||
assert all(value is failure for name, value in logger.calls if name.endswith('_handler'))
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::synchronous(false)]
|
||||
#[case::asynchronous(true)]
|
||||
fn a_limit_rejected_before_the_call_surfaces_as_the_callers_error(#[case] asynchronous: bool) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(
|
||||
py,
|
||||
c"
|
||||
class BudgetExceeded(Exception):
|
||||
pass
|
||||
|
||||
rejection = BudgetExceeded('over budget')
|
||||
|
||||
class LimitedLogger(StubLogger):
|
||||
def check_limits(self, arguments):
|
||||
raise rejection
|
||||
|
||||
logger = LimitedLogger()
|
||||
logger.hooks = {'pre': lambda kwargs: kwargs}
|
||||
kwargs = {'logger': logger}
|
||||
",
|
||||
);
|
||||
let mut logging = legacy_call(py, &locals, asynchronous);
|
||||
let kwargs = local(&locals, "kwargs")
|
||||
.cast_into::<PyDict>()
|
||||
.unwrap()
|
||||
.unbind();
|
||||
let result = logging.begin(py, kwargs, 0.0).and_then(|step| match step {
|
||||
LifecycleStep::Await(_) => logging.resume(py, Ok(local(&locals, "kwargs").unbind())),
|
||||
step => Ok(step),
|
||||
});
|
||||
let error = result.err().unwrap();
|
||||
assert!(error.value(py).is(local(&locals, "rejection")));
|
||||
});
|
||||
}
|
||||
|
|
@ -1,523 +0,0 @@
|
|||
use std::ffi::CStr;
|
||||
|
||||
use litellm_auth::SecretValue;
|
||||
use litellm_host::event::{MachineEvent, RawResponse, RequestContext, WireRequest};
|
||||
use litellm_host_python::{LifecycleEvent, LifecycleStep, PythonLifecycle, to_py};
|
||||
use proptest::prelude::*;
|
||||
use pyo3::prelude::*;
|
||||
use rstest::rstest;
|
||||
use serde_json::{Map, Value, json};
|
||||
|
||||
use super::LegacyLogging;
|
||||
use crate::PythonLogger;
|
||||
use crate::test_support::{legacy_call, local, namespace, run};
|
||||
|
||||
/// The payload phases of `Logging` on top of `StubLogger`, with `pre_call` handing the
|
||||
/// payload to the case's `on_pre_call`.
|
||||
const PAYLOAD_LOGGER: &CStr = c"
|
||||
class Request:
|
||||
pass
|
||||
|
||||
class PayloadLogger(StubLogger):
|
||||
def update_from_kwargs(self, **update):
|
||||
self.update = update
|
||||
|
||||
def pre_call(self, input, api_key, additional_args):
|
||||
self.record('pre_call', None)
|
||||
self.pre = additional_args
|
||||
self.pre_api_key = api_key
|
||||
on_pre_call(additional_args)
|
||||
|
||||
def post_call(self, original_response, api_key, additional_args):
|
||||
self.record('post_call', None)
|
||||
self.post = (original_response, api_key, additional_args)
|
||||
|
||||
request = Request()
|
||||
kwargs = {}
|
||||
logger = PayloadLogger()
|
||||
on_pre_call = lambda additional_args: None
|
||||
check = lambda: None
|
||||
";
|
||||
|
||||
const DOCUMENT: &str = "data:application/pdf;base64,YWJj";
|
||||
const EDITED: &str = "data:application/pdf;base64,ZWRpdGVk";
|
||||
|
||||
fn document(source: &str) -> Value {
|
||||
json!({"type": "document_url", "document_url": source})
|
||||
}
|
||||
|
||||
fn before_send(script: &CStr, body: Value) -> WireRequest {
|
||||
before_send_with_secrets(script, json!({}), body, &[])
|
||||
}
|
||||
|
||||
/// Runs `before_send` over `body` for a route whose parameters are `optional_params`, with
|
||||
/// the Python objects `script` binds, then delivers the provider's raw response the way the
|
||||
/// driver does and runs the script's `check()`.
|
||||
fn before_send_with_secrets(
|
||||
script: &CStr,
|
||||
optional_params: Value,
|
||||
body: Value,
|
||||
secret_fields: &[&str],
|
||||
) -> WireRequest {
|
||||
before_send_bound(&[], script, optional_params, body, secret_fields)
|
||||
}
|
||||
|
||||
/// [`before_send_with_secrets`] with `bindings` placed in the namespace before `script` runs.
|
||||
fn before_send_bound(
|
||||
bindings: &[(&str, &Value)],
|
||||
script: &CStr,
|
||||
optional_params: Value,
|
||||
body: Value,
|
||||
secret_fields: &[&str],
|
||||
) -> WireRequest {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(py, PAYLOAD_LOGGER);
|
||||
for &(name, value) in bindings {
|
||||
locals.set_item(name, to_py(py, value).unwrap()).unwrap();
|
||||
}
|
||||
run(py, &locals, script);
|
||||
let mut logging = LegacyLogging {
|
||||
logger: Some(PythonLogger::new(local(&locals, "logger").unbind())),
|
||||
..legacy_call(py, &locals, false)
|
||||
};
|
||||
let context = RequestContext {
|
||||
model: "model".into(),
|
||||
custom_llm_provider: "provider".into(),
|
||||
optional_params,
|
||||
secret_fields: secret_fields.iter().map(|name| name.to_string()).collect(),
|
||||
api_key: Some(SecretValue::new("route-key")),
|
||||
};
|
||||
let wire = WireRequest {
|
||||
url: "https://provider.invalid/ocr".into(),
|
||||
headers: vec![("x-route".into(), "route".into())],
|
||||
body,
|
||||
};
|
||||
let step = logging.before_send(py, Box::new(wire), &context).unwrap();
|
||||
let raw = MachineEvent::ResponseReceived {
|
||||
raw: RawResponse {
|
||||
body: "raw response".into(),
|
||||
},
|
||||
};
|
||||
assert!(matches!(
|
||||
logging.emit(py, LifecycleEvent::Machine(&raw)).unwrap(),
|
||||
LifecycleStep::Done
|
||||
));
|
||||
run(py, &locals, c"check()");
|
||||
let LifecycleStep::Wire(wire) = step else {
|
||||
panic!("before_send did not hand back the wire request");
|
||||
};
|
||||
*wire
|
||||
})
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::caller_keyword(c"
|
||||
document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
|
||||
pages = [0]
|
||||
kwargs = {'document': document, 'pages': pages}
|
||||
observed = []
|
||||
on_pre_call = lambda args: observed.append(
|
||||
(args['complete_input_dict']['document'] is document, args['complete_input_dict']['pages'] is pages)
|
||||
)
|
||||
def check():
|
||||
assert observed == [(True, True)], observed
|
||||
")]
|
||||
#[case::request_attribute_behind_an_omitted_keyword(c"
|
||||
document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
|
||||
pages = [0]
|
||||
request.document = document
|
||||
kwargs = {'pages': pages}
|
||||
observed = []
|
||||
on_pre_call = lambda args: observed.append(
|
||||
(args['complete_input_dict']['document'] is document, args['complete_input_dict']['pages'] is pages)
|
||||
)
|
||||
def check():
|
||||
assert observed == [(True, True)], observed
|
||||
")]
|
||||
fn passthrough_keys_reach_pre_call_as_the_callers_own_objects(#[case] script: &CStr) {
|
||||
let body = json!({"model": "model", "document": document(DOCUMENT), "pages": [0]});
|
||||
let wire = before_send(script, body.clone());
|
||||
assert_eq!(wire.body, body);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn pre_call_edit_of_a_passthrough_object_reaches_the_caller_and_the_wire() {
|
||||
let wire = before_send(
|
||||
c"
|
||||
document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
|
||||
kwargs = {'document': document}
|
||||
def on_pre_call(args):
|
||||
args['complete_input_dict']['document']['document_url'] = 'data:application/pdf;base64,ZWRpdGVk'
|
||||
def check():
|
||||
assert document['document_url'] == 'data:application/pdf;base64,ZWRpdGVk'
|
||||
",
|
||||
json!({"document": document(DOCUMENT)}),
|
||||
);
|
||||
assert_eq!(wire.body["document"], document(EDITED));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_body_key_the_route_rewrote_is_not_the_callers_object() {
|
||||
let wire = before_send(
|
||||
c"
|
||||
document = {'type': 'document_url', 'document_url': 'https://example.invalid/scan.pdf'}
|
||||
kwargs = {'document': document}
|
||||
observed = []
|
||||
def on_pre_call(args):
|
||||
observed.append(args['complete_input_dict']['document'] is document)
|
||||
args['complete_input_dict']['document']['document_name'] = 'edited.pdf'
|
||||
def check():
|
||||
assert observed == [False], observed
|
||||
assert document == {'type': 'document_url', 'document_url': 'https://example.invalid/scan.pdf'}
|
||||
",
|
||||
json!({"document": document(DOCUMENT)}),
|
||||
);
|
||||
assert_eq!(
|
||||
wire.body["document"],
|
||||
json!({"type": "document_url", "document_url": DOCUMENT, "document_name": "edited.pdf"})
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_caller_value_with_no_json_form_is_left_out_of_realiasing() {
|
||||
let body = json!({"pages": [0]});
|
||||
let wire = before_send(
|
||||
c"
|
||||
opaque = object()
|
||||
kwargs = {'pages': opaque}
|
||||
observed = []
|
||||
on_pre_call = lambda args: observed.append(args['complete_input_dict']['pages'])
|
||||
def check():
|
||||
assert observed == [[0]], observed
|
||||
",
|
||||
body.clone(),
|
||||
);
|
||||
assert_eq!(wire.body, body);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::body(
|
||||
c"
|
||||
def on_pre_call(args):
|
||||
args['complete_input_dict'] = {'replacement': True}
|
||||
"
|
||||
)]
|
||||
#[case::headers(
|
||||
c"
|
||||
def on_pre_call(args):
|
||||
args['headers'] = {'x-replacement': 'yes'}
|
||||
"
|
||||
)]
|
||||
fn rebinding_the_payload_envelope_does_not_reach_the_wire(#[case] script: &CStr) {
|
||||
let body = json!({"document": document(DOCUMENT)});
|
||||
let wire = before_send(script, body.clone());
|
||||
assert_eq!(wire.body, body);
|
||||
assert_eq!(wire.headers, [("x-route".to_string(), "route".to_string())]);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn pre_call_header_edit_reaches_the_wire() {
|
||||
let wire = before_send(
|
||||
c"
|
||||
def on_pre_call(args):
|
||||
args['headers']['x-callback'] = 'edited'
|
||||
",
|
||||
json!({}),
|
||||
);
|
||||
assert_eq!(
|
||||
wire.headers,
|
||||
[
|
||||
("x-route".to_string(), "route".to_string()),
|
||||
("x-callback".to_string(), "edited".to_string()),
|
||||
]
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn pre_call_receives_the_wire_request_and_the_logger_its_redacted_request() {
|
||||
let body = json!({"model": "model", "document": document(DOCUMENT)});
|
||||
before_send_with_secrets(
|
||||
c"
|
||||
logger_fn = lambda *args: None
|
||||
kwargs = {
|
||||
'litellm_call_id': 'call-1',
|
||||
'client_secret': 'shh',
|
||||
'proxy_server_request': {'body': {}},
|
||||
'logger_fn': logger_fn,
|
||||
'litellm_request_debug': True,
|
||||
'ocr_cost_per_page': 0.05,
|
||||
}
|
||||
observed = []
|
||||
on_pre_call = observed.append
|
||||
def check():
|
||||
[args] = observed
|
||||
assert args['api_base'] == 'https://provider.invalid/ocr', args
|
||||
assert args['complete_input_dict'] == {
|
||||
'model': 'model',
|
||||
'document': {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'},
|
||||
}, args
|
||||
update = logger.update
|
||||
assert update['model'] == 'model' and update['custom_llm_provider'] == 'provider', update
|
||||
assert update['litellm_params']['litellm_call_id'] == 'call-1', update
|
||||
assert update['litellm_params']['api_base'] == 'https://provider.invalid/ocr', update
|
||||
assert update['litellm_params']['logger_fn'] is logger_fn, update
|
||||
assert update['litellm_params']['litellm_request_debug'] is True, update
|
||||
assert update['litellm_params']['ocr_cost_per_page'] == 0.05, update
|
||||
assert update['kwargs']['client_secret'] == '****', update
|
||||
assert 'proxy_server_request' not in update['kwargs'], update
|
||||
assert update['optional_params']['client_secret'] == '****', update
|
||||
",
|
||||
json!({"client_secret": "shh"}),
|
||||
body,
|
||||
&["client_secret"],
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::added_key(
|
||||
c"
|
||||
def on_pre_call(args):
|
||||
args['complete_input_dict']['include_image_base64'] = True
|
||||
",
|
||||
json!({"document": document(DOCUMENT), "include_image_base64": true})
|
||||
)]
|
||||
#[case::replaced_document(
|
||||
c"
|
||||
document = {'type': 'document_url', 'document_url': 'data:application/pdf;base64,YWJj'}
|
||||
kwargs = {'document': document}
|
||||
def on_pre_call(args):
|
||||
args['complete_input_dict']['document'] = {
|
||||
'type': 'document_url', 'document_url': 'data:application/pdf;base64,ZWRpdGVk'
|
||||
}
|
||||
def check():
|
||||
assert document['document_url'] == 'data:application/pdf;base64,YWJj', document
|
||||
",
|
||||
json!({"document": document(EDITED)})
|
||||
)]
|
||||
#[case::retained_body_edited_after_rebinding(
|
||||
c"
|
||||
def on_pre_call(args):
|
||||
retained = args['complete_input_dict']
|
||||
args['complete_input_dict'] = {'rebound': True}
|
||||
retained['include_image_base64'] = True
|
||||
",
|
||||
json!({"document": document(DOCUMENT), "include_image_base64": true})
|
||||
)]
|
||||
fn pre_call_body_edits_reach_the_wire(#[case] script: &CStr, #[case] expected: Value) {
|
||||
let body = json!({"document": document(DOCUMENT)});
|
||||
let wire = before_send(script, body);
|
||||
assert_eq!(wire.body, expected);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn retained_headers_edited_after_rebinding_reach_the_wire() {
|
||||
let wire = before_send(
|
||||
c"
|
||||
def on_pre_call(args):
|
||||
retained = args['headers']
|
||||
args['headers'] = {'x-rebound': 'rebound'}
|
||||
retained['x-retained'] = 'sent'
|
||||
",
|
||||
json!({}),
|
||||
);
|
||||
assert_eq!(
|
||||
wire.headers,
|
||||
[
|
||||
("x-route".to_string(), "route".to_string()),
|
||||
("x-retained".to_string(), "sent".to_string()),
|
||||
]
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn post_call_receives_the_raw_response_the_route_key_and_the_body_and_headers_pre_call_saw() {
|
||||
before_send(
|
||||
c"
|
||||
def check():
|
||||
original_response, api_key, additional_args = logger.post
|
||||
assert original_response == 'raw response', original_response
|
||||
assert api_key == logger.pre_api_key == 'route-key', (api_key, logger.pre_api_key)
|
||||
assert additional_args == {
|
||||
'complete_input_dict': logger.pre['complete_input_dict'],
|
||||
'headers': logger.pre['headers'],
|
||||
}, additional_args
|
||||
assert additional_args['complete_input_dict'] is logger.pre['complete_input_dict']
|
||||
assert additional_args['headers'] is logger.pre['headers']
|
||||
",
|
||||
json!({"document": document(DOCUMENT)}),
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn every_request_runs_the_full_pre_call_and_post_call() {
|
||||
let wire = before_send(
|
||||
c"
|
||||
def on_pre_call(args):
|
||||
args['complete_input_dict']['include_image_base64'] = True
|
||||
def check():
|
||||
assert logger.names() == ['pre_call', 'post_call'], logger.calls
|
||||
",
|
||||
json!({"document": document(DOCUMENT)}),
|
||||
);
|
||||
assert_eq!(
|
||||
wire.body,
|
||||
json!({"document": document(DOCUMENT), "include_image_base64": true})
|
||||
);
|
||||
}
|
||||
|
||||
/// What one pre-call callback does to the payload it is handed.
|
||||
#[derive(Clone, Debug)]
|
||||
enum Edit {
|
||||
Nothing,
|
||||
Set(String, Value),
|
||||
Remove(String),
|
||||
Rebind(Value),
|
||||
RebindThenSetRetained(String, Value),
|
||||
}
|
||||
|
||||
impl Edit {
|
||||
fn script(&self) -> Value {
|
||||
match self {
|
||||
Self::Nothing => json!({"kind": "nothing"}),
|
||||
Self::Set(key, value) => json!({"kind": "set", "key": key, "value": value}),
|
||||
Self::Remove(key) => json!({"kind": "remove", "key": key}),
|
||||
Self::Rebind(value) => json!({"kind": "rebind", "value": value}),
|
||||
Self::RebindThenSetRetained(key, value) => {
|
||||
json!({"kind": "rebind_then_set_retained", "key": key, "value": value})
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// The legacy contract: the provider is sent the body object `pre_call` received, as
|
||||
/// the callback left it. Rebinding the envelope's key points the envelope elsewhere and
|
||||
/// leaves that object alone.
|
||||
fn sent(&self, body: &Map<String, Value>) -> Value {
|
||||
let mut sent = body.clone();
|
||||
match self {
|
||||
Self::Nothing | Self::Rebind(_) => {}
|
||||
Self::Set(key, value) | Self::RebindThenSetRetained(key, value) => {
|
||||
sent.insert(key.clone(), value.clone());
|
||||
}
|
||||
Self::Remove(key) => {
|
||||
sent.remove(key);
|
||||
}
|
||||
}
|
||||
Value::Object(sent)
|
||||
}
|
||||
}
|
||||
|
||||
/// How the caller's keyword for a body key relates to what the route sends under it.
|
||||
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
|
||||
enum Caller {
|
||||
PassedUnchanged,
|
||||
RewrittenByTheRoute,
|
||||
NotPassed,
|
||||
}
|
||||
|
||||
const MODEL: &CStr = c"
|
||||
aliased = {}
|
||||
def on_pre_call(args):
|
||||
body = args['complete_input_dict']
|
||||
aliased.update({name: body[name] is kwargs[name] for name in unchanged})
|
||||
kind = edit['kind']
|
||||
if kind == 'set':
|
||||
body[edit['key']] = edit['value']
|
||||
elif kind == 'remove':
|
||||
body.pop(edit['key'], None)
|
||||
elif kind == 'rebind':
|
||||
args['complete_input_dict'] = edit['value']
|
||||
elif kind == 'rebind_then_set_retained':
|
||||
args['complete_input_dict'] = {}
|
||||
body[edit['key']] = edit['value']
|
||||
def check():
|
||||
assert aliased == {name: True for name in unchanged}, aliased
|
||||
assert logger.names() == ['pre_call', 'post_call'], logger.calls
|
||||
";
|
||||
|
||||
fn json_value() -> impl Strategy<Value = Value> {
|
||||
let leaf = prop_oneof![
|
||||
Just(Value::Null),
|
||||
any::<bool>().prop_map(Value::from),
|
||||
any::<i64>().prop_map(Value::from),
|
||||
any::<f64>()
|
||||
.prop_filter("JSON has no NaN or infinity", |number| number.is_finite())
|
||||
.prop_map(Value::from),
|
||||
".{0,8}".prop_map(Value::from),
|
||||
];
|
||||
leaf.prop_recursive(3, 24, 4, |inner| {
|
||||
prop_oneof![
|
||||
prop::collection::vec(inner.clone(), 0..4).prop_map(Value::from),
|
||||
prop::collection::btree_map(key(), inner, 0..4)
|
||||
.prop_map(|fields| Value::Object(fields.into_iter().collect())),
|
||||
]
|
||||
})
|
||||
}
|
||||
|
||||
fn key() -> impl Strategy<Value = String> {
|
||||
"[a-z]{1,6}"
|
||||
}
|
||||
|
||||
fn caller() -> impl Strategy<Value = Caller> {
|
||||
prop_oneof![
|
||||
Just(Caller::PassedUnchanged),
|
||||
Just(Caller::RewrittenByTheRoute),
|
||||
Just(Caller::NotPassed),
|
||||
]
|
||||
}
|
||||
|
||||
fn edit() -> impl Strategy<Value = Edit> {
|
||||
prop_oneof![
|
||||
Just(Edit::Nothing),
|
||||
(key(), json_value()).prop_map(|(key, value)| Edit::Set(key, value)),
|
||||
key().prop_map(Edit::Remove),
|
||||
json_value().prop_map(Edit::Rebind),
|
||||
(key(), json_value()).prop_map(|(key, value)| Edit::RebindThenSetRetained(key, value)),
|
||||
]
|
||||
}
|
||||
|
||||
proptest! {
|
||||
#![proptest_config(ProptestConfig::with_cases(128))]
|
||||
|
||||
/// For any body, any caller keywords and any callback edit: every keyword the route
|
||||
/// sends unchanged reaches `pre_call` as the caller's own object, and the provider is
|
||||
/// sent exactly what the model says, so a callback that edits nothing changes nothing.
|
||||
#[test]
|
||||
fn the_wire_is_the_body_pre_call_received_as_the_callback_left_it(
|
||||
fields in prop::collection::btree_map(key(), (json_value(), caller()), 0..5),
|
||||
edit in edit(),
|
||||
) {
|
||||
let body: Map<String, Value> = fields
|
||||
.iter()
|
||||
.map(|(name, (value, _))| (name.clone(), value.clone()))
|
||||
.collect();
|
||||
let kwargs: Map<String, Value> = fields
|
||||
.iter()
|
||||
.filter_map(|(name, (value, caller))| match caller {
|
||||
Caller::PassedUnchanged => Some((name.clone(), value.clone())),
|
||||
Caller::RewrittenByTheRoute => Some((name.clone(), json!([value]))),
|
||||
Caller::NotPassed => None,
|
||||
})
|
||||
.collect();
|
||||
let unchanged: Value = fields
|
||||
.iter()
|
||||
.filter(|(_, (_, caller))| *caller == Caller::PassedUnchanged)
|
||||
.map(|(name, _)| Value::from(name.clone()))
|
||||
.collect();
|
||||
|
||||
let wire = before_send_bound(
|
||||
&[
|
||||
("kwargs", &Value::Object(kwargs)),
|
||||
("unchanged", &unchanged),
|
||||
("edit", &edit.script()),
|
||||
],
|
||||
MODEL,
|
||||
json!({}),
|
||||
Value::Object(body.clone()),
|
||||
&[],
|
||||
);
|
||||
|
||||
prop_assert_eq!(wire.body, edit.sent(&body));
|
||||
prop_assert_eq!(wire.headers, [("x-route".to_string(), "route".to_string())]);
|
||||
}
|
||||
}
|
||||
|
|
@ -1,205 +0,0 @@
|
|||
use std::ffi::CStr;
|
||||
|
||||
use pyo3::prelude::*;
|
||||
use pyo3::types::{PyDict, PyTuple};
|
||||
|
||||
use crate::{LegacyLogging, LegacySurface, PublicCall};
|
||||
|
||||
/// The parameters of every `callbacks_legacy_python` function, as the real module declares them.
|
||||
/// `tests/test_litellm/rust_bridge/test_callbacks_legacy_python.py` pins this file to the Python
|
||||
/// signatures, and [`namespace`] binds every fake call against it.
|
||||
pub(crate) const PYTHON_CONTRACT: &str = include_str!("../python_contract.json");
|
||||
|
||||
/// Stand-ins for `callbacks_legacy_python`, the only Python module the crate calls. Tests
|
||||
/// share one interpreter and run concurrently, so each fake is installed idempotently and
|
||||
/// forwards to the per-test `StubLogger` it is handed (directly, or as `kwargs['logger']`).
|
||||
/// Every fake is bound against the contract first, so a call the real module would reject
|
||||
/// fails here too.
|
||||
const STUBS: &CStr = c"
|
||||
import contextvars
|
||||
import inspect
|
||||
import json
|
||||
import sys
|
||||
import traceback
|
||||
import types
|
||||
|
||||
for name in ('litellm', 'litellm.rust_bridge', 'litellm.rust_bridge.callbacks_legacy_python'):
|
||||
sys.modules.setdefault(name, types.ModuleType(name))
|
||||
|
||||
legacy = sys.modules['litellm.rust_bridge.callbacks_legacy_python']
|
||||
CONTRACT = json.loads(python_contract)
|
||||
|
||||
|
||||
def contracted(name, fake):
|
||||
signature = inspect.Signature(
|
||||
[inspect.Parameter(parameter, inspect.Parameter.POSITIONAL_OR_KEYWORD) for parameter in CONTRACT[name]]
|
||||
)
|
||||
|
||||
def checked(*args, **kwargs):
|
||||
signature.bind(*args, **kwargs)
|
||||
return fake(*args, **kwargs)
|
||||
|
||||
return checked
|
||||
|
||||
|
||||
if not hasattr(legacy, 'is_internal'):
|
||||
legacy.is_internal = contextvars.ContextVar('is_internal_call', default=False)
|
||||
|
||||
FAKES = {
|
||||
'setup': lambda call_type, args, kwargs, start, asynchronous: types.SimpleNamespace(
|
||||
logger=kwargs['logger_factory'](kwargs) if 'logger_factory' in kwargs else kwargs['logger'],
|
||||
kwargs=kwargs,
|
||||
),
|
||||
'check_limits': lambda arguments: arguments['logger'].check_limits(arguments),
|
||||
'finalize': lambda response, logger, kwargs, start, end: logger.record('finalize', response),
|
||||
'update_logging': lambda logger, kwargs, model, optional_params, litellm_params, provider: logger.update_from_kwargs(
|
||||
kwargs=kwargs,
|
||||
model=model,
|
||||
optional_params=optional_params,
|
||||
litellm_params=litellm_params,
|
||||
custom_llm_provider=provider,
|
||||
),
|
||||
'pre_call': lambda logger, input, api_key, additional_args: logger.pre_call(input, api_key, additional_args),
|
||||
'post_call': lambda logger, original_response, api_key, additional_args: logger.post_call(
|
||||
original_response, api_key, additional_args
|
||||
),
|
||||
'defers_async_logging': lambda logger: bool(getattr(logger, '_defer_async_logging', False)),
|
||||
'defer_success': lambda logger, pending: setattr(logger, '_native_pending_logging', pending),
|
||||
'sync_success_for_async_call': lambda logger, response, start, end: logger.handle_sync_success_callbacks_for_async_calls(
|
||||
response, start, end
|
||||
),
|
||||
'failure_handler': lambda logger, error, start, end, asynchronous: (
|
||||
logger.async_failure_handler if asynchronous else logger.failure_handler
|
||||
)(error, ''.join(traceback.format_exception(error)), start, end),
|
||||
'submit_success': lambda logger, response, start, end: logger.record('submit', (response, start, end)),
|
||||
'async_success_handler': lambda logger, response, start, end: logger.async_success_handler(response, start, end),
|
||||
'enqueue_logging': lambda coroutine: coroutine.enqueue(),
|
||||
'restore_context': lambda logger: logger.record('restore', None),
|
||||
'custom_pricing_fields': lambda: ('ocr_cost_per_page',),
|
||||
'is_internal_call': lambda: legacy.is_internal.get(),
|
||||
'credential_list': lambda: [],
|
||||
'warn_unknown_credential': lambda name, loaded: None,
|
||||
'before_deployment_call': lambda kwargs, call_type: kwargs['logger'].hook('pre', kwargs, call_type),
|
||||
'after_deployment_success': lambda kwargs, response, call_type: kwargs['logger'].hook(
|
||||
'success', response, call_type
|
||||
),
|
||||
'after_deployment_failure': lambda kwargs, error, call_type: kwargs['logger'].hook('failure', error, call_type),
|
||||
'stream_opened': lambda logger: logger.record('stream_opened', None),
|
||||
'stream_success': lambda logger, request_body, chunks, start, end, first_chunk: logger.record(
|
||||
'stream_success', list(chunks)
|
||||
),
|
||||
'stream_failure': lambda logger, request_body, chunks, error: logger.record('stream_failure', error),
|
||||
}
|
||||
assert FAKES.keys() == CONTRACT.keys(), sorted(FAKES.keys() ^ CONTRACT.keys())
|
||||
for name, fake in FAKES.items():
|
||||
setattr(legacy, name, contracted(name, fake))
|
||||
|
||||
|
||||
unraisable = sys.modules.setdefault(
|
||||
'litellm_test_unraisable', types.ModuleType('litellm_test_unraisable')
|
||||
)
|
||||
if not hasattr(unraisable, 'events'):
|
||||
unraisable.events = []
|
||||
sys.unraisablehook = lambda event: unraisable.events.append((event.object, event.exc_value))
|
||||
|
||||
|
||||
def unraisable_from(owner):
|
||||
return [error for source, error in unraisable.events if source is owner]
|
||||
|
||||
|
||||
class StubCoroutine:
|
||||
def __init__(self, logger):
|
||||
self.logger = logger
|
||||
|
||||
def enqueue(self):
|
||||
self.logger.record('enqueued', None)
|
||||
self.logger.on_enqueue(self)
|
||||
|
||||
def close(self):
|
||||
self.logger.record('closed', None)
|
||||
|
||||
|
||||
class StubLogger:
|
||||
def __init__(self):
|
||||
self.calls = []
|
||||
self.hooks = {}
|
||||
self.on_enqueue = lambda coroutine: None
|
||||
|
||||
def record(self, name, value):
|
||||
self.calls.append((name, value))
|
||||
|
||||
def names(self):
|
||||
return [name for name, _ in self.calls]
|
||||
|
||||
def hook(self, phase, value, call_type):
|
||||
self.record(phase + '_hook', call_type)
|
||||
return self.hooks.get(phase, lambda value: 'awaitable')(value)
|
||||
|
||||
def check_limits(self, arguments):
|
||||
self.record('check_limits', arguments)
|
||||
|
||||
def failure_handler(self, error, trace, start, end):
|
||||
self.record('failure_handler', error)
|
||||
|
||||
def async_failure_handler(self, error, trace, start, end):
|
||||
self.record('async_failure_handler', error)
|
||||
return 'awaitable'
|
||||
|
||||
def success_handler(self, response, start, end):
|
||||
self.record('success_handler', response)
|
||||
|
||||
def async_success_handler(self, response, start, end):
|
||||
self.record('async_success_handler', response)
|
||||
return StubCoroutine(self)
|
||||
|
||||
def handle_sync_success_callbacks_for_async_calls(self, response, start, end):
|
||||
self.record('sync_success_for_async_call', response)
|
||||
|
||||
|
||||
logger = StubLogger()
|
||||
";
|
||||
|
||||
/// A namespace with the stubs, `StubLogger` and a fresh `logger`, after `script` ran in it.
|
||||
pub(crate) fn namespace<'py>(py: Python<'py>, script: &CStr) -> Bound<'py, PyDict> {
|
||||
let locals = PyDict::new(py);
|
||||
locals.set_item("python_contract", PYTHON_CONTRACT).unwrap();
|
||||
py.run(STUBS, Some(&locals), Some(&locals)).unwrap();
|
||||
py.run(script, Some(&locals), Some(&locals)).unwrap();
|
||||
locals
|
||||
}
|
||||
|
||||
pub(crate) fn run(py: Python<'_>, locals: &Bound<'_, PyDict>, code: &CStr) {
|
||||
py.run(code, Some(locals), Some(locals)).unwrap();
|
||||
}
|
||||
|
||||
pub(crate) fn local<'py>(locals: &Bound<'py, PyDict>, name: &str) -> Bound<'py, PyAny> {
|
||||
locals.get_item(name).unwrap().unwrap()
|
||||
}
|
||||
|
||||
/// A legacy call over the namespace's `kwargs` (or none) and `request` (or `None`).
|
||||
pub(crate) fn legacy_call(
|
||||
py: Python<'_>,
|
||||
locals: &Bound<'_, PyDict>,
|
||||
asynchronous: bool,
|
||||
) -> LegacyLogging {
|
||||
let request = locals
|
||||
.get_item("request")
|
||||
.unwrap()
|
||||
.unwrap_or_else(|| py.None().into_bound(py));
|
||||
let kwargs = locals
|
||||
.get_item("kwargs")
|
||||
.unwrap()
|
||||
.map(|kwargs| kwargs.cast_into::<PyDict>().unwrap())
|
||||
.unwrap_or_else(|| PyDict::new(py));
|
||||
let call = PublicCall::capture(&request, &PyTuple::empty(py), &kwargs).unwrap();
|
||||
LegacyLogging::new(
|
||||
py,
|
||||
LegacySurface {
|
||||
call_type: "test",
|
||||
input_description: "test input",
|
||||
stream: None,
|
||||
},
|
||||
call,
|
||||
asynchronous,
|
||||
)
|
||||
}
|
||||
|
|
@ -1,291 +0,0 @@
|
|||
use std::ffi::CStr;
|
||||
|
||||
use litellm_host::event::{FailureOrigin, Timing};
|
||||
use litellm_host_python::{LifecycleEvent, LifecycleStep, PythonLifecycle};
|
||||
use pyo3::exceptions::PyRuntimeError;
|
||||
use pyo3::exceptions::asyncio::CancelledError;
|
||||
use pyo3::prelude::*;
|
||||
use pyo3::types::PyDict;
|
||||
use rstest::rstest;
|
||||
|
||||
use super::LegacyLogging;
|
||||
use crate::PythonLogger;
|
||||
use crate::test_support::{legacy_call, local, namespace, run};
|
||||
|
||||
const TIMING: Timing = Timing {
|
||||
start_time: 0.0,
|
||||
end_time: 1.0,
|
||||
};
|
||||
|
||||
fn logged(py: Python<'_>, locals: &Bound<'_, PyDict>, asynchronous: bool) -> LegacyLogging {
|
||||
LegacyLogging {
|
||||
logger: Some(PythonLogger::new(local(locals, "logger").unbind())),
|
||||
..legacy_call(py, locals, asynchronous)
|
||||
}
|
||||
}
|
||||
|
||||
fn succeed(
|
||||
py: Python<'_>,
|
||||
locals: &Bound<'_, PyDict>,
|
||||
logging: &mut LegacyLogging,
|
||||
) -> LifecycleStep {
|
||||
let response = local(locals, "response").unbind();
|
||||
logging
|
||||
.emit(
|
||||
py,
|
||||
LifecycleEvent::Succeeded {
|
||||
timing: TIMING,
|
||||
response: &response,
|
||||
},
|
||||
)
|
||||
.unwrap()
|
||||
}
|
||||
|
||||
fn fail(py: Python<'_>, locals: &Bound<'_, PyDict>, logging: &mut LegacyLogging) -> LifecycleStep {
|
||||
let failure = PyErr::from_value(local(locals, "failure"));
|
||||
logging
|
||||
.emit(
|
||||
py,
|
||||
LifecycleEvent::Failed {
|
||||
timing: TIMING,
|
||||
origin: FailureOrigin::Host,
|
||||
error: &failure,
|
||||
},
|
||||
)
|
||||
.unwrap()
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::sync_listened(false, c"", &["submit"])]
|
||||
#[case::async_listened(
|
||||
true,
|
||||
c"",
|
||||
&["async_success_handler", "enqueued", "sync_success_for_async_call"]
|
||||
)]
|
||||
#[case::async_deferred(true, c"logger._defer_async_logging = True", &["sync_success_for_async_call"])]
|
||||
#[case::async_with_fallbacks(true, c"kwargs = {'fallbacks': ['other']}", &["sync_success_for_async_call"])]
|
||||
fn success_reaches_the_logging_handlers(
|
||||
#[case] asynchronous: bool,
|
||||
#[case] script: &CStr,
|
||||
#[case] expected: &[&str],
|
||||
) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(py, c"response = object()");
|
||||
run(py, &locals, script);
|
||||
let mut logging = logged(py, &locals, asynchronous);
|
||||
assert!(matches!(
|
||||
succeed(py, &locals, &mut logging),
|
||||
LifecycleStep::Done
|
||||
));
|
||||
let names: Vec<String> = local(&locals, "logger")
|
||||
.call_method0("names")
|
||||
.unwrap()
|
||||
.extract()
|
||||
.unwrap();
|
||||
assert_eq!(names, expected);
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"
|
||||
assert all(value is response for name, value in logger.calls if name.endswith('_handler'))
|
||||
assert hasattr(logger, '_native_pending_logging') == getattr(logger, '_defer_async_logging', False)
|
||||
",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::synchronous(false, &["failure_handler"])]
|
||||
#[case::asynchronous(true, &[])]
|
||||
fn internal_calls_skip_failure_callbacks_only_when_asynchronous(
|
||||
#[case] asynchronous: bool,
|
||||
#[case] expected: &[&str],
|
||||
) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(py, c"failure = ValueError('provider')");
|
||||
let mut logging = LegacyLogging {
|
||||
internal: true,
|
||||
..logged(py, &locals, asynchronous)
|
||||
};
|
||||
assert!(matches!(
|
||||
fail(py, &locals, &mut logging),
|
||||
LifecycleStep::Done
|
||||
));
|
||||
let names: Vec<String> = local(&locals, "logger")
|
||||
.call_method0("names")
|
||||
.unwrap()
|
||||
.extract()
|
||||
.unwrap();
|
||||
assert_eq!(names, expected);
|
||||
});
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn internal_async_calls_skip_the_async_success_fan_out() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(py, c"response = object()");
|
||||
let mut logging = LegacyLogging {
|
||||
internal: true,
|
||||
..logged(py, &locals, true)
|
||||
};
|
||||
succeed(py, &locals, &mut logging);
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"assert logger.names() == ['sync_success_for_async_call'], logger.calls",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_failing_success_callback_is_reported_without_replacing_the_response() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(
|
||||
py,
|
||||
c"
|
||||
response = object()
|
||||
failure = ValueError('terminal diagnostic')
|
||||
|
||||
class FailingLogger(StubLogger):
|
||||
def handle_sync_success_callbacks_for_async_calls(self, *args):
|
||||
raise failure
|
||||
|
||||
logger = FailingLogger()
|
||||
",
|
||||
);
|
||||
let mut logging = logged(py, &locals, true);
|
||||
assert!(matches!(
|
||||
succeed(py, &locals, &mut logging),
|
||||
LifecycleStep::Done
|
||||
));
|
||||
assert!(
|
||||
logging
|
||||
.response
|
||||
.as_ref()
|
||||
.unwrap()
|
||||
.bind(py)
|
||||
.is(local(&locals, "response"))
|
||||
);
|
||||
run(py, &locals, c"assert unraisable_from(logger) == [failure]");
|
||||
});
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::sync_listened(false, c"", &["failure_handler"])]
|
||||
#[case::async_listened(true, c"", &["failure_handler", "async_failure_handler"])]
|
||||
fn failure_reaches_the_logging_handlers(
|
||||
#[case] asynchronous: bool,
|
||||
#[case] script: &CStr,
|
||||
#[case] expected: &[&str],
|
||||
) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(py, c"failure = ValueError('provider')");
|
||||
run(py, &locals, script);
|
||||
let mut logging = logged(py, &locals, asynchronous);
|
||||
let step = fail(py, &locals, &mut logging);
|
||||
let awaits_async_handler = expected.contains(&"async_failure_handler");
|
||||
assert_eq!(
|
||||
matches!(step, LifecycleStep::Await(_)),
|
||||
awaits_async_handler
|
||||
);
|
||||
let names: Vec<String> = local(&locals, "logger")
|
||||
.call_method0("names")
|
||||
.unwrap()
|
||||
.extract()
|
||||
.unwrap();
|
||||
assert_eq!(names, expected);
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"assert all(value is failure for name, value in logger.calls if name.endswith('_handler'))",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_failing_sync_failure_callback_keeps_the_error_and_still_runs_the_async_family() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(
|
||||
py,
|
||||
c"
|
||||
failure = ValueError('selected')
|
||||
|
||||
class FailingLogger(StubLogger):
|
||||
def failure_handler(self, error, trace, start, end):
|
||||
self.record('failure_handler', error)
|
||||
raise RuntimeError('handler failed')
|
||||
|
||||
logger = FailingLogger()
|
||||
",
|
||||
);
|
||||
let mut logging = logged(py, &locals, true);
|
||||
assert!(matches!(
|
||||
fail(py, &locals, &mut logging),
|
||||
LifecycleStep::Await(_)
|
||||
));
|
||||
assert!(
|
||||
logging
|
||||
.error
|
||||
.as_ref()
|
||||
.unwrap()
|
||||
.bind(py)
|
||||
.is(local(&locals, "failure"))
|
||||
);
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"assert logger.names() == ['failure_handler', 'async_failure_handler'], logger.calls",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::completed(None, true)]
|
||||
#[case::handler_error(Some(false), true)]
|
||||
#[case::cancelled(Some(true), false)]
|
||||
fn the_async_failure_handler_ends_the_call_unless_it_was_cancelled(
|
||||
#[case] error: Option<bool>,
|
||||
#[case] done: bool,
|
||||
) {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(py, c"failure = ValueError('provider')");
|
||||
let mut logging = logged(py, &locals, true);
|
||||
fail(py, &locals, &mut logging);
|
||||
let result = match error {
|
||||
None => Ok(py.None()),
|
||||
Some(false) => Err(PyRuntimeError::new_err("handler failed")),
|
||||
Some(true) => Err(CancelledError::new_err("cancelled")),
|
||||
};
|
||||
let expected = result.as_ref().err().map(|error| error.value(py).clone());
|
||||
match logging.resume(py, result) {
|
||||
Ok(step) => assert!(done && matches!(step, LifecycleStep::Done)),
|
||||
Err(propagated) => {
|
||||
assert!(!done);
|
||||
assert!(propagated.value(py).is(expected.unwrap()));
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn closing_restores_the_correlation_context_once() {
|
||||
Python::initialize();
|
||||
Python::attach(|py| {
|
||||
let locals = namespace(py, c"");
|
||||
let mut logging = logged(py, &locals, true);
|
||||
logging.close(py);
|
||||
logging.close(py);
|
||||
run(
|
||||
py,
|
||||
&locals,
|
||||
c"assert logger.names() == ['restore'], logger.calls",
|
||||
);
|
||||
});
|
||||
}
|
||||
|
|
@ -4,7 +4,6 @@ version = "0.1.0"
|
|||
edition.workspace = true
|
||||
license.workspace = true
|
||||
repository.workspace = true
|
||||
autotests = false
|
||||
|
||||
[dependencies]
|
||||
litellm-secrets.workspace = true
|
||||
|
|
|
|||
|
|
@ -14,6 +14,3 @@ pub async fn audio_transcription(request: AudioTranscriptionRequest<'_>) -> Resu
|
|||
execute_audio_transcription_provider_call(prepare_audio_transcription_provider_call(request)?)
|
||||
.await
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests;
|
||||
|
|
|
|||
|
|
@ -51,6 +51,3 @@ pub fn chat_completions_decline_reason(
|
|||
.unsupported_reason(&messages, optional_params)
|
||||
.map(|reason| reason.0)
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests;
|
||||
|
|
|
|||
|
|
@ -143,3 +143,841 @@ pub(super) fn prepare_provider_request(
|
|||
timeout: request.timeout,
|
||||
})
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use litellm_llms::base_llm::chat::transformation::RequestAuth;
|
||||
use serde_json::{Map, Value, json};
|
||||
|
||||
use super::{prepare_provider_request, resolve_request};
|
||||
use crate::chat_completions::{
|
||||
Error,
|
||||
types::{ChatCompletionsRequest, ProviderChatCompletionsRequest},
|
||||
};
|
||||
|
||||
fn prepare_chat_completions_call(
|
||||
request: ChatCompletionsRequest<'_>,
|
||||
) -> Result<ProviderChatCompletionsRequest, Error> {
|
||||
prepare_provider_request(resolve_request(request)?)
|
||||
}
|
||||
|
||||
fn request<'a>(
|
||||
model: &'a str,
|
||||
provider: Option<&'a str>,
|
||||
messages: Value,
|
||||
optional_params: Value,
|
||||
) -> ChatCompletionsRequest<'a> {
|
||||
ChatCompletionsRequest {
|
||||
model,
|
||||
messages,
|
||||
optional_params: match optional_params {
|
||||
Value::Object(map) => map,
|
||||
other => panic!("params must be an object, got {other}"),
|
||||
},
|
||||
api_key: Some("sk-test"),
|
||||
api_base: None,
|
||||
custom_llm_provider: provider,
|
||||
extra_headers: None,
|
||||
timeout: None,
|
||||
}
|
||||
}
|
||||
|
||||
/// `ProviderChatCompletionsRequest` deliberately has no `Debug` (its headers
|
||||
/// carry resolved credentials), so unwrap the failure case by hand.
|
||||
fn decline(request: ChatCompletionsRequest<'_>) -> Error {
|
||||
match prepare_chat_completions_call(request) {
|
||||
Err(error) => error,
|
||||
Ok(prepared) => panic!("expected a decline, prepared a call to {}", prepared.url),
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn resolves_the_provider_from_the_model_prefix() {
|
||||
let prepared = prepare_chat_completions_call(request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.expect("prepares");
|
||||
assert_eq!(prepared.model, "claude-sonnet-4-5");
|
||||
assert_eq!(prepared.url, "https://api.anthropic.com/v1/messages");
|
||||
assert_eq!(prepared.body["model"], json!("claude-sonnet-4-5"));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn strips_an_explicit_provider_prefix_from_the_model() {
|
||||
let prepared = prepare_chat_completions_call(request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
))
|
||||
.expect("prepares");
|
||||
assert_eq!(prepared.model, "claude-sonnet-4-5");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn adds_the_auth_and_default_headers() {
|
||||
let prepared = prepare_chat_completions_call(request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
))
|
||||
.expect("prepares");
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.contains(&("x-api-key".to_string(), "sk-test".to_string()))
|
||||
);
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.contains(&("anthropic-version".to_string(), "2023-06-01".to_string()))
|
||||
);
|
||||
assert!(matches!(
|
||||
prepared.auth,
|
||||
RequestAuth::Header {
|
||||
name: "x-api-key",
|
||||
..
|
||||
}
|
||||
));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn the_deployment_credential_replaces_a_caller_supplied_auth_header() {
|
||||
// Python builds `{**headers, **anthropic_headers}`, so the deployment's key
|
||||
// overwrites a forwarded one. Honouring the caller's would let whoever sends
|
||||
// the request choose the Anthropic principal it bills to.
|
||||
let mut call = request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([(
|
||||
"X-Api-Key".to_string(),
|
||||
json!("sk-caller"),
|
||||
)]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let keys: Vec<_> = prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
|
||||
.collect();
|
||||
assert_eq!(keys.len(), 1, "got {:?}", prepared.upstream_headers);
|
||||
assert_eq!(keys[0].1, "sk-test");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_forwarded_authorization_header_suppresses_the_resolved_api_key_header() {
|
||||
// Anthropic's `validate_environment` pops `x-api-key` and sets `authorization`
|
||||
// for an OAuth token, so re-adding the key here would put the credential into
|
||||
// a header the host removed on purpose.
|
||||
let mut call = request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([
|
||||
(
|
||||
"Authorization".to_string(),
|
||||
json!("Bearer sk-ant-oat01-token"),
|
||||
),
|
||||
("X-Api-Key".to_string(), json!("sk-caller")),
|
||||
]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
assert!(
|
||||
!prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, value)| name.eq_ignore_ascii_case("x-api-key") && value == "sk-test"),
|
||||
"the resolved key must not be applied over an OAuth bearer, got {:?}",
|
||||
prepared.upstream_headers
|
||||
);
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, value)| name.eq_ignore_ascii_case("authorization")
|
||||
&& value == "Bearer sk-ant-oat01-token")
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn an_unrelated_forwarded_authorization_does_not_defer_the_resolved_key() {
|
||||
// Only an OAuth bearer replaces the credential. Python sends the deployment's
|
||||
// `x-api-key` alongside any other forwarded `authorization`, so deferring on
|
||||
// the mere presence of that header would drop the deployment's auth.
|
||||
let mut call = request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([
|
||||
("Authorization".to_string(), json!("Bearer unrelated")),
|
||||
("X-Api-Key".to_string(), json!("sk-caller")),
|
||||
]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let keys: Vec<_> = prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
|
||||
.collect();
|
||||
assert_eq!(keys.len(), 1, "got {:?}", prepared.upstream_headers);
|
||||
assert_eq!(keys[0].1, "sk-test");
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, value)| name.eq_ignore_ascii_case("authorization")
|
||||
&& value == "Bearer unrelated"),
|
||||
"the unrelated authorization must survive, got {:?}",
|
||||
prepared.upstream_headers
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn declines_an_unsupported_request_before_resolving_credentials() {
|
||||
let mut call = request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"stream": true}),
|
||||
);
|
||||
call.api_key = None;
|
||||
// No api_key is set and no env is consulted: the gate must run first, so the
|
||||
// error is the decline rather than a missing-credential error.
|
||||
assert_eq!(decline(call), Error::Unsupported("streaming"));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_an_unknown_provider() {
|
||||
assert_eq!(
|
||||
decline(request(
|
||||
"openai/gpt-4o",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
)),
|
||||
Error::InvalidProvider("openai".to_string())
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_a_model_with_no_resolvable_provider() {
|
||||
assert!(matches!(
|
||||
decline(request(
|
||||
"claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
)),
|
||||
Error::InvalidProvider(_)
|
||||
));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_an_empty_or_malformed_message_list() {
|
||||
assert_eq!(
|
||||
decline(request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!([]),
|
||||
json!({}),
|
||||
)),
|
||||
Error::InvalidRequest("chat completions requires at least one message".to_string())
|
||||
);
|
||||
assert!(matches!(
|
||||
decline(request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!("not a list"),
|
||||
json!({}),
|
||||
)),
|
||||
Error::InvalidRequest(_)
|
||||
));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_non_string_extra_headers() {
|
||||
let mut call = request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([("x-trace".to_string(), json!(7))]));
|
||||
assert_eq!(
|
||||
decline(call),
|
||||
Error::Headers(litellm_http::request::HeaderError {
|
||||
context: "chat completions",
|
||||
name: "x-trace".to_string(),
|
||||
actual: "number",
|
||||
})
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn prepares_a_bedrock_call_without_resolving_credentials() {
|
||||
let mut call = request(
|
||||
"bedrock/us-east-1/anthropic.claude-v2",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"maxTokens": 16}),
|
||||
);
|
||||
call.api_key = None;
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
assert_eq!(
|
||||
prepared.url,
|
||||
"https://bedrock-runtime.us-east-1.amazonaws.com/model/anthropic.claude-v2/converse"
|
||||
);
|
||||
assert_eq!(
|
||||
prepared.auth,
|
||||
RequestAuth::AwsSigV4 {
|
||||
region: "us-east-1".to_string(),
|
||||
service: "bedrock",
|
||||
}
|
||||
);
|
||||
// SigV4 signs the serialized body, so prepare must not have added an
|
||||
// Authorization header; the handler does it.
|
||||
assert!(
|
||||
!prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, _)| name.eq_ignore_ascii_case("authorization"))
|
||||
);
|
||||
assert_eq!(prepared.body["inferenceConfig"], json!({"maxTokens": 16}));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_forwarded_client_header_does_not_enter_the_bedrock_signature() {
|
||||
// Python signs only the AWS header set and reattaches the rest, so a header
|
||||
// the caller forwarded rides along without joining the canonical request.
|
||||
// Signing it makes Converse 403 on a deployment that works on Python.
|
||||
let mut call = request(
|
||||
"bedrock/us-east-1/anthropic.claude-v2",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({
|
||||
"maxTokens": 16,
|
||||
"aws_access_key_id": "AKIDEXAMPLE",
|
||||
"aws_secret_access_key": "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY"
|
||||
}),
|
||||
);
|
||||
// A key would resolve to a bearer token and never reach the signer.
|
||||
call.api_key = None;
|
||||
call.extra_headers = Some(Map::from_iter([(
|
||||
"x-request-id".to_string(),
|
||||
json!("abc-123"),
|
||||
)]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let signed = crate::chat_completions::handler::outbound_request(&prepared)
|
||||
.await
|
||||
.expect("signs");
|
||||
|
||||
let authorization = signed
|
||||
.header("authorization")
|
||||
.expect("carries an authorization header")
|
||||
.to_string();
|
||||
assert!(
|
||||
authorization.starts_with("AWS4-HMAC-SHA256"),
|
||||
"expected a SigV4 signature, got {authorization}"
|
||||
);
|
||||
assert!(
|
||||
!authorization.contains("x-request-id"),
|
||||
"forwarded header reached SignedHeaders: {authorization}"
|
||||
);
|
||||
// It still goes on the wire, it is just not part of the signature.
|
||||
assert!(
|
||||
signed
|
||||
.headers()
|
||||
.iter()
|
||||
.any(|(name, value)| name == "x-request-id" && value == "abc-123"),
|
||||
"forwarded header was dropped instead of reattached"
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_forwarded_header_the_signer_computes_declines_to_python() {
|
||||
// Reattaching the caller's copy next to the computed one puts the name on
|
||||
// the wire twice and Bedrock rejects the pair, so a request carrying one
|
||||
// has to go to Python instead of being signed here.
|
||||
for forwarded in [
|
||||
"Authorization",
|
||||
"x-amz-date",
|
||||
"x-amz-security-token",
|
||||
"Date",
|
||||
] {
|
||||
let mut call = request(
|
||||
"bedrock/us-east-1/anthropic.claude-v2",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({
|
||||
"maxTokens": 16,
|
||||
"aws_access_key_id": "AKIDEXAMPLE",
|
||||
"aws_secret_access_key": "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY"
|
||||
}),
|
||||
);
|
||||
call.api_key = None;
|
||||
call.extra_headers = Some(Map::from_iter([(forwarded.to_string(), json!("forged"))]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let error = crate::chat_completions::handler::outbound_request(&prepared)
|
||||
.await
|
||||
.expect_err("{forwarded} should decline instead of being signed");
|
||||
assert!(
|
||||
matches!(error, Error::Unsupported(_)),
|
||||
"{forwarded} declined as {error:?}, which the host would not fall back on"
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_bedrock_deployment_bearer_outranks_a_forwarded_authorization() {
|
||||
// `get_request_headers` assigns `headers["Authorization"]` unconditionally
|
||||
// once a bearer token resolves, so the deployment's identity wins on
|
||||
// Python. Keeping the caller's would authorize and bill the call as a
|
||||
// different principal, and only when the deployment carries `rust: true`.
|
||||
let mut call = request(
|
||||
"bedrock/us-east-1/anthropic.claude-v2",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"maxTokens": 16}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([(
|
||||
"Authorization".to_string(),
|
||||
json!("Bearer caller-supplied"),
|
||||
)]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let authorizations: Vec<_> = prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.filter(|(name, _)| name.eq_ignore_ascii_case("authorization"))
|
||||
.map(|(_, value)| value.as_str())
|
||||
.collect();
|
||||
assert_eq!(
|
||||
authorizations,
|
||||
vec!["Bearer sk-test"],
|
||||
"the deployment token must be the only authorization on the wire"
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn an_anthropic_forwarded_oauth_bearer_still_outranks_the_resolved_key() {
|
||||
// The opposite precedence, and deliberate: Anthropic's own transform
|
||||
// honours a forwarded OAuth bearer, so the Bedrock fix above must not be
|
||||
// generalized into a rule that the configured key always wins.
|
||||
//
|
||||
// An OAuth bearer is the whole of that exception. This forwarded a plain
|
||||
// `x-api-key` until round 17, which read as the same claim and was not:
|
||||
// Python overwrites a forwarded `x-api-key` with the deployment's.
|
||||
let mut call = request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([(
|
||||
"authorization".to_string(),
|
||||
json!("Bearer sk-ant-oat01-forwarded"),
|
||||
)]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let keys: Vec<_> = prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
|
||||
.map(|(_, value)| value.as_str())
|
||||
.collect();
|
||||
assert!(keys.is_empty(), "got {:?}", prepared.upstream_headers);
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, value)| name.eq_ignore_ascii_case("authorization")
|
||||
&& value == "Bearer sk-ant-oat01-forwarded")
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_bedrock_api_key_is_sent_as_a_bearer_token_instead_of_being_signed() {
|
||||
// The configured bearer identity has its own account and quota boundary,
|
||||
// so a request carrying one must not be signed as whatever principal the
|
||||
// host's AWS credentials resolve to.
|
||||
let prepared = prepare_chat_completions_call(request(
|
||||
"bedrock/us-east-1/anthropic.claude-v2",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"maxTokens": 16}),
|
||||
))
|
||||
.expect("prepares");
|
||||
assert_eq!(
|
||||
prepared.auth,
|
||||
RequestAuth::Bearer {
|
||||
token: "sk-test".to_string()
|
||||
}
|
||||
);
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, value)| name.eq_ignore_ascii_case("authorization")
|
||||
&& value == "Bearer sk-test"),
|
||||
"prepare did not carry the bearer token"
|
||||
);
|
||||
}
|
||||
|
||||
fn decline_reason(
|
||||
model: &str,
|
||||
provider: Option<&str>,
|
||||
messages: Value,
|
||||
params: Value,
|
||||
) -> Option<&'static str> {
|
||||
let params = match params {
|
||||
Value::Object(map) => map,
|
||||
other => panic!("params must be an object, got {other}"),
|
||||
};
|
||||
crate::chat_completions::chat_completions_decline_reason(model, provider, messages, ¶ms)
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn the_gate_accepts_what_prepare_accepts() {
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
),
|
||||
None
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn the_gate_declines_without_resolving_credentials_or_calling_out() {
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"stream": true}),
|
||||
),
|
||||
Some("streaming")
|
||||
);
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"openai/gpt-4o",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
),
|
||||
Some("provider is not on the rust chat completions path")
|
||||
);
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
),
|
||||
Some("provider is not on the rust chat completions path")
|
||||
);
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!("nope"),
|
||||
json!({})
|
||||
),
|
||||
Some("unreadable message list")
|
||||
);
|
||||
assert_eq!(
|
||||
decline_reason("anthropic/claude-sonnet-4-5", None, json!([]), json!({})),
|
||||
Some("empty message list")
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn the_gate_agrees_with_prepare_on_every_case_it_accepts() {
|
||||
// A gate that accepts what prepare then declines would make the host emit
|
||||
// its pre-call logging on a path that falls back, so pin the agreement.
|
||||
for (messages, params) in [
|
||||
(
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 8}),
|
||||
),
|
||||
(
|
||||
json!([{"role": "system", "content": "s"}, {"role": "user", "content": "hi"}]),
|
||||
json!({"temperature": 0.1}),
|
||||
),
|
||||
(
|
||||
json!([{"role": "user", "content": "hi"}, {"role": "assistant", "content": "yo"}]),
|
||||
json!({}),
|
||||
),
|
||||
] {
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
messages.clone(),
|
||||
params.clone()
|
||||
),
|
||||
None,
|
||||
"gate declined {messages}"
|
||||
);
|
||||
prepare_chat_completions_call(request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
messages.clone(),
|
||||
params,
|
||||
))
|
||||
.unwrap_or_else(|error| panic!("prepare declined {messages}: {error}"));
|
||||
}
|
||||
}
|
||||
|
||||
mod round_trip {
|
||||
use tokio::{
|
||||
io::{AsyncReadExt, AsyncWriteExt},
|
||||
net::{TcpListener, TcpStream},
|
||||
};
|
||||
|
||||
use super::*;
|
||||
use crate::chat_completions::chat_completions;
|
||||
|
||||
async fn read_http_request(socket: &mut TcpStream) -> String {
|
||||
let mut request = Vec::new();
|
||||
let mut buffer = [0_u8; 1024];
|
||||
let header_end = loop {
|
||||
let n = socket.read(&mut buffer).await.expect("reads request");
|
||||
if n == 0 {
|
||||
break request.len();
|
||||
}
|
||||
request.extend_from_slice(&buffer[..n]);
|
||||
if let Some(position) = request.windows(4).position(|window| window == b"\r\n\r\n")
|
||||
{
|
||||
break position + 4;
|
||||
}
|
||||
};
|
||||
let headers = String::from_utf8_lossy(&request[..header_end]);
|
||||
let content_length = headers
|
||||
.lines()
|
||||
.find_map(|line| {
|
||||
let (name, value) = line.split_once(':')?;
|
||||
name.eq_ignore_ascii_case("content-length")
|
||||
.then(|| value.trim().parse::<usize>().ok())
|
||||
.flatten()
|
||||
})
|
||||
.unwrap_or(0);
|
||||
while request.len().saturating_sub(header_end) < content_length {
|
||||
let n = socket.read(&mut buffer).await.expect("reads body");
|
||||
if n == 0 {
|
||||
break;
|
||||
}
|
||||
request.extend_from_slice(&buffer[..n]);
|
||||
}
|
||||
String::from_utf8(request).expect("request is utf8")
|
||||
}
|
||||
|
||||
fn http_response(status: &str, body: &str) -> String {
|
||||
format!(
|
||||
"HTTP/1.1 {status}\r\ncontent-type: application/json\r\ncontent-length: {}\r\nconnection: close\r\n\r\n{}",
|
||||
body.len(),
|
||||
body
|
||||
)
|
||||
}
|
||||
|
||||
/// Serve one request from a stub upstream and hand back what it received.
|
||||
async fn serve_once(
|
||||
status: &'static str,
|
||||
body: &'static str,
|
||||
) -> (String, tokio::task::JoinHandle<String>) {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
|
||||
let port = listener.local_addr().expect("addr").port();
|
||||
let handle = tokio::spawn(async move {
|
||||
let (mut socket, _) = listener.accept().await.expect("accepts");
|
||||
let received = read_http_request(&mut socket).await;
|
||||
socket
|
||||
.write_all(http_response(status, body).as_bytes())
|
||||
.await
|
||||
.expect("writes response");
|
||||
socket.flush().await.expect("flushes");
|
||||
received
|
||||
});
|
||||
(format!("http://127.0.0.1:{port}/v1/messages"), handle)
|
||||
}
|
||||
|
||||
fn call(api_base: &str, messages: Value, params: Value) -> ChatCompletionsRequest<'_> {
|
||||
ChatCompletionsRequest {
|
||||
model: "anthropic/claude-sonnet-4-5",
|
||||
messages,
|
||||
optional_params: match params {
|
||||
Value::Object(map) => map,
|
||||
other => panic!("params must be an object, got {other}"),
|
||||
},
|
||||
api_key: Some("sk-test"),
|
||||
api_base: Some(api_base),
|
||||
custom_llm_provider: None,
|
||||
extra_headers: None,
|
||||
timeout: Some(std::time::Duration::from_secs(10)),
|
||||
}
|
||||
}
|
||||
|
||||
const GOOD_BODY: &str = r#"{"id":"msg_1","type":"message","role":"assistant","model":"claude-sonnet-4-5-20260101","content":[{"type":"text","text":"hello"}],"stop_reason":"end_turn","stop_sequence":null,"usage":{"input_tokens":11,"output_tokens":4}}"#;
|
||||
|
||||
#[tokio::test]
|
||||
async fn round_trip_sends_the_translated_body_and_normalizes_the_response() {
|
||||
let (api_base, handle) = serve_once("200 OK", GOOD_BODY).await;
|
||||
let response = chat_completions(call(
|
||||
&api_base,
|
||||
json!([
|
||||
{"role": "system", "content": "be terse"},
|
||||
{"role": "user", "content": "hi"}
|
||||
]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.await
|
||||
.expect("call succeeds");
|
||||
|
||||
let received = handle.await.expect("server task");
|
||||
let sent: Value = serde_json::from_str(
|
||||
received
|
||||
.split_once("\r\n\r\n")
|
||||
.expect("request has a body")
|
||||
.1,
|
||||
)
|
||||
.expect("body is json");
|
||||
assert_eq!(
|
||||
sent["messages"],
|
||||
json!([{"role": "user", "content": [{"type": "text", "text": "hi"}]}])
|
||||
);
|
||||
assert_eq!(
|
||||
sent["system"],
|
||||
json!([{"type": "text", "text": "be terse"}])
|
||||
);
|
||||
assert_eq!(sent["max_tokens"], json!(16));
|
||||
assert!(received.to_lowercase().contains("x-api-key: sk-test"));
|
||||
|
||||
assert_eq!(
|
||||
response.choices[0].message.content.as_deref(),
|
||||
Some("hello")
|
||||
);
|
||||
assert_eq!(response.usage.total_tokens, 15);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_response_it_cannot_normalize_is_reported_as_already_sent() {
|
||||
// The provider was called and billed, so the host must not retry this
|
||||
// on its own path. `MissingField` here would read as a pre-send
|
||||
// decline and be retried; `InvalidResponse` cannot.
|
||||
const NO_USAGE: &str =
|
||||
r#"{"model":"m","content":[{"type":"text","text":"hi"}],"stop_reason":"end_turn"}"#;
|
||||
let (api_base, handle) = serve_once("200 OK", NO_USAGE).await;
|
||||
let err = chat_completions(call(
|
||||
&api_base,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.await
|
||||
.expect_err("response cannot be normalized");
|
||||
handle.await.expect("server task");
|
||||
assert!(
|
||||
matches!(err, Error::InvalidResponse(_)),
|
||||
"expected a post-send error, got {err:?}"
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_tool_use_block_in_the_response_is_also_reported_as_already_sent() {
|
||||
const TOOL_USE: &str = r#"{"model":"m","content":[{"type":"tool_use","id":"t","name":"f","input":{}}],"stop_reason":"tool_use","usage":{"input_tokens":1,"output_tokens":1}}"#;
|
||||
let (api_base, handle) = serve_once("200 OK", TOOL_USE).await;
|
||||
let err = chat_completions(call(
|
||||
&api_base,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.await
|
||||
.expect_err("response cannot be normalized");
|
||||
handle.await.expect("server task");
|
||||
assert!(
|
||||
matches!(err, Error::InvalidResponse(_)),
|
||||
"expected a post-send error, got {err:?}"
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn an_upstream_error_status_keeps_its_code() {
|
||||
let (api_base, handle) =
|
||||
serve_once("429 Too Many Requests", r#"{"error":"slow down"}"#).await;
|
||||
let err = chat_completions(call(
|
||||
&api_base,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.await
|
||||
.expect_err("upstream rejects");
|
||||
handle.await.expect("server task");
|
||||
assert!(
|
||||
matches!(
|
||||
err,
|
||||
Error::Transport(litellm_http::transport::Error::Http { status: 429, .. })
|
||||
),
|
||||
"expected a 429, got {err:?}"
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_connection_that_is_never_established_declines_instead_of_failing() {
|
||||
// Nothing was sent, so nothing was billed and the host can still serve
|
||||
// the request. Classing this with the post-send failures would turn a
|
||||
// recoverable fallback into a user-facing error on exactly the
|
||||
// deployments whose transport is configured only on the Python client.
|
||||
let port = {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
|
||||
listener.local_addr().expect("has an address").port()
|
||||
// Dropped here, so the port is closed and the connect is refused.
|
||||
};
|
||||
let err = chat_completions(call(
|
||||
&format!("http://127.0.0.1:{port}/v1/messages"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.await
|
||||
.expect_err("nothing is listening");
|
||||
assert!(
|
||||
matches!(
|
||||
err,
|
||||
Error::Transport(litellm_http::transport::Error::Connect(_))
|
||||
),
|
||||
"expected a pre-send connect failure, got {err:?}"
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn response_errors_collapse_to_one_variant_that_can_only_mean_already_sent() {
|
||||
use crate::chat_completions::handler::as_response_error;
|
||||
|
||||
for original in [
|
||||
Error::MissingField("usage"),
|
||||
Error::Unsupported("non-text response content block"),
|
||||
Error::InvalidRequest("whatever".to_string()),
|
||||
Error::Auth(litellm_auth::Error::InvalidHeader),
|
||||
] {
|
||||
let label = format!("{original:?}");
|
||||
assert!(
|
||||
matches!(as_response_error(original), Error::InvalidResponse(_)),
|
||||
"{label} must not stay retryable once the provider has answered"
|
||||
);
|
||||
}
|
||||
// An upstream status is already unambiguous, so it survives intact.
|
||||
assert!(matches!(
|
||||
as_response_error(Error::Transport(litellm_http::transport::Error::Http {
|
||||
status: 500,
|
||||
body: "boom".to_string()
|
||||
})),
|
||||
Error::Transport(litellm_http::transport::Error::Http { status: 500, .. })
|
||||
));
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -1,833 +0,0 @@
|
|||
use litellm_llms::base_llm::chat::transformation::RequestAuth;
|
||||
use serde_json::{Map, Value, json};
|
||||
|
||||
use super::{
|
||||
Error,
|
||||
prepare::{prepare_provider_request, resolve_request},
|
||||
};
|
||||
use crate::chat_completions::types::{ChatCompletionsRequest, ProviderChatCompletionsRequest};
|
||||
|
||||
fn prepare_chat_completions_call(
|
||||
request: ChatCompletionsRequest<'_>,
|
||||
) -> Result<ProviderChatCompletionsRequest, Error> {
|
||||
prepare_provider_request(resolve_request(request)?)
|
||||
}
|
||||
|
||||
fn request<'a>(
|
||||
model: &'a str,
|
||||
provider: Option<&'a str>,
|
||||
messages: Value,
|
||||
optional_params: Value,
|
||||
) -> ChatCompletionsRequest<'a> {
|
||||
ChatCompletionsRequest {
|
||||
model,
|
||||
messages,
|
||||
optional_params: match optional_params {
|
||||
Value::Object(map) => map,
|
||||
other => panic!("params must be an object, got {other}"),
|
||||
},
|
||||
api_key: Some("sk-test"),
|
||||
api_base: None,
|
||||
custom_llm_provider: provider,
|
||||
extra_headers: None,
|
||||
timeout: None,
|
||||
}
|
||||
}
|
||||
|
||||
/// `ProviderChatCompletionsRequest` deliberately has no `Debug` (its headers
|
||||
/// carry resolved credentials), so unwrap the failure case by hand.
|
||||
fn decline(request: ChatCompletionsRequest<'_>) -> Error {
|
||||
match prepare_chat_completions_call(request) {
|
||||
Err(error) => error,
|
||||
Ok(prepared) => panic!("expected a decline, prepared a call to {}", prepared.url),
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn resolves_the_provider_from_the_model_prefix() {
|
||||
let prepared = prepare_chat_completions_call(request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.expect("prepares");
|
||||
assert_eq!(prepared.model, "claude-sonnet-4-5");
|
||||
assert_eq!(prepared.url, "https://api.anthropic.com/v1/messages");
|
||||
assert_eq!(prepared.body["model"], json!("claude-sonnet-4-5"));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn strips_an_explicit_provider_prefix_from_the_model() {
|
||||
let prepared = prepare_chat_completions_call(request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
))
|
||||
.expect("prepares");
|
||||
assert_eq!(prepared.model, "claude-sonnet-4-5");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn adds_the_auth_and_default_headers() {
|
||||
let prepared = prepare_chat_completions_call(request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
))
|
||||
.expect("prepares");
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.contains(&("x-api-key".to_string(), "sk-test".to_string()))
|
||||
);
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.contains(&("anthropic-version".to_string(), "2023-06-01".to_string()))
|
||||
);
|
||||
assert!(matches!(
|
||||
prepared.auth,
|
||||
RequestAuth::Header {
|
||||
name: "x-api-key",
|
||||
..
|
||||
}
|
||||
));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn the_deployment_credential_replaces_a_caller_supplied_auth_header() {
|
||||
// Python builds `{**headers, **anthropic_headers}`, so the deployment's key
|
||||
// overwrites a forwarded one. Honouring the caller's would let whoever sends
|
||||
// the request choose the Anthropic principal it bills to.
|
||||
let mut call = request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([(
|
||||
"X-Api-Key".to_string(),
|
||||
json!("sk-caller"),
|
||||
)]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let keys: Vec<_> = prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
|
||||
.collect();
|
||||
assert_eq!(keys.len(), 1, "got {:?}", prepared.upstream_headers);
|
||||
assert_eq!(keys[0].1, "sk-test");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_forwarded_authorization_header_suppresses_the_resolved_api_key_header() {
|
||||
// Anthropic's `validate_environment` pops `x-api-key` and sets `authorization`
|
||||
// for an OAuth token, so re-adding the key here would put the credential into
|
||||
// a header the host removed on purpose.
|
||||
let mut call = request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([
|
||||
(
|
||||
"Authorization".to_string(),
|
||||
json!("Bearer sk-ant-oat01-token"),
|
||||
),
|
||||
("X-Api-Key".to_string(), json!("sk-caller")),
|
||||
]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
assert!(
|
||||
!prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, value)| name.eq_ignore_ascii_case("x-api-key") && value == "sk-test"),
|
||||
"the resolved key must not be applied over an OAuth bearer, got {:?}",
|
||||
prepared.upstream_headers
|
||||
);
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, value)| name.eq_ignore_ascii_case("authorization")
|
||||
&& value == "Bearer sk-ant-oat01-token")
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn an_unrelated_forwarded_authorization_does_not_defer_the_resolved_key() {
|
||||
// Only an OAuth bearer replaces the credential. Python sends the deployment's
|
||||
// `x-api-key` alongside any other forwarded `authorization`, so deferring on
|
||||
// the mere presence of that header would drop the deployment's auth.
|
||||
let mut call = request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([
|
||||
("Authorization".to_string(), json!("Bearer unrelated")),
|
||||
("X-Api-Key".to_string(), json!("sk-caller")),
|
||||
]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let keys: Vec<_> = prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
|
||||
.collect();
|
||||
assert_eq!(keys.len(), 1, "got {:?}", prepared.upstream_headers);
|
||||
assert_eq!(keys[0].1, "sk-test");
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, value)| name.eq_ignore_ascii_case("authorization")
|
||||
&& value == "Bearer unrelated"),
|
||||
"the unrelated authorization must survive, got {:?}",
|
||||
prepared.upstream_headers
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn declines_an_unsupported_request_before_resolving_credentials() {
|
||||
let mut call = request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"stream": true}),
|
||||
);
|
||||
call.api_key = None;
|
||||
// No api_key is set and no env is consulted: the gate must run first, so the
|
||||
// error is the decline rather than a missing-credential error.
|
||||
assert_eq!(decline(call), Error::Unsupported("streaming"));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_an_unknown_provider() {
|
||||
assert_eq!(
|
||||
decline(request(
|
||||
"openai/gpt-4o",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
)),
|
||||
Error::InvalidProvider("openai".to_string())
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_a_model_with_no_resolvable_provider() {
|
||||
assert!(matches!(
|
||||
decline(request(
|
||||
"claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
)),
|
||||
Error::InvalidProvider(_)
|
||||
));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_an_empty_or_malformed_message_list() {
|
||||
assert_eq!(
|
||||
decline(request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!([]),
|
||||
json!({}),
|
||||
)),
|
||||
Error::InvalidRequest("chat completions requires at least one message".to_string())
|
||||
);
|
||||
assert!(matches!(
|
||||
decline(request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!("not a list"),
|
||||
json!({}),
|
||||
)),
|
||||
Error::InvalidRequest(_)
|
||||
));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_non_string_extra_headers() {
|
||||
let mut call = request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([("x-trace".to_string(), json!(7))]));
|
||||
assert_eq!(
|
||||
decline(call),
|
||||
Error::Headers(litellm_http::request::HeaderError {
|
||||
context: "chat completions",
|
||||
name: "x-trace".to_string(),
|
||||
actual: "number",
|
||||
})
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn prepares_a_bedrock_call_without_resolving_credentials() {
|
||||
let mut call = request(
|
||||
"bedrock/us-east-1/anthropic.claude-v2",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"maxTokens": 16}),
|
||||
);
|
||||
call.api_key = None;
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
assert_eq!(
|
||||
prepared.url,
|
||||
"https://bedrock-runtime.us-east-1.amazonaws.com/model/anthropic.claude-v2/converse"
|
||||
);
|
||||
assert_eq!(
|
||||
prepared.auth,
|
||||
RequestAuth::AwsSigV4 {
|
||||
region: "us-east-1".to_string(),
|
||||
service: "bedrock",
|
||||
}
|
||||
);
|
||||
// SigV4 signs the serialized body, so prepare must not have added an
|
||||
// Authorization header; the handler does it.
|
||||
assert!(
|
||||
!prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, _)| name.eq_ignore_ascii_case("authorization"))
|
||||
);
|
||||
assert_eq!(prepared.body["inferenceConfig"], json!({"maxTokens": 16}));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_forwarded_client_header_does_not_enter_the_bedrock_signature() {
|
||||
// Python signs only the AWS header set and reattaches the rest, so a header
|
||||
// the caller forwarded rides along without joining the canonical request.
|
||||
// Signing it makes Converse 403 on a deployment that works on Python.
|
||||
let mut call = request(
|
||||
"bedrock/us-east-1/anthropic.claude-v2",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({
|
||||
"maxTokens": 16,
|
||||
"aws_access_key_id": "AKIDEXAMPLE",
|
||||
"aws_secret_access_key": "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY"
|
||||
}),
|
||||
);
|
||||
// A key would resolve to a bearer token and never reach the signer.
|
||||
call.api_key = None;
|
||||
call.extra_headers = Some(Map::from_iter([(
|
||||
"x-request-id".to_string(),
|
||||
json!("abc-123"),
|
||||
)]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let signed = super::handler::outbound_request(&prepared)
|
||||
.await
|
||||
.expect("signs");
|
||||
|
||||
let authorization = signed
|
||||
.header("authorization")
|
||||
.expect("carries an authorization header")
|
||||
.to_string();
|
||||
assert!(
|
||||
authorization.starts_with("AWS4-HMAC-SHA256"),
|
||||
"expected a SigV4 signature, got {authorization}"
|
||||
);
|
||||
assert!(
|
||||
!authorization.contains("x-request-id"),
|
||||
"forwarded header reached SignedHeaders: {authorization}"
|
||||
);
|
||||
// It still goes on the wire, it is just not part of the signature.
|
||||
assert!(
|
||||
signed
|
||||
.headers()
|
||||
.iter()
|
||||
.any(|(name, value)| name == "x-request-id" && value == "abc-123"),
|
||||
"forwarded header was dropped instead of reattached"
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_forwarded_header_the_signer_computes_declines_to_python() {
|
||||
// Reattaching the caller's copy next to the computed one puts the name on
|
||||
// the wire twice and Bedrock rejects the pair, so a request carrying one
|
||||
// has to go to Python instead of being signed here.
|
||||
for forwarded in [
|
||||
"Authorization",
|
||||
"x-amz-date",
|
||||
"x-amz-security-token",
|
||||
"Date",
|
||||
] {
|
||||
let mut call = request(
|
||||
"bedrock/us-east-1/anthropic.claude-v2",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({
|
||||
"maxTokens": 16,
|
||||
"aws_access_key_id": "AKIDEXAMPLE",
|
||||
"aws_secret_access_key": "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY"
|
||||
}),
|
||||
);
|
||||
call.api_key = None;
|
||||
call.extra_headers = Some(Map::from_iter([(forwarded.to_string(), json!("forged"))]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let error = super::handler::outbound_request(&prepared)
|
||||
.await
|
||||
.expect_err("{forwarded} should decline instead of being signed");
|
||||
assert!(
|
||||
matches!(error, Error::Unsupported(_)),
|
||||
"{forwarded} declined as {error:?}, which the host would not fall back on"
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_bedrock_deployment_bearer_outranks_a_forwarded_authorization() {
|
||||
// `get_request_headers` assigns `headers["Authorization"]` unconditionally
|
||||
// once a bearer token resolves, so the deployment's identity wins on
|
||||
// Python. Keeping the caller's would authorize and bill the call as a
|
||||
// different principal, and only when the deployment carries `rust: true`.
|
||||
let mut call = request(
|
||||
"bedrock/us-east-1/anthropic.claude-v2",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"maxTokens": 16}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([(
|
||||
"Authorization".to_string(),
|
||||
json!("Bearer caller-supplied"),
|
||||
)]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let authorizations: Vec<_> = prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.filter(|(name, _)| name.eq_ignore_ascii_case("authorization"))
|
||||
.map(|(_, value)| value.as_str())
|
||||
.collect();
|
||||
assert_eq!(
|
||||
authorizations,
|
||||
vec!["Bearer sk-test"],
|
||||
"the deployment token must be the only authorization on the wire"
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn an_anthropic_forwarded_oauth_bearer_still_outranks_the_resolved_key() {
|
||||
// The opposite precedence, and deliberate: Anthropic's own transform
|
||||
// honours a forwarded OAuth bearer, so the Bedrock fix above must not be
|
||||
// generalized into a rule that the configured key always wins.
|
||||
//
|
||||
// An OAuth bearer is the whole of that exception. This forwarded a plain
|
||||
// `x-api-key` until round 17, which read as the same claim and was not:
|
||||
// Python overwrites a forwarded `x-api-key` with the deployment's.
|
||||
let mut call = request(
|
||||
"claude-sonnet-4-5",
|
||||
Some("anthropic"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
);
|
||||
call.extra_headers = Some(Map::from_iter([(
|
||||
"authorization".to_string(),
|
||||
json!("Bearer sk-ant-oat01-forwarded"),
|
||||
)]));
|
||||
let prepared = prepare_chat_completions_call(call).expect("prepares");
|
||||
let keys: Vec<_> = prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.filter(|(name, _)| name.eq_ignore_ascii_case("x-api-key"))
|
||||
.map(|(_, value)| value.as_str())
|
||||
.collect();
|
||||
assert!(keys.is_empty(), "got {:?}", prepared.upstream_headers);
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, value)| name.eq_ignore_ascii_case("authorization")
|
||||
&& value == "Bearer sk-ant-oat01-forwarded")
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_bedrock_api_key_is_sent_as_a_bearer_token_instead_of_being_signed() {
|
||||
// The configured bearer identity has its own account and quota boundary,
|
||||
// so a request carrying one must not be signed as whatever principal the
|
||||
// host's AWS credentials resolve to.
|
||||
let prepared = prepare_chat_completions_call(request(
|
||||
"bedrock/us-east-1/anthropic.claude-v2",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"maxTokens": 16}),
|
||||
))
|
||||
.expect("prepares");
|
||||
assert_eq!(
|
||||
prepared.auth,
|
||||
RequestAuth::Bearer {
|
||||
token: "sk-test".to_string()
|
||||
}
|
||||
);
|
||||
assert!(
|
||||
prepared
|
||||
.upstream_headers
|
||||
.iter()
|
||||
.any(|(name, value)| name.eq_ignore_ascii_case("authorization")
|
||||
&& value == "Bearer sk-test"),
|
||||
"prepare did not carry the bearer token"
|
||||
);
|
||||
}
|
||||
|
||||
fn decline_reason(
|
||||
model: &str,
|
||||
provider: Option<&str>,
|
||||
messages: Value,
|
||||
params: Value,
|
||||
) -> Option<&'static str> {
|
||||
let params = match params {
|
||||
Value::Object(map) => map,
|
||||
other => panic!("params must be an object, got {other}"),
|
||||
};
|
||||
super::chat_completions_decline_reason(model, provider, messages, ¶ms)
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn the_gate_accepts_what_prepare_accepts() {
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
),
|
||||
None
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn the_gate_declines_without_resolving_credentials_or_calling_out() {
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"stream": true}),
|
||||
),
|
||||
Some("streaming")
|
||||
);
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"openai/gpt-4o",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
),
|
||||
Some("provider is not on the rust chat completions path")
|
||||
);
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"claude-sonnet-4-5",
|
||||
None,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({}),
|
||||
),
|
||||
Some("provider is not on the rust chat completions path")
|
||||
);
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
json!("nope"),
|
||||
json!({})
|
||||
),
|
||||
Some("unreadable message list")
|
||||
);
|
||||
assert_eq!(
|
||||
decline_reason("anthropic/claude-sonnet-4-5", None, json!([]), json!({})),
|
||||
Some("empty message list")
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn the_gate_agrees_with_prepare_on_every_case_it_accepts() {
|
||||
// A gate that accepts what prepare then declines would make the host emit
|
||||
// its pre-call logging on a path that falls back, so pin the agreement.
|
||||
for (messages, params) in [
|
||||
(
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 8}),
|
||||
),
|
||||
(
|
||||
json!([{"role": "system", "content": "s"}, {"role": "user", "content": "hi"}]),
|
||||
json!({"temperature": 0.1}),
|
||||
),
|
||||
(
|
||||
json!([{"role": "user", "content": "hi"}, {"role": "assistant", "content": "yo"}]),
|
||||
json!({}),
|
||||
),
|
||||
] {
|
||||
assert_eq!(
|
||||
decline_reason(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
messages.clone(),
|
||||
params.clone()
|
||||
),
|
||||
None,
|
||||
"gate declined {messages}"
|
||||
);
|
||||
prepare_chat_completions_call(request(
|
||||
"anthropic/claude-sonnet-4-5",
|
||||
None,
|
||||
messages.clone(),
|
||||
params,
|
||||
))
|
||||
.unwrap_or_else(|error| panic!("prepare declined {messages}: {error}"));
|
||||
}
|
||||
}
|
||||
|
||||
mod round_trip {
|
||||
use tokio::{
|
||||
io::{AsyncReadExt, AsyncWriteExt},
|
||||
net::{TcpListener, TcpStream},
|
||||
};
|
||||
|
||||
use super::*;
|
||||
use crate::chat_completions::chat_completions;
|
||||
|
||||
async fn read_http_request(socket: &mut TcpStream) -> String {
|
||||
let mut request = Vec::new();
|
||||
let mut buffer = [0_u8; 1024];
|
||||
let header_end = loop {
|
||||
let n = socket.read(&mut buffer).await.expect("reads request");
|
||||
if n == 0 {
|
||||
break request.len();
|
||||
}
|
||||
request.extend_from_slice(&buffer[..n]);
|
||||
if let Some(position) = request.windows(4).position(|window| window == b"\r\n\r\n") {
|
||||
break position + 4;
|
||||
}
|
||||
};
|
||||
let headers = String::from_utf8_lossy(&request[..header_end]);
|
||||
let content_length = headers
|
||||
.lines()
|
||||
.find_map(|line| {
|
||||
let (name, value) = line.split_once(':')?;
|
||||
name.eq_ignore_ascii_case("content-length")
|
||||
.then(|| value.trim().parse::<usize>().ok())
|
||||
.flatten()
|
||||
})
|
||||
.unwrap_or(0);
|
||||
while request.len().saturating_sub(header_end) < content_length {
|
||||
let n = socket.read(&mut buffer).await.expect("reads body");
|
||||
if n == 0 {
|
||||
break;
|
||||
}
|
||||
request.extend_from_slice(&buffer[..n]);
|
||||
}
|
||||
String::from_utf8(request).expect("request is utf8")
|
||||
}
|
||||
|
||||
fn http_response(status: &str, body: &str) -> String {
|
||||
format!(
|
||||
"HTTP/1.1 {status}\r\ncontent-type: application/json\r\ncontent-length: {}\r\nconnection: close\r\n\r\n{}",
|
||||
body.len(),
|
||||
body
|
||||
)
|
||||
}
|
||||
|
||||
/// Serve one request from a stub upstream and hand back what it received.
|
||||
async fn serve_once(
|
||||
status: &'static str,
|
||||
body: &'static str,
|
||||
) -> (String, tokio::task::JoinHandle<String>) {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
|
||||
let port = listener.local_addr().expect("addr").port();
|
||||
let handle = tokio::spawn(async move {
|
||||
let (mut socket, _) = listener.accept().await.expect("accepts");
|
||||
let received = read_http_request(&mut socket).await;
|
||||
socket
|
||||
.write_all(http_response(status, body).as_bytes())
|
||||
.await
|
||||
.expect("writes response");
|
||||
socket.flush().await.expect("flushes");
|
||||
received
|
||||
});
|
||||
(format!("http://127.0.0.1:{port}/v1/messages"), handle)
|
||||
}
|
||||
|
||||
fn call(api_base: &str, messages: Value, params: Value) -> ChatCompletionsRequest<'_> {
|
||||
ChatCompletionsRequest {
|
||||
model: "anthropic/claude-sonnet-4-5",
|
||||
messages,
|
||||
optional_params: match params {
|
||||
Value::Object(map) => map,
|
||||
other => panic!("params must be an object, got {other}"),
|
||||
},
|
||||
api_key: Some("sk-test"),
|
||||
api_base: Some(api_base),
|
||||
custom_llm_provider: None,
|
||||
extra_headers: None,
|
||||
timeout: Some(std::time::Duration::from_secs(10)),
|
||||
}
|
||||
}
|
||||
|
||||
const GOOD_BODY: &str = r#"{"id":"msg_1","type":"message","role":"assistant","model":"claude-sonnet-4-5-20260101","content":[{"type":"text","text":"hello"}],"stop_reason":"end_turn","stop_sequence":null,"usage":{"input_tokens":11,"output_tokens":4}}"#;
|
||||
|
||||
#[tokio::test]
|
||||
async fn round_trip_sends_the_translated_body_and_normalizes_the_response() {
|
||||
let (api_base, handle) = serve_once("200 OK", GOOD_BODY).await;
|
||||
let response = chat_completions(call(
|
||||
&api_base,
|
||||
json!([
|
||||
{"role": "system", "content": "be terse"},
|
||||
{"role": "user", "content": "hi"}
|
||||
]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.await
|
||||
.expect("call succeeds");
|
||||
|
||||
let received = handle.await.expect("server task");
|
||||
let sent: Value = serde_json::from_str(
|
||||
received
|
||||
.split_once("\r\n\r\n")
|
||||
.expect("request has a body")
|
||||
.1,
|
||||
)
|
||||
.expect("body is json");
|
||||
assert_eq!(
|
||||
sent["messages"],
|
||||
json!([{"role": "user", "content": [{"type": "text", "text": "hi"}]}])
|
||||
);
|
||||
assert_eq!(
|
||||
sent["system"],
|
||||
json!([{"type": "text", "text": "be terse"}])
|
||||
);
|
||||
assert_eq!(sent["max_tokens"], json!(16));
|
||||
assert!(received.to_lowercase().contains("x-api-key: sk-test"));
|
||||
|
||||
assert_eq!(
|
||||
response.choices[0].message.content.as_deref(),
|
||||
Some("hello")
|
||||
);
|
||||
assert_eq!(response.usage.total_tokens, 15);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_response_it_cannot_normalize_is_reported_as_already_sent() {
|
||||
// The provider was called and billed, so the host must not retry this
|
||||
// on its own path. `MissingField` here would read as a pre-send
|
||||
// decline and be retried; `InvalidResponse` cannot.
|
||||
const NO_USAGE: &str =
|
||||
r#"{"model":"m","content":[{"type":"text","text":"hi"}],"stop_reason":"end_turn"}"#;
|
||||
let (api_base, handle) = serve_once("200 OK", NO_USAGE).await;
|
||||
let err = chat_completions(call(
|
||||
&api_base,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.await
|
||||
.expect_err("response cannot be normalized");
|
||||
handle.await.expect("server task");
|
||||
assert!(
|
||||
matches!(err, Error::InvalidResponse(_)),
|
||||
"expected a post-send error, got {err:?}"
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_tool_use_block_in_the_response_is_also_reported_as_already_sent() {
|
||||
const TOOL_USE: &str = r#"{"model":"m","content":[{"type":"tool_use","id":"t","name":"f","input":{}}],"stop_reason":"tool_use","usage":{"input_tokens":1,"output_tokens":1}}"#;
|
||||
let (api_base, handle) = serve_once("200 OK", TOOL_USE).await;
|
||||
let err = chat_completions(call(
|
||||
&api_base,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.await
|
||||
.expect_err("response cannot be normalized");
|
||||
handle.await.expect("server task");
|
||||
assert!(
|
||||
matches!(err, Error::InvalidResponse(_)),
|
||||
"expected a post-send error, got {err:?}"
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn an_upstream_error_status_keeps_its_code() {
|
||||
let (api_base, handle) =
|
||||
serve_once("429 Too Many Requests", r#"{"error":"slow down"}"#).await;
|
||||
let err = chat_completions(call(
|
||||
&api_base,
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.await
|
||||
.expect_err("upstream rejects");
|
||||
handle.await.expect("server task");
|
||||
assert!(
|
||||
matches!(
|
||||
err,
|
||||
Error::Transport(litellm_http::transport::Error::Http { status: 429, .. })
|
||||
),
|
||||
"expected a 429, got {err:?}"
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_connection_that_is_never_established_declines_instead_of_failing() {
|
||||
// Nothing was sent, so nothing was billed and the host can still serve
|
||||
// the request. Classing this with the post-send failures would turn a
|
||||
// recoverable fallback into a user-facing error on exactly the
|
||||
// deployments whose transport is configured only on the Python client.
|
||||
let port = {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
|
||||
listener.local_addr().expect("has an address").port()
|
||||
// Dropped here, so the port is closed and the connect is refused.
|
||||
};
|
||||
let err = chat_completions(call(
|
||||
&format!("http://127.0.0.1:{port}/v1/messages"),
|
||||
json!([{"role": "user", "content": "hi"}]),
|
||||
json!({"max_tokens": 16}),
|
||||
))
|
||||
.await
|
||||
.expect_err("nothing is listening");
|
||||
assert!(
|
||||
matches!(
|
||||
err,
|
||||
Error::Transport(litellm_http::transport::Error::Connect(_))
|
||||
),
|
||||
"expected a pre-send connect failure, got {err:?}"
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn response_errors_collapse_to_one_variant_that_can_only_mean_already_sent() {
|
||||
use crate::chat_completions::handler::as_response_error;
|
||||
|
||||
for original in [
|
||||
Error::MissingField("usage"),
|
||||
Error::Unsupported("non-text response content block"),
|
||||
Error::InvalidRequest("whatever".to_string()),
|
||||
Error::Auth(litellm_auth::Error::InvalidHeader),
|
||||
] {
|
||||
let label = format!("{original:?}");
|
||||
assert!(
|
||||
matches!(as_response_error(original), Error::InvalidResponse(_)),
|
||||
"{label} must not stay retryable once the provider has answered"
|
||||
);
|
||||
}
|
||||
// An upstream status is already unambiguous, so it survives intact.
|
||||
assert!(matches!(
|
||||
as_response_error(Error::Transport(litellm_http::transport::Error::Http {
|
||||
status: 500,
|
||||
body: "boom".to_string()
|
||||
})),
|
||||
Error::Transport(litellm_http::transport::Error::Http { status: 500, .. })
|
||||
));
|
||||
}
|
||||
}
|
||||
|
|
@ -26,3 +26,186 @@ pub(super) fn string_headers(
|
|||
) -> Result<Vec<(String, String)>, Error> {
|
||||
shared_string_headers(HEADER_CONTEXT, extra_headers).map_err(Error::from)
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use std::{sync::Arc, time::Duration};
|
||||
|
||||
use futures_util::future::BoxFuture;
|
||||
use litellm_secrets::{SecretValue, source::SecretSource};
|
||||
use serde_json::{Value, json};
|
||||
use tokio::{
|
||||
io::{AsyncReadExt, AsyncWriteExt},
|
||||
net::{TcpListener, TcpStream},
|
||||
};
|
||||
|
||||
use super::{messages_provider_config, string_headers, truncate_error_body};
|
||||
use crate::messages::{
|
||||
Error,
|
||||
route::{LocalMessagesHost, MessagesCall, MessagesOutput, messages_machine},
|
||||
types::MessagesShaping,
|
||||
};
|
||||
|
||||
struct RecordingSecrets {
|
||||
values: Vec<(&'static str, String)>,
|
||||
requested: std::sync::Mutex<Vec<String>>,
|
||||
}
|
||||
|
||||
impl SecretSource for RecordingSecrets {
|
||||
fn get_secret_str<'a>(
|
||||
&'a self,
|
||||
name: &'a str,
|
||||
) -> BoxFuture<'a, Result<Option<SecretValue>, litellm_secrets::Error>> {
|
||||
Box::pin(async move {
|
||||
self.requested.lock().unwrap().push(name.to_string());
|
||||
Ok(self
|
||||
.values
|
||||
.iter()
|
||||
.find(|(key, _)| *key == name)
|
||||
.map(|(_, value)| SecretValue::new(value.clone())))
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
fn secrets_call() -> MessagesCall {
|
||||
let Value::Object(body) = json!({
|
||||
"model": "claude-sonnet-4-5",
|
||||
"max_tokens": 16,
|
||||
"messages": [{"role": "user", "content": "hi"}]
|
||||
}) else {
|
||||
unreachable!("literal object")
|
||||
};
|
||||
MessagesCall {
|
||||
model: "claude-sonnet-4-5".into(),
|
||||
body,
|
||||
api_key: None,
|
||||
api_base: None,
|
||||
custom_llm_provider: Some("anthropic".into()),
|
||||
extra_headers: None,
|
||||
provider_specific_header: None,
|
||||
timeout: Some(Duration::from_secs(5)),
|
||||
shaping: MessagesShaping::default(),
|
||||
}
|
||||
}
|
||||
|
||||
async fn read_http_request(socket: &mut TcpStream) -> String {
|
||||
let mut request = Vec::new();
|
||||
let mut buffer = [0_u8; 1024];
|
||||
let header_end = loop {
|
||||
let n = socket.read(&mut buffer).await.expect("reads request");
|
||||
if n == 0 {
|
||||
break request.len();
|
||||
}
|
||||
request.extend_from_slice(&buffer[..n]);
|
||||
if let Some(position) = request.windows(4).position(|window| window == b"\r\n\r\n") {
|
||||
break position + 4;
|
||||
}
|
||||
};
|
||||
let headers = String::from_utf8_lossy(&request[..header_end]);
|
||||
let content_length = headers
|
||||
.lines()
|
||||
.find_map(|line| {
|
||||
let (name, value) = line.split_once(':')?;
|
||||
name.eq_ignore_ascii_case("content-length")
|
||||
.then(|| value.trim().parse::<usize>().ok())
|
||||
.flatten()
|
||||
})
|
||||
.unwrap_or(0);
|
||||
while request.len().saturating_sub(header_end) < content_length {
|
||||
let n = socket.read(&mut buffer).await.expect("reads body");
|
||||
if n == 0 {
|
||||
break;
|
||||
}
|
||||
request.extend_from_slice(&buffer[..n]);
|
||||
}
|
||||
String::from_utf8(request).expect("request is utf8")
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn route_reads_the_provider_credential_and_base_from_the_secret_source() {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
|
||||
let addr = listener.local_addr().expect("addr");
|
||||
let server = tokio::spawn(async move {
|
||||
let (mut socket, _) = listener.accept().await.expect("accepts request");
|
||||
let request = read_http_request(&mut socket).await;
|
||||
let response_body = r#"{"id":"msg_1","type":"message","role":"assistant","content":[],"model":"claude-sonnet-4-5","stop_reason":"end_turn","usage":{"input_tokens":1,"output_tokens":1}}"#;
|
||||
let response = format!(
|
||||
"HTTP/1.1 200 OK\r\ncontent-type: application/json\r\ncontent-length: {}\r\nconnection: close\r\n\r\n{}",
|
||||
response_body.len(),
|
||||
response_body
|
||||
);
|
||||
socket
|
||||
.write_all(response.as_bytes())
|
||||
.await
|
||||
.expect("writes response");
|
||||
request
|
||||
});
|
||||
let secrets = Arc::new(RecordingSecrets {
|
||||
values: vec![
|
||||
("ANTHROPIC_API_KEY", "sk-from-manager".to_string()),
|
||||
("ANTHROPIC_BASE_URL", format!("http://{addr}")),
|
||||
],
|
||||
requested: std::sync::Mutex::new(Vec::new()),
|
||||
});
|
||||
|
||||
let output = litellm_host::run::run(
|
||||
messages_machine(secrets.clone()),
|
||||
&LocalMessagesHost::new(secrets_call()),
|
||||
)
|
||||
.await
|
||||
.expect("messages request succeeds");
|
||||
|
||||
assert!(matches!(output, MessagesOutput::Message(_)));
|
||||
let request = server.await.expect("server task completes");
|
||||
assert!(
|
||||
request
|
||||
.to_ascii_lowercase()
|
||||
.contains("x-api-key: sk-from-manager"),
|
||||
"{request}"
|
||||
);
|
||||
let requested = secrets.requested.lock().unwrap().clone();
|
||||
assert_eq!(
|
||||
requested,
|
||||
messages_provider_config("anthropic")
|
||||
.unwrap()
|
||||
.secret_names()
|
||||
.iter()
|
||||
.map(ToString::to_string)
|
||||
.collect::<Vec<_>>()
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn provider_config_resolves_anthropic_and_azure_ai() {
|
||||
assert!(messages_provider_config("anthropic").is_some());
|
||||
assert!(messages_provider_config("azure_ai").is_some());
|
||||
assert!(messages_provider_config("openai").is_none());
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn truncate_error_body_caps_long_payloads() {
|
||||
let body = "x".repeat(400);
|
||||
let truncated = truncate_error_body(&body);
|
||||
assert!(truncated.ends_with("... (truncated)"));
|
||||
let prefix_chars = truncated
|
||||
.strip_suffix("... (truncated)")
|
||||
.expect("truncated marker present")
|
||||
.chars()
|
||||
.count();
|
||||
assert_eq!(prefix_chars, 256);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn string_headers_rejects_non_string_values() {
|
||||
let headers = json!({"x-count": 3}).as_object().unwrap().clone();
|
||||
let err = string_headers(Some(headers)).expect_err("non-string header rejected");
|
||||
assert_eq!(
|
||||
err,
|
||||
Error::Headers(litellm_http::request::HeaderError {
|
||||
context: "messages",
|
||||
name: "x-count".to_string(),
|
||||
actual: "number",
|
||||
})
|
||||
);
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -46,6 +46,3 @@ pub async fn messages(request: MessagesRequest<'_>) -> Result<AnthropicMessagesR
|
|||
)),
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests;
|
||||
|
|
|
|||
|
|
@ -248,3 +248,160 @@ mod tests {
|
|||
}
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod document_tests {
|
||||
use litellm_host::event::WireRequest;
|
||||
use litellm_llms::base_llm::ocr::error::Error;
|
||||
use rstest::rstest;
|
||||
use serde_json::{Value, json};
|
||||
|
||||
use crate::ocr::route::LocalOcrHost;
|
||||
use crate::ocr::test_support::{
|
||||
MockResponse, SERVED_DOCUMENT, document_server, mock_server, perform_ocr_with,
|
||||
request_body, wire_request_with_document,
|
||||
};
|
||||
|
||||
#[derive(Clone, Copy, Debug)]
|
||||
enum Route {
|
||||
Mistral,
|
||||
AzureAi,
|
||||
VertexMistral,
|
||||
AzureCohereParse,
|
||||
Cohere,
|
||||
}
|
||||
|
||||
impl Route {
|
||||
fn model(self) -> &'static str {
|
||||
match self {
|
||||
Self::Mistral => "mistral/model",
|
||||
Self::AzureAi => "azure_ai/model",
|
||||
Self::VertexMistral => "vertex_ai/mistral-ocr-maas",
|
||||
Self::AzureCohereParse => "azure_ai/cohere-parse",
|
||||
Self::Cohere => "cohere/model",
|
||||
}
|
||||
}
|
||||
|
||||
fn document_type(self) -> &'static str {
|
||||
match self {
|
||||
Self::Mistral | Self::AzureAi | Self::VertexMistral => "document_url",
|
||||
Self::AzureCohereParse | Self::Cohere => "image_url",
|
||||
}
|
||||
}
|
||||
|
||||
fn options(self) -> Value {
|
||||
match self {
|
||||
Self::Mistral | Self::AzureAi => json!({"pages": [0]}),
|
||||
Self::VertexMistral => json!({"pages": [0], "vertex_project": "project-1"}),
|
||||
Self::AzureCohereParse | Self::Cohere => json!({"output_format": "markdown"}),
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// What the host does to the wire request in `before_send`.
|
||||
#[derive(Clone, Copy, Debug)]
|
||||
enum Host {
|
||||
Detached,
|
||||
ReplacesDocument,
|
||||
}
|
||||
|
||||
const REPLACED_DOCUMENT: &str = "data:image/png;base64,cmVwbGFjZWQ=";
|
||||
|
||||
impl Host {
|
||||
fn before_send(self, wire: WireRequest) -> WireRequest {
|
||||
let Value::Object(fields) = wire.body else {
|
||||
return wire;
|
||||
};
|
||||
let body = fields
|
||||
.into_iter()
|
||||
.map(|(name, value)| match self {
|
||||
Self::Detached => (name, value),
|
||||
Self::ReplacesDocument if name == "document" => {
|
||||
let document_type = value["type"].clone();
|
||||
let key = document_type.as_str().unwrap_or_default().to_string();
|
||||
(name, json!({"type": document_type, key: REPLACED_DOCUMENT}))
|
||||
}
|
||||
Self::ReplacesDocument => (name, value),
|
||||
})
|
||||
.collect();
|
||||
WireRequest {
|
||||
body: Value::Object(body),
|
||||
..wire
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
struct Sent {
|
||||
result: Result<(), Error>,
|
||||
provider_body: Option<Value>,
|
||||
}
|
||||
|
||||
async fn send(route: Route, host: Host, document_base: &str) -> Sent {
|
||||
let (base, seen, provider) =
|
||||
mock_server(vec![MockResponse::json(json!({"pages": []}))]).await;
|
||||
let document_type = route.document_type();
|
||||
let document =
|
||||
json!({"type": document_type, document_type: format!("{document_base}/scan.png")});
|
||||
let request = wire_request_with_document(route.model(), &base, document, route.options());
|
||||
let local =
|
||||
LocalOcrHost::new(request).with_before_send(move |wire, _| Ok(host.before_send(wire)));
|
||||
let result = perform_ocr_with(local).await.map(|_| ());
|
||||
match result {
|
||||
Ok(()) => provider.await.unwrap(),
|
||||
Err(_) => provider.abort(),
|
||||
}
|
||||
let provider_body = seen
|
||||
.lock()
|
||||
.unwrap()
|
||||
.first()
|
||||
.map(|request| request_body(request));
|
||||
Sent {
|
||||
result,
|
||||
provider_body,
|
||||
}
|
||||
}
|
||||
|
||||
fn served_document_uri() -> String {
|
||||
use base64::Engine;
|
||||
format!(
|
||||
"data:image/png;base64,{}",
|
||||
base64::engine::general_purpose::STANDARD.encode(SERVED_DOCUMENT)
|
||||
)
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::azure_ai(Route::AzureAi)]
|
||||
#[case::vertex_mistral(Route::VertexMistral)]
|
||||
#[case::azure_cohere_parse(Route::AzureCohereParse)]
|
||||
#[tokio::test]
|
||||
async fn inlining_routes_send_the_downloaded_document(#[case] route: Route) {
|
||||
let (document_base, _documents) = document_server().await;
|
||||
let sent = send(route, Host::Detached, &document_base).await;
|
||||
sent.result.unwrap();
|
||||
assert_eq!(
|
||||
sent.provider_body.unwrap()["document"][route.document_type()],
|
||||
json!(served_document_uri())
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[tokio::test]
|
||||
async fn document_replaced_by_the_host_reaches_the_provider(
|
||||
#[values(
|
||||
Route::Mistral,
|
||||
Route::AzureAi,
|
||||
Route::VertexMistral,
|
||||
Route::AzureCohereParse,
|
||||
Route::Cohere
|
||||
)]
|
||||
route: Route,
|
||||
) {
|
||||
let (document_base, _documents) = document_server().await;
|
||||
let sent = send(route, Host::ReplacesDocument, &document_base).await;
|
||||
sent.result.unwrap();
|
||||
assert_eq!(
|
||||
sent.provider_body.unwrap()["document"][route.document_type()],
|
||||
json!(REPLACED_DOCUMENT)
|
||||
);
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -9,36 +9,210 @@ pub mod types;
|
|||
pub mod wire;
|
||||
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/aws_textract_ocr.rs"]
|
||||
mod aws_textract_tests;
|
||||
pub(crate) mod test_support {
|
||||
use std::sync::{Arc, Mutex};
|
||||
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/azure_ai_ocr.rs"]
|
||||
mod azure_ai_tests;
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/azure_document_intelligence_ocr.rs"]
|
||||
mod azure_document_intelligence_tests;
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/cohere_ocr.rs"]
|
||||
mod cohere_tests;
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/deepseek_ocr.rs"]
|
||||
mod deepseek_tests;
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/ocr/document.rs"]
|
||||
mod document_tests;
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/reducto_ocr.rs"]
|
||||
mod reducto_tests;
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/ocr/support.rs"]
|
||||
pub(crate) mod test_support;
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/ocr.rs"]
|
||||
pub(crate) mod tests;
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/vertex_ai_deepseek_ocr.rs"]
|
||||
mod vertex_ai_deepseek_tests;
|
||||
#[cfg(test)]
|
||||
#[path = "../../tests/vertex_ai_ocr.rs"]
|
||||
mod vertex_ai_tests;
|
||||
use futures_util::future::BoxFuture;
|
||||
use litellm_host::event::WireRequest;
|
||||
use litellm_llms::base_llm::ocr::{
|
||||
error::Error,
|
||||
handler::{CallHooks, OcrClient},
|
||||
transformation::LiteLLMOcrResponse,
|
||||
};
|
||||
use serde_json::{Value, json};
|
||||
use tokio::{
|
||||
io::{AsyncReadExt, AsyncWriteExt},
|
||||
net::TcpListener,
|
||||
};
|
||||
|
||||
use crate::ocr::{
|
||||
route::{LocalOcrHost, ocr_machine},
|
||||
types::LiteLLMOcrRequest,
|
||||
wire::{OcrWireRequest, decode_request},
|
||||
};
|
||||
|
||||
/// Stands in for a host with no hooks registered: the wire request goes out unchanged
|
||||
/// and response events go nowhere.
|
||||
pub(crate) struct NoHooks;
|
||||
|
||||
impl CallHooks<Error> for NoHooks {
|
||||
fn before_send(&self, wire: WireRequest) -> BoxFuture<'_, Result<WireRequest, Error>> {
|
||||
Box::pin(async move { Ok(wire) })
|
||||
}
|
||||
|
||||
fn response_received<'a>(&'a self, _body: &'a [u8]) -> BoxFuture<'a, Result<(), Error>> {
|
||||
Box::pin(async { Ok(()) })
|
||||
}
|
||||
}
|
||||
|
||||
pub(crate) fn ocr_client() -> OcrClient {
|
||||
let document_http = reqwest::Client::builder()
|
||||
.redirect(reqwest::redirect::Policy::none())
|
||||
.build()
|
||||
.expect("test document client builds");
|
||||
OcrClient::for_test(reqwest::Client::new(), document_http)
|
||||
}
|
||||
|
||||
pub(crate) async fn perform_ocr(
|
||||
request: LiteLLMOcrRequest,
|
||||
) -> Result<LiteLLMOcrResponse, Error> {
|
||||
crate::ocr::client::perform(&ocr_client(), request).await
|
||||
}
|
||||
|
||||
pub(crate) async fn perform_ocr_with(host: LocalOcrHost) -> Result<LiteLLMOcrResponse, Error> {
|
||||
litellm_host::run::run(ocr_machine(ocr_client()), &host).await
|
||||
}
|
||||
|
||||
pub(crate) fn wire_request(model: &str, base: &str, options: Value) -> LiteLLMOcrRequest {
|
||||
wire_request_with_document(
|
||||
model,
|
||||
base,
|
||||
json!({"type":"document_url","document_url":"data:application/pdf;base64,YWJj"}),
|
||||
options,
|
||||
)
|
||||
}
|
||||
|
||||
pub(crate) fn wire_request_with_document(
|
||||
model: &str,
|
||||
base: &str,
|
||||
document: Value,
|
||||
options: Value,
|
||||
) -> LiteLLMOcrRequest {
|
||||
decode_request(OcrWireRequest {
|
||||
model: model.into(),
|
||||
document,
|
||||
api_key: Some(litellm_auth::SecretValue::new("test-key")),
|
||||
api_base: Some(base.into()),
|
||||
custom_llm_provider: None,
|
||||
extra_headers: None,
|
||||
optional_params: options.as_object().unwrap().clone(),
|
||||
input_sources: Default::default(),
|
||||
timeout_seconds: Some(2.0),
|
||||
})
|
||||
.unwrap()
|
||||
}
|
||||
|
||||
pub(crate) fn resolved_request(
|
||||
request: LiteLLMOcrRequest,
|
||||
) -> crate::ocr::types::ResolvedOcrRequest {
|
||||
request
|
||||
.map_document(crate::ocr::document::prepare_document)
|
||||
.unwrap()
|
||||
}
|
||||
|
||||
pub(crate) fn with_source(request: LiteLLMOcrRequest, source: &str) -> LiteLLMOcrRequest {
|
||||
let request = resolved_request(request);
|
||||
let document = request.document.clone().with_source(source.into());
|
||||
request.with_document(document.into())
|
||||
}
|
||||
|
||||
pub(crate) fn request_body(request: &str) -> Value {
|
||||
serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap()
|
||||
}
|
||||
|
||||
pub(crate) const SERVED_DOCUMENT: &[u8] = b"\x89PNG served document";
|
||||
|
||||
/// Serves [`SERVED_DOCUMENT`] as `image/png` to every connection until aborted.
|
||||
pub(crate) async fn document_server() -> (String, tokio::task::JoinHandle<()>) {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
|
||||
let base = format!("http://{}", listener.local_addr().unwrap());
|
||||
let task = tokio::spawn(async move {
|
||||
loop {
|
||||
let (mut socket, _) = listener.accept().await.unwrap();
|
||||
let mut buffer = [0u8; 4096];
|
||||
let _ = socket.read(&mut buffer).await.unwrap();
|
||||
let head = format!(
|
||||
"HTTP/1.1 200 OK\r\nContent-Type: image/png\r\nContent-Length: {}\r\nConnection: close\r\n\r\n",
|
||||
SERVED_DOCUMENT.len()
|
||||
);
|
||||
socket.write_all(head.as_bytes()).await.unwrap();
|
||||
socket.write_all(SERVED_DOCUMENT).await.unwrap();
|
||||
}
|
||||
});
|
||||
(base, task)
|
||||
}
|
||||
|
||||
pub(crate) struct MockResponse {
|
||||
pub status: u16,
|
||||
pub headers: Vec<(&'static str, String)>,
|
||||
pub body: Value,
|
||||
}
|
||||
|
||||
impl MockResponse {
|
||||
pub fn json(body: Value) -> Self {
|
||||
Self {
|
||||
status: 200,
|
||||
headers: vec![],
|
||||
body,
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
pub(crate) async fn mock_server(
|
||||
responses: Vec<MockResponse>,
|
||||
) -> (String, Arc<Mutex<Vec<String>>>, tokio::task::JoinHandle<()>) {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
|
||||
let base = format!("http://{}", listener.local_addr().unwrap());
|
||||
let requests = Arc::new(Mutex::new(Vec::new()));
|
||||
let seen = requests.clone();
|
||||
let server_base = base.clone();
|
||||
let task = tokio::spawn(async move {
|
||||
for response in responses {
|
||||
let (mut socket, _) = listener.accept().await.unwrap();
|
||||
let mut bytes = Vec::new();
|
||||
let mut buffer = [0u8; 4096];
|
||||
let header_end = loop {
|
||||
let n = socket.read(&mut buffer).await.unwrap();
|
||||
assert!(n > 0);
|
||||
bytes.extend_from_slice(&buffer[..n]);
|
||||
if let Some(index) = bytes.windows(4).position(|s| s == b"\r\n\r\n") {
|
||||
break index + 4;
|
||||
}
|
||||
};
|
||||
let length = String::from_utf8_lossy(&bytes[..header_end])
|
||||
.lines()
|
||||
.find_map(|line| {
|
||||
let (name, value) = line.split_once(':')?;
|
||||
name.eq_ignore_ascii_case("content-length")
|
||||
.then(|| value.trim().parse::<usize>().unwrap())
|
||||
})
|
||||
.unwrap_or(0);
|
||||
while bytes.len() < header_end + length {
|
||||
let n = socket.read(&mut buffer).await.unwrap();
|
||||
assert!(n > 0);
|
||||
bytes.extend_from_slice(&buffer[..n]);
|
||||
}
|
||||
seen.lock()
|
||||
.unwrap()
|
||||
.push(String::from_utf8_lossy(&bytes).into_owned());
|
||||
let body = serde_json::to_vec(&response.body).unwrap();
|
||||
let headers = response
|
||||
.headers
|
||||
.into_iter()
|
||||
.map(|(name, value)| {
|
||||
format!("{name}: {}\r\n", value.replace("{base}", &server_base))
|
||||
})
|
||||
.collect::<String>();
|
||||
let head = format!(
|
||||
"HTTP/1.1 {} OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n{}\r\n",
|
||||
response.status,
|
||||
body.len(),
|
||||
headers
|
||||
);
|
||||
socket.write_all(head.as_bytes()).await.unwrap();
|
||||
socket.write_all(&body).await.unwrap();
|
||||
}
|
||||
});
|
||||
(base, requests, task)
|
||||
}
|
||||
|
||||
pub(crate) fn header<'a>(request: &'a str, name: &str) -> Option<&'a str> {
|
||||
request
|
||||
.lines()
|
||||
.take_while(|line| !line.is_empty())
|
||||
.find_map(|line| {
|
||||
let (key, value) = line.split_once(':')?;
|
||||
key.eq_ignore_ascii_case(name).then(|| value.trim())
|
||||
})
|
||||
}
|
||||
}
|
||||
|
|
|
|||
File diff suppressed because it is too large
Load diff
|
|
@ -4,11 +4,9 @@ use std::{
|
|||
thread,
|
||||
};
|
||||
|
||||
use litellm_core::audio_transcription::{audio_transcription, types::AudioTranscriptionRequest};
|
||||
use serde_json::{Map, json};
|
||||
|
||||
use super::audio_transcription;
|
||||
use crate::audio_transcription::types::AudioTranscriptionRequest;
|
||||
|
||||
#[tokio::test]
|
||||
async fn bedrock_request_is_signed_and_contains_audio() {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").expect("listener");
|
||||
|
|
@ -1,193 +0,0 @@
|
|||
use std::{collections::BTreeMap, time::SystemTime};
|
||||
|
||||
use litellm_auth_aws::{Credentials, aws_signature_headers, sign_post};
|
||||
use litellm_llms::base_llm::ocr::error::Error;
|
||||
use serde_json::{Value, json};
|
||||
use time::{PrimitiveDateTime, format_description};
|
||||
|
||||
use crate::ocr::{
|
||||
route::LocalOcrHost,
|
||||
test_support::{
|
||||
MockResponse, header, mock_server, perform_ocr_with, request_body,
|
||||
wire_request_with_document,
|
||||
},
|
||||
types::LiteLLMOcrRequest,
|
||||
};
|
||||
|
||||
const ACCESS_KEY_ID: &str = "AKIDEXAMPLE";
|
||||
const SECRET_ACCESS_KEY: &str = "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY";
|
||||
|
||||
fn textract_request(base: &str) -> LiteLLMOcrRequest {
|
||||
textract_request_for("aws_textract/detect-document-text", base)
|
||||
}
|
||||
|
||||
fn textract_request_for(model: &str, base: &str) -> LiteLLMOcrRequest {
|
||||
wire_request_with_document(
|
||||
model,
|
||||
&format!("{base}/"),
|
||||
json!({"type": "image_url", "image_url": "data:image/png;base64,b3JpZ2luYWw="}),
|
||||
json!({
|
||||
"aws_access_key_id": ACCESS_KEY_ID,
|
||||
"aws_secret_access_key": SECRET_ACCESS_KEY,
|
||||
"aws_region_name": "eu-west-1"
|
||||
}),
|
||||
)
|
||||
}
|
||||
|
||||
fn textract_response() -> MockResponse {
|
||||
MockResponse::json(json!({
|
||||
"DocumentMetadata": {"Pages": 1},
|
||||
"Blocks": [{"BlockType": "PAGE"}, {"BlockType": "LINE", "Text": "Invoice 12345"}]
|
||||
}))
|
||||
}
|
||||
|
||||
/// Recomputes SigV4 over the bytes the server received, at the time the client claimed.
|
||||
fn expected_authorization(url: &str, raw_request: &str) -> String {
|
||||
let format =
|
||||
format_description::parse_borrowed::<2>("[year][month][day]T[hour][minute][second]Z")
|
||||
.unwrap();
|
||||
let signed_at: SystemTime =
|
||||
PrimitiveDateTime::parse(header(raw_request, "x-amz-date").unwrap(), &format)
|
||||
.unwrap()
|
||||
.assume_utc()
|
||||
.into();
|
||||
let headers: BTreeMap<String, String> = ["content-type", "x-amz-target"]
|
||||
.into_iter()
|
||||
.map(|name| {
|
||||
(
|
||||
name.to_string(),
|
||||
header(raw_request, name).unwrap().to_string(),
|
||||
)
|
||||
})
|
||||
.collect();
|
||||
let body = raw_request.split_once("\r\n\r\n").unwrap().1;
|
||||
sign_post(
|
||||
url,
|
||||
body.as_bytes(),
|
||||
&aws_signature_headers(&headers),
|
||||
"eu-west-1",
|
||||
"textract",
|
||||
&Credentials::new(ACCESS_KEY_ID, SECRET_ACCESS_KEY, None, None, "test"),
|
||||
signed_at,
|
||||
)
|
||||
.unwrap()["Authorization"]
|
||||
.clone()
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn the_request_is_signed_for_textract_and_lines_become_the_page() {
|
||||
let (base, seen, server) = mock_server(vec![textract_response()]).await;
|
||||
|
||||
let response = perform_ocr_with(LocalOcrHost::new(textract_request(&base)))
|
||||
.await
|
||||
.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
let raw = seen.lock().unwrap()[0].clone();
|
||||
assert_eq!(
|
||||
header(&raw, "x-amz-target"),
|
||||
Some("Textract.DetectDocumentText")
|
||||
);
|
||||
assert_eq!(
|
||||
header(&raw, "content-type"),
|
||||
Some("application/x-amz-json-1.1")
|
||||
);
|
||||
assert_eq!(
|
||||
request_body(&raw),
|
||||
json!({"Document": {"Bytes": "b3JpZ2luYWw="}})
|
||||
);
|
||||
assert_eq!(
|
||||
header(&raw, "authorization"),
|
||||
Some(expected_authorization(&format!("{base}/"), &raw).as_str())
|
||||
);
|
||||
assert_eq!(response.pages[0].markdown, "Invoice 12345");
|
||||
assert_eq!(response.usage_info.unwrap().pages_processed, Some(1));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_body_rewritten_by_before_send_is_what_gets_signed_and_sent() {
|
||||
let (base, seen, server) = mock_server(vec![textract_response()]).await;
|
||||
let host = LocalOcrHost::new(textract_request(&base)).with_before_send(|mut wire, _| {
|
||||
assert!(
|
||||
!wire
|
||||
.headers
|
||||
.iter()
|
||||
.any(|(name, _)| name.eq_ignore_ascii_case("authorization")),
|
||||
"the hook ran after signing"
|
||||
);
|
||||
wire.body["Document"]["Bytes"] = Value::from("cmVkYWN0ZWQ=");
|
||||
Ok(wire)
|
||||
});
|
||||
|
||||
perform_ocr_with(host).await.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
let raw = seen.lock().unwrap()[0].clone();
|
||||
assert_eq!(
|
||||
request_body(&raw),
|
||||
json!({"Document": {"Bytes": "cmVkYWN0ZWQ="}})
|
||||
);
|
||||
assert_eq!(
|
||||
header(&raw, "authorization"),
|
||||
Some(expected_authorization(&format!("{base}/"), &raw).as_str())
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn a_multi_page_rejection_reaches_the_caller_with_the_single_page_limit() {
|
||||
let (base, _, server) = mock_server(vec![MockResponse {
|
||||
status: 400,
|
||||
headers: vec![],
|
||||
body: json!({
|
||||
"__type": "UnsupportedDocumentException",
|
||||
"Message": "Request has unsupported document format"
|
||||
}),
|
||||
}])
|
||||
.await;
|
||||
|
||||
let error = perform_ocr_with(LocalOcrHost::new(textract_request(&base)))
|
||||
.await
|
||||
.unwrap_err();
|
||||
server.await.unwrap();
|
||||
|
||||
let Error::Provider { status, body, .. } = error else {
|
||||
panic!("expected a provider error, got {error:?}");
|
||||
};
|
||||
assert_eq!(status, 400);
|
||||
assert!(
|
||||
body.contains("multi-page documents are not supported"),
|
||||
"{body}"
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn analyze_document_asks_for_layout_and_tables_and_returns_markdown() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"DocumentMetadata": {"Pages": 1},
|
||||
"Blocks": [
|
||||
{"Id": "l1", "BlockType": "LINE", "Text": "Quarterly Report"},
|
||||
{"Id": "t", "BlockType": "LAYOUT_TITLE",
|
||||
"Relationships": [{"Type": "CHILD", "Ids": ["l1"]}]}
|
||||
]
|
||||
}))])
|
||||
.await;
|
||||
let request = textract_request_for("aws_textract/analyze-document", &base);
|
||||
|
||||
let response = perform_ocr_with(LocalOcrHost::new(request)).await.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
let raw = seen.lock().unwrap()[0].clone();
|
||||
assert_eq!(
|
||||
header(&raw, "x-amz-target"),
|
||||
Some("Textract.AnalyzeDocument")
|
||||
);
|
||||
assert_eq!(
|
||||
request_body(&raw)["FeatureTypes"],
|
||||
json!(["LAYOUT", "TABLES"])
|
||||
);
|
||||
assert_eq!(
|
||||
header(&raw, "authorization"),
|
||||
Some(expected_authorization(&format!("{base}/"), &raw).as_str())
|
||||
);
|
||||
assert_eq!(response.pages[0].markdown, "# Quarterly Report");
|
||||
}
|
||||
|
|
@ -1,293 +0,0 @@
|
|||
use litellm_llms::base_llm::ocr::error::Error;
|
||||
use serde_json::{Value, json};
|
||||
|
||||
use super::test_support::{MockResponse, mock_server, perform_ocr, perform_ocr_with, wire_request};
|
||||
use crate::ocr::route::LocalOcrHost;
|
||||
|
||||
#[tokio::test]
|
||||
async fn facade_executes_azure_mistral_with_prepared_auth() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"pages":[{"index":0,"markdown":"hello"}],
|
||||
"usage_info":{"pages_processed":1}
|
||||
}))])
|
||||
.await;
|
||||
let mut request = wire_request(
|
||||
"azure_ai/model",
|
||||
&base,
|
||||
json!({"include_image_base64":true}),
|
||||
);
|
||||
request.credentials.api_key = None;
|
||||
request.transport.extra_headers = vec![(
|
||||
"Authorization".into(),
|
||||
"Bearer python-prepared-token".into(),
|
||||
)];
|
||||
|
||||
let result = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(result.pages[0].markdown, "hello");
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 1);
|
||||
assert!(requests[0].starts_with("POST /providers/mistral/azure/ocr "));
|
||||
assert!(
|
||||
requests[0]
|
||||
.to_ascii_lowercase()
|
||||
.contains("authorization: bearer python-prepared-token\r\n")
|
||||
);
|
||||
let body: Value = serde_json::from_str(requests[0].split_once("\r\n\r\n").unwrap().1).unwrap();
|
||||
assert_eq!(
|
||||
body,
|
||||
json!({
|
||||
"model":"model",
|
||||
"document":{"type":"document_url","document_url":"data:application/pdf;base64,YWJj"},
|
||||
"include_image_base64":true
|
||||
})
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn facade_acquires_supplied_entra_token_for_final_request() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({"pages":[]}))]).await;
|
||||
let mut request = wire_request(
|
||||
"azure_ai/model",
|
||||
&base,
|
||||
json!({"azure_ad_token":"rust-owned-token"}),
|
||||
);
|
||||
request.credentials.api_key = None;
|
||||
|
||||
perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 1);
|
||||
assert!(
|
||||
requests[0]
|
||||
.to_ascii_lowercase()
|
||||
.contains("authorization: bearer rust-owned-token\r\n")
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn rejects_non_inline_body_after_guardrails() {
|
||||
let request = wire_request("azure_ai/model", "http://127.0.0.1:1", json!({}));
|
||||
let host = LocalOcrHost::new(request).with_before_send(|mut wire, _| {
|
||||
wire.body["document"] = json!({
|
||||
"type":"document_url",
|
||||
"document_url":"https://example.com/not-inline.pdf"
|
||||
});
|
||||
Ok(wire)
|
||||
});
|
||||
let error = perform_ocr_with(host).await.unwrap_err();
|
||||
assert!(error.to_string().contains("data URI"));
|
||||
}
|
||||
|
||||
mod transformation {
|
||||
use std::sync::{
|
||||
Arc,
|
||||
atomic::{AtomicUsize, Ordering},
|
||||
};
|
||||
|
||||
use litellm_auth::{
|
||||
ResolvedCredential, SecretValue, TokenFuture, TokenProvider, TokenProviderHandle,
|
||||
};
|
||||
use rstest::rstest;
|
||||
use serde_json::json;
|
||||
|
||||
use super::*;
|
||||
use crate::ocr::{
|
||||
test_support::{MockResponse, header, mock_server, perform_ocr},
|
||||
types::LiteLLMOcrRequest,
|
||||
wire::decode_request,
|
||||
};
|
||||
|
||||
#[derive(Debug)]
|
||||
struct CountingToken {
|
||||
token: fn(usize) -> String,
|
||||
calls: AtomicUsize,
|
||||
}
|
||||
|
||||
impl CountingToken {
|
||||
fn new(token: fn(usize) -> String) -> Arc<Self> {
|
||||
Arc::new(Self {
|
||||
token,
|
||||
calls: AtomicUsize::new(0),
|
||||
})
|
||||
}
|
||||
|
||||
fn calls(&self) -> usize {
|
||||
self.calls.load(Ordering::SeqCst)
|
||||
}
|
||||
}
|
||||
|
||||
impl TokenProvider for CountingToken {
|
||||
fn acquire(&self) -> TokenFuture<'_> {
|
||||
let call = self.calls.fetch_add(1, Ordering::SeqCst) + 1;
|
||||
let token = SecretValue::new((self.token)(call));
|
||||
Box::pin(async move {
|
||||
Ok(ResolvedCredential::AccessToken {
|
||||
token,
|
||||
expires_on: None,
|
||||
})
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
fn numbered_token(call: usize) -> String {
|
||||
format!("callback-{call}")
|
||||
}
|
||||
|
||||
fn azure_request(
|
||||
provider: &Arc<CountingToken>,
|
||||
api_base: Option<&str>,
|
||||
api_key: Option<&str>,
|
||||
extra_headers: Value,
|
||||
optional_params: Value,
|
||||
) -> LiteLLMOcrRequest {
|
||||
let wire = serde_json::from_value(json!({
|
||||
"model": "azure_ai/mistral-ocr-latest",
|
||||
"document": {"type":"document_url","document_url":"data:application/pdf;base64,YWJj"},
|
||||
"api_key": api_key,
|
||||
"api_base": api_base,
|
||||
"custom_llm_provider": null,
|
||||
"extra_headers": extra_headers,
|
||||
"optional_params": optional_params,
|
||||
"timeout_seconds": 2.0
|
||||
}))
|
||||
.unwrap();
|
||||
LiteLLMOcrRequest {
|
||||
azure_ad_token_provider: Some(TokenProviderHandle::new(provider.clone())),
|
||||
..decode_request(wire).unwrap()
|
||||
}
|
||||
}
|
||||
|
||||
fn ocr_page() -> MockResponse {
|
||||
MockResponse::json(json!({"pages":[{"index":0,"markdown":"hello"}]}))
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn token_provider_result_is_the_bearer_and_is_acquired_for_each_request() {
|
||||
let provider = CountingToken::new(numbered_token);
|
||||
let (base, seen, server) = mock_server(vec![ocr_page(), ocr_page()]).await;
|
||||
|
||||
for _ in 0..2 {
|
||||
perform_ocr(azure_request(
|
||||
&provider,
|
||||
Some(&base),
|
||||
None,
|
||||
Value::Null,
|
||||
json!({}),
|
||||
))
|
||||
.await
|
||||
.unwrap();
|
||||
}
|
||||
server.await.unwrap();
|
||||
|
||||
assert_eq!(provider.calls(), 2);
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(
|
||||
requests
|
||||
.iter()
|
||||
.map(|request| header(request, "authorization"))
|
||||
.collect::<Vec<_>>(),
|
||||
[Some("Bearer callback-1"), Some("Bearer callback-2")]
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::api_key_skips_provider(Some("resource-key"), Value::Null, json!({}), "Bearer resource-key", 0)]
|
||||
#[case::provider_beats_static_token(
|
||||
None,
|
||||
Value::Null,
|
||||
json!({"azure_ad_token":"static-token"}),
|
||||
"Bearer callback-1",
|
||||
1
|
||||
)]
|
||||
#[case::header_wins_on_the_wire_but_provider_still_runs(
|
||||
None,
|
||||
json!({"Authorization":"Bearer override"}),
|
||||
json!({}),
|
||||
"Bearer override",
|
||||
1
|
||||
)]
|
||||
#[tokio::test]
|
||||
async fn credential_precedence(
|
||||
#[case] api_key: Option<&str>,
|
||||
#[case] extra_headers: Value,
|
||||
#[case] optional_params: Value,
|
||||
#[case] expected_authorization: &str,
|
||||
#[case] expected_calls: usize,
|
||||
) {
|
||||
let provider = CountingToken::new(numbered_token);
|
||||
let (base, seen, server) = mock_server(vec![ocr_page()]).await;
|
||||
|
||||
perform_ocr(azure_request(
|
||||
&provider,
|
||||
Some(&base),
|
||||
api_key,
|
||||
extra_headers,
|
||||
optional_params,
|
||||
))
|
||||
.await
|
||||
.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
assert_eq!(provider.calls(), expected_calls);
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 1);
|
||||
assert_eq!(
|
||||
header(&requests[0], "authorization"),
|
||||
Some(expected_authorization)
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::missing_api_base(
|
||||
false,
|
||||
json!({}),
|
||||
numbered_token,
|
||||
|error: &Error| matches!(error, Error::Auth(litellm_auth::Error::MissingApiBase {
|
||||
provider: "Azure AI",
|
||||
environment_variable: "AZURE_AI_API_BASE",
|
||||
})),
|
||||
0
|
||||
)]
|
||||
#[case::unsupported_oidc_reference(
|
||||
true,
|
||||
json!({"azure_ad_token":"oidc/assertion","client_id":"client","tenant_id":"tenant"}),
|
||||
numbered_token,
|
||||
|error: &Error| matches!(error, Error::Auth(litellm_auth::Error::UnsupportedOidcReference)),
|
||||
0
|
||||
)]
|
||||
#[case::empty_provider_token_ignores_static_token(
|
||||
true,
|
||||
json!({"azure_ad_token":"static-token"}),
|
||||
|_| String::new(),
|
||||
|error: &Error| matches!(error, Error::MissingAzureAiCredentials),
|
||||
1
|
||||
)]
|
||||
#[tokio::test]
|
||||
async fn credential_failures_send_no_provider_request(
|
||||
#[case] with_api_base: bool,
|
||||
#[case] optional_params: Value,
|
||||
#[case] token: fn(usize) -> String,
|
||||
#[case] expected: fn(&Error) -> bool,
|
||||
#[case] expected_calls: usize,
|
||||
) {
|
||||
let provider = CountingToken::new(token);
|
||||
let (base, seen, server) = mock_server(vec![ocr_page()]).await;
|
||||
|
||||
let error = perform_ocr(azure_request(
|
||||
&provider,
|
||||
with_api_base.then_some(base.as_str()),
|
||||
None,
|
||||
Value::Null,
|
||||
optional_params,
|
||||
))
|
||||
.await
|
||||
.unwrap_err();
|
||||
server.abort();
|
||||
|
||||
assert!(expected(&error), "unexpected error: {error:?}");
|
||||
assert_eq!(provider.calls(), expected_calls);
|
||||
assert!(seen.lock().unwrap().is_empty());
|
||||
}
|
||||
}
|
||||
|
|
@ -1,712 +0,0 @@
|
|||
use litellm_host::event::{CallEvent, MachineEvent};
|
||||
use litellm_llms::base_llm::ocr::{error::Error, settings::OcrSettings};
|
||||
use rstest::rstest;
|
||||
use serde_json::{Value, json};
|
||||
|
||||
use super::{
|
||||
test_support::{
|
||||
MockResponse, mock_server, ocr_client, perform_ocr, perform_ocr_with, wire_request,
|
||||
},
|
||||
wire::{OcrWireRequest, decode_request},
|
||||
};
|
||||
use crate::ocr::route::LocalOcrHost;
|
||||
|
||||
fn query_value(url: &str, key: &str) -> Option<String> {
|
||||
url::Url::parse(url)
|
||||
.unwrap()
|
||||
.query_pairs()
|
||||
.find_map(|(name, value)| (name == key).then(|| value.into_owned()))
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn facade_maps_pages_features_and_url_document() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"status":"succeeded",
|
||||
"analyzeResult":{"pages":[]}
|
||||
}))])
|
||||
.await;
|
||||
let mut request = wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({"pages":[2,0,0,1],"features":["keyValuePairs","languages"]}),
|
||||
);
|
||||
request.document =
|
||||
serde_json::from_value::<litellm_llms::base_llm::ocr::transformation::OcrDocument>(json!({
|
||||
"type":"document_url",
|
||||
"document_url":"https://example.com/document.pdf"
|
||||
}))
|
||||
.unwrap()
|
||||
.into();
|
||||
|
||||
perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
let request = &seen.lock().unwrap()[0];
|
||||
let target = request.split_whitespace().nth(1).unwrap();
|
||||
let url = format!("{base}{target}");
|
||||
assert_eq!(query_value(&url, "pages").as_deref(), Some("1,2,3"));
|
||||
assert_eq!(
|
||||
query_value(&url, "features").as_deref(),
|
||||
Some("keyValuePairs,languages")
|
||||
);
|
||||
let body: Value = serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap();
|
||||
assert_eq!(
|
||||
body,
|
||||
json!({"urlSource":"https://example.com/document.pdf"})
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case(json!({"pages":[true]}), Error::Pages("expected only integers or only strings".into()))]
|
||||
#[case(json!({"pages":[1,"2"]}), Error::Pages("expected only integers or only strings".into()))]
|
||||
#[case(json!({"pages":[-1]}), Error::Pages("negative page index".into()))]
|
||||
#[case(json!({"pages":"1&&features=bad"}), Error::Pages("invalid native page range".into()))]
|
||||
#[case(json!({"features":"languages&pages=1"}), Error::Features)]
|
||||
#[case(json!({"req_format":"azure"}), Error::RequestFormat)]
|
||||
#[tokio::test]
|
||||
async fn rejects_invalid_pages_features_and_format(
|
||||
#[case] options: Value,
|
||||
#[case] expected: Error,
|
||||
) {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({}))]).await;
|
||||
let result = decode_request(OcrWireRequest {
|
||||
model: "azure_ai/doc-intelligence/prebuilt-read".into(),
|
||||
document: json!({"type":"document_url","document_url":"https://example.com/a.pdf"}),
|
||||
api_key: Some(litellm_auth::SecretValue::new("key")),
|
||||
api_base: Some(base),
|
||||
custom_llm_provider: None,
|
||||
extra_headers: None,
|
||||
optional_params: options.as_object().unwrap().clone(),
|
||||
input_sources: Default::default(),
|
||||
timeout_seconds: Some(2.0),
|
||||
});
|
||||
let result = match result {
|
||||
Ok(request) => perform_ocr(request).await,
|
||||
Err(error) => Err(error),
|
||||
};
|
||||
server.abort();
|
||||
let _ = server.await;
|
||||
assert!(
|
||||
seen.lock().unwrap().is_empty(),
|
||||
"sent invalid options: {options}"
|
||||
);
|
||||
let error = result.unwrap_err();
|
||||
assert_eq!(
|
||||
std::mem::discriminant(&error),
|
||||
std::mem::discriminant(&expected)
|
||||
);
|
||||
assert_eq!(error.http_status_code(), Some(400));
|
||||
assert_eq!(error.to_string(), expected.to_string());
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case(json!({}))]
|
||||
#[case(json!({"req_format":"litellm"}))]
|
||||
#[tokio::test]
|
||||
async fn missing_native_fields_keep_page_text_without_retaining_raw_response(
|
||||
#[case] options: Value,
|
||||
) {
|
||||
let operation = json!({
|
||||
"status":"succeeded",
|
||||
"analyzeResult":{"pages":[{"pageNumber":1,"lines":[{"content":"hello"}]}]}
|
||||
});
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(operation)]).await;
|
||||
let response = perform_ocr(wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
options,
|
||||
))
|
||||
.await
|
||||
.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
assert_eq!(response.pages.len(), 1);
|
||||
assert_eq!(response.pages[0].index, 0);
|
||||
assert_eq!(response.pages[0].markdown, "hello");
|
||||
assert_eq!(response.provider_native_response, None);
|
||||
let serialized = response.into_json();
|
||||
assert_eq!(serialized.get("content"), Some(&Value::Null));
|
||||
assert_eq!(serialized.get("tables"), Some(&Value::Null));
|
||||
assert_eq!(serialized.get("keyValuePairs"), Some(&Value::Null));
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 1);
|
||||
let target = requests[0].split_whitespace().nth(1).unwrap();
|
||||
let url = format!("{base}{target}");
|
||||
for field in ["pages", "features", "req_format"] {
|
||||
assert_eq!(query_value(&url, field), None);
|
||||
}
|
||||
let body: Value = serde_json::from_str(requests[0].split_once("\r\n\r\n").unwrap().1).unwrap();
|
||||
assert_eq!(body, json!({"base64Source":"YWJj"}));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn inline_document_decodes_to_base64_source() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"status":"succeeded"
|
||||
}))])
|
||||
.await;
|
||||
let request = wire_request("azure_ai/doc-intelligence/prebuilt-read", &base, json!({}));
|
||||
|
||||
perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
let request = &seen.lock().unwrap()[0];
|
||||
let body: Value = serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap();
|
||||
assert_eq!(body, json!({"base64Source":"YWJj"}));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn immediate_response_normalizes_pages_and_preserves_native() {
|
||||
let operation = json!({
|
||||
"status":"succeeded",
|
||||
"operationExtension":42,
|
||||
"analyzeResult":{
|
||||
"content":"A\n\nB",
|
||||
"tables":[{"cells":[]}],
|
||||
"keyValuePairs":[{"key":{"content":"A"}}],
|
||||
"pages":[{
|
||||
"pageNumber":"2",
|
||||
"width":"8.5",
|
||||
"height":11,
|
||||
"unit":"inch",
|
||||
"lines":[{"content":"A"},{"content":null},{"content":"B"}]
|
||||
}]
|
||||
}
|
||||
});
|
||||
let (base, _, server) = mock_server(vec![MockResponse::json(operation.clone())]).await;
|
||||
let result = perform_ocr(wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({"req_format":"native"}),
|
||||
))
|
||||
.await
|
||||
.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
assert_eq!(result.pages[0].index, 1);
|
||||
assert_eq!(result.pages[0].markdown, "A\n\nB");
|
||||
assert_eq!(
|
||||
serde_json::to_value(&result.pages[0].dimensions).unwrap(),
|
||||
json!({"width":816,"height":1056,"dpi":96})
|
||||
);
|
||||
assert_eq!(result.usage_info.as_ref().unwrap().pages_processed, Some(1));
|
||||
let serialized = result.clone().into_json();
|
||||
assert_eq!(serialized["content"], "A\n\nB");
|
||||
assert_eq!(serialized["tables"], json!([{"cells":[]}]));
|
||||
assert_eq!(
|
||||
serialized["keyValuePairs"],
|
||||
json!([{"key":{"content":"A"}}])
|
||||
);
|
||||
assert!(serialized.get("key_value_pairs").is_none());
|
||||
assert_eq!(
|
||||
result.provider_native_response.map(Value::Object),
|
||||
Some(operation)
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn client_settings_choose_the_api_version_and_the_inch_to_pixel_dpi() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"status":"succeeded",
|
||||
"analyzeResult":{"pages":[{"pageNumber":1,"width":8.5,"height":11,"unit":"inch"}]}
|
||||
}))])
|
||||
.await;
|
||||
let client = ocr_client().with_settings(OcrSettings {
|
||||
document_intelligence_api_version: "2099-01-01".into(),
|
||||
document_intelligence_dpi: 72,
|
||||
..OcrSettings::default()
|
||||
});
|
||||
|
||||
let result = crate::ocr::client::perform(
|
||||
&client,
|
||||
wire_request("azure_ai/doc-intelligence/prebuilt-read", &base, json!({})),
|
||||
)
|
||||
.await
|
||||
.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
let target = seen.lock().unwrap()[0]
|
||||
.split_whitespace()
|
||||
.nth(1)
|
||||
.unwrap()
|
||||
.to_string();
|
||||
assert_eq!(
|
||||
query_value(&format!("{base}{target}"), "api-version").as_deref(),
|
||||
Some("2099-01-01")
|
||||
);
|
||||
assert_eq!(
|
||||
serde_json::to_value(&result.pages[0].dimensions).unwrap(),
|
||||
json!({"width":612,"height":792,"dpi":72})
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn accepted_response_polls_to_success_with_only_credentials() {
|
||||
let operation = json!({"status":"succeeded","analyzeResult":{"pages":[]}});
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse {
|
||||
status: 202,
|
||||
headers: vec![("Operation-Location", "{base}/operation".into())],
|
||||
body: json!({}),
|
||||
},
|
||||
MockResponse {
|
||||
status: 200,
|
||||
headers: vec![("Retry-After", "0".into())],
|
||||
body: json!({"status":"running"}),
|
||||
},
|
||||
MockResponse::json(operation.clone()),
|
||||
])
|
||||
.await;
|
||||
let mut request = wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({"req_format":"native"}),
|
||||
);
|
||||
request
|
||||
.transport
|
||||
.extra_headers
|
||||
.push(("X-Trace".into(), "initial-only".into()));
|
||||
|
||||
let result = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(
|
||||
result.provider_native_response.map(Value::Object),
|
||||
Some(operation)
|
||||
);
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 3);
|
||||
assert!(requests[0].to_ascii_lowercase().contains("x-trace:"));
|
||||
for poll in &requests[1..] {
|
||||
assert!(!poll.to_ascii_lowercase().contains("x-trace:"));
|
||||
assert!(
|
||||
poll.to_ascii_lowercase()
|
||||
.contains("ocp-apim-subscription-key: test-key")
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn accepted_response_emits_response_received_before_polling() {
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse {
|
||||
status: 202,
|
||||
headers: vec![("Operation-Location", "{base}/operation".into())],
|
||||
body: json!({"submitted": true}),
|
||||
},
|
||||
MockResponse::json(json!({"status":"succeeded"})),
|
||||
])
|
||||
.await;
|
||||
let request_count = seen.clone();
|
||||
let host = LocalOcrHost::new(wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({}),
|
||||
))
|
||||
.with_observer(move |event| {
|
||||
let CallEvent::Machine(MachineEvent::ResponseReceived { raw }) = event else {
|
||||
return;
|
||||
};
|
||||
match request_count.lock().unwrap().len() {
|
||||
1 => assert_eq!(raw.body, r#"{"submitted":true}"#),
|
||||
2 => assert!(raw.body.contains("succeeded")),
|
||||
count => panic!("unexpected callback after {count} requests"),
|
||||
}
|
||||
});
|
||||
|
||||
perform_ocr_with(host).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(seen.lock().unwrap().len(), 2);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn polling_forwards_bearer_credentials() {
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse {
|
||||
status: 202,
|
||||
headers: vec![("Operation-Location", "{base}/operation".into())],
|
||||
body: json!({}),
|
||||
},
|
||||
MockResponse::json(json!({"status":"succeeded"})),
|
||||
])
|
||||
.await;
|
||||
let mut request = wire_request("azure_ai/doc-intelligence/prebuilt-read", &base, json!({}));
|
||||
request.credentials.api_key = None;
|
||||
request.transport.extra_headers = vec![("Authorization".into(), "Bearer token".into())];
|
||||
|
||||
perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
let requests = seen.lock().unwrap();
|
||||
assert!(
|
||||
requests[1]
|
||||
.to_ascii_lowercase()
|
||||
.contains("authorization: bearer token")
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn polling_does_not_follow_redirects() {
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse {
|
||||
status: 202,
|
||||
headers: vec![("Operation-Location", "{base}/operation".into())],
|
||||
body: json!({}),
|
||||
},
|
||||
MockResponse {
|
||||
status: 302,
|
||||
headers: vec![("Location", "{base}/redirected".into())],
|
||||
body: json!({}),
|
||||
},
|
||||
MockResponse::json(json!({"status":"succeeded"})),
|
||||
])
|
||||
.await;
|
||||
|
||||
let error = perform_ocr(wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({}),
|
||||
))
|
||||
.await
|
||||
.unwrap_err();
|
||||
|
||||
assert!(error.to_string().contains("status 302"), "{error}");
|
||||
assert_eq!(seen.lock().unwrap().len(), 2);
|
||||
server.abort();
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn polling_rejects_terminal_failure() {
|
||||
let (base, _, server) = mock_server(vec![
|
||||
MockResponse {
|
||||
status: 202,
|
||||
headers: vec![("Operation-Location", "{base}/operation".into())],
|
||||
body: json!({}),
|
||||
},
|
||||
MockResponse::json(json!({"status":"failed"})),
|
||||
])
|
||||
.await;
|
||||
|
||||
let error = perform_ocr(wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({}),
|
||||
))
|
||||
.await
|
||||
.unwrap_err();
|
||||
server.await.unwrap();
|
||||
assert!(error.to_string().contains("status failed"));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn malformed_provider_pages_report_response_paths() {
|
||||
for (analysis, path) in [
|
||||
(json!({"pages":null}), "pages"),
|
||||
(json!({"pages":[null]}), "pages[0]"),
|
||||
(json!({"pages":[{"lines":null}]}), "lines"),
|
||||
(json!({"pages":[{"width":"bad"}]}), "width"),
|
||||
] {
|
||||
let (base, _, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"status":"succeeded",
|
||||
"analyzeResult":analysis
|
||||
}))])
|
||||
.await;
|
||||
let error = perform_ocr(wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({}),
|
||||
))
|
||||
.await
|
||||
.unwrap_err();
|
||||
server.await.unwrap();
|
||||
assert!(error.to_string().contains(path), "{error}");
|
||||
}
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn rejects_missing_invalid_and_cross_origin_operation_locations() {
|
||||
for headers in [
|
||||
Vec::new(),
|
||||
vec![("Operation-Location", "/relative".into())],
|
||||
vec![("Operation-Location", "http://example.com/operation".into())],
|
||||
vec![(
|
||||
"Operation-Location",
|
||||
"http://user:password@127.0.0.1/operation".into(),
|
||||
)],
|
||||
] {
|
||||
let (base, _, server) = mock_server(vec![MockResponse {
|
||||
status: 202,
|
||||
headers,
|
||||
body: json!({}),
|
||||
}])
|
||||
.await;
|
||||
let error = perform_ocr(wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({}),
|
||||
))
|
||||
.await
|
||||
.unwrap_err();
|
||||
server.await.unwrap();
|
||||
assert!(error.to_string().contains("operation-location"));
|
||||
}
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn polling_deadline_bounds_retry_delay() {
|
||||
let (base, _, server) = mock_server(vec![
|
||||
MockResponse {
|
||||
status: 202,
|
||||
headers: vec![("Operation-Location", "{base}/operation".into())],
|
||||
body: json!({}),
|
||||
},
|
||||
MockResponse {
|
||||
status: 200,
|
||||
headers: vec![("Retry-After", "9999".into())],
|
||||
body: json!({"status":"notStarted"}),
|
||||
},
|
||||
])
|
||||
.await;
|
||||
let request = wire_request("azure_ai/doc-intelligence/prebuilt-read", &base, json!({}));
|
||||
let client = ocr_client().with_settings(OcrSettings {
|
||||
poll_timeout: std::time::Duration::from_millis(100),
|
||||
..OcrSettings::default()
|
||||
});
|
||||
|
||||
let error = tokio::time::timeout(
|
||||
std::time::Duration::from_secs(1),
|
||||
crate::ocr::client::perform(&client, request),
|
||||
)
|
||||
.await
|
||||
.unwrap()
|
||||
.unwrap_err();
|
||||
server.await.unwrap();
|
||||
assert!(error.to_string().contains("timed out"));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn model_id_is_encoded_and_dot_segments_are_rejected() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"status":"succeeded"
|
||||
}))])
|
||||
.await;
|
||||
perform_ocr(wire_request(
|
||||
"azure_ai/doc-intelligence/a ?#é",
|
||||
&base,
|
||||
json!({}),
|
||||
))
|
||||
.await
|
||||
.unwrap();
|
||||
server.await.unwrap();
|
||||
assert!(seen.lock().unwrap()[0].contains("a%20%3F%23%C3%A9:analyze"));
|
||||
|
||||
for model in [
|
||||
"azure_ai/doc-intelligence/.",
|
||||
"azure_ai/doc-intelligence/..",
|
||||
] {
|
||||
let error = perform_ocr(wire_request(model, "http://127.0.0.1:1", json!({})))
|
||||
.await
|
||||
.unwrap_err();
|
||||
assert!(error.to_string().contains("dot segment"));
|
||||
}
|
||||
}
|
||||
|
||||
mod transformation {
|
||||
use std::sync::{Arc, Mutex};
|
||||
|
||||
use litellm_host::event::{CallEvent, MachineEvent};
|
||||
use litellm_llms::base_llm::ocr::transformation::OcrDocument;
|
||||
use serde_json::{Value, json};
|
||||
|
||||
use super::*;
|
||||
use crate::ocr::{
|
||||
route::LocalOcrHost,
|
||||
test_support::{MockResponse, mock_server, perform_ocr, perform_ocr_with, wire_request},
|
||||
};
|
||||
|
||||
#[tokio::test]
|
||||
async fn facade_maps_pages_features_and_url_document() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"status":"succeeded",
|
||||
"analyzeResult":{"pages":[]}
|
||||
}))])
|
||||
.await;
|
||||
let mut request = wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({"pages":[2,0,0,1],"features":["keyValuePairs","languages"], "future_option": {"nested":null}, "extra_body":{"provider_option":false}}),
|
||||
);
|
||||
request.document = serde_json::from_value::<OcrDocument>(json!({
|
||||
"type":"document_url",
|
||||
"document_url":"https://example.com/document.pdf"
|
||||
}))
|
||||
.unwrap()
|
||||
.into();
|
||||
|
||||
perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
let request = &seen.lock().unwrap()[0];
|
||||
let target = request.split_whitespace().nth(1).unwrap();
|
||||
let url = format!("{base}{target}");
|
||||
assert_eq!(query_value(&url, "pages").as_deref(), Some("1,2,3"));
|
||||
assert_eq!(
|
||||
query_value(&url, "features").as_deref(),
|
||||
Some("keyValuePairs,languages")
|
||||
);
|
||||
let body: Value = serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap();
|
||||
assert_eq!(
|
||||
body,
|
||||
json!({"urlSource":"https://example.com/document.pdf", "future_option":{"nested":null}, "provider_option":false})
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn rejects_invalid_pages_features_and_format() {
|
||||
for options in [
|
||||
json!({"pages":[true]}),
|
||||
json!({"pages":[1,"2"]}),
|
||||
json!({"pages":[-1]}),
|
||||
json!({"pages":"1&&features=bad"}),
|
||||
json!({"features":"languages&pages=1"}),
|
||||
json!({"req_format":"azure"}),
|
||||
] {
|
||||
let request = wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
"http://127.0.0.1:1",
|
||||
options.clone(),
|
||||
);
|
||||
let rejected = perform_ocr(request).await.is_err();
|
||||
assert!(rejected, "accepted {options}");
|
||||
}
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn immediate_response_normalizes_pages_and_preserves_native() {
|
||||
let operation = json!({
|
||||
"status":"succeeded",
|
||||
"operationExtension":42,
|
||||
"analyzeResult":{
|
||||
"content":"A\n\nB",
|
||||
"tables":[{"cells":[]}],
|
||||
"keyValuePairs":[{"key":{"content":"A"}}],
|
||||
"pages":[{
|
||||
"pageNumber":"2",
|
||||
"width":"8.5",
|
||||
"height":11,
|
||||
"unit":"inch",
|
||||
"lines":[{"content":"A"},{"content":null},{"content":"B"}]
|
||||
}]
|
||||
}
|
||||
});
|
||||
let (base, _, server) = mock_server(vec![MockResponse::json(operation.clone())]).await;
|
||||
let result = perform_ocr(wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({"req_format":"native"}),
|
||||
))
|
||||
.await
|
||||
.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
assert_eq!(result.pages[0].index, 1);
|
||||
assert_eq!(result.pages[0].markdown, "A\n\nB");
|
||||
assert_eq!(
|
||||
serde_json::to_value(&result.pages[0].dimensions).unwrap(),
|
||||
json!({"width":816,"height":1056,"dpi":96})
|
||||
);
|
||||
assert_eq!(result.usage_info.as_ref().unwrap().pages_processed, Some(1));
|
||||
let serialized = result.clone().into_json();
|
||||
assert_eq!(serialized["content"], "A\n\nB");
|
||||
assert_eq!(serialized["tables"], json!([{"cells":[]}]));
|
||||
assert_eq!(
|
||||
serialized["keyValuePairs"],
|
||||
json!([{"key":{"content":"A"}}])
|
||||
);
|
||||
assert!(serialized.get("key_value_pairs").is_none());
|
||||
assert_eq!(
|
||||
result.provider_native_response.as_ref(),
|
||||
operation.as_object()
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn accepted_response_polls_to_success_with_only_credentials() {
|
||||
let operation = json!({"status":"succeeded","analyzeResult":{"pages":[]}});
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse {
|
||||
status: 202,
|
||||
headers: vec![("Operation-Location", "{base}/operation".into())],
|
||||
body: json!({}),
|
||||
},
|
||||
MockResponse {
|
||||
status: 200,
|
||||
headers: vec![("Retry-After", "0".into())],
|
||||
body: json!({"status":"running"}),
|
||||
},
|
||||
MockResponse::json(operation.clone()),
|
||||
])
|
||||
.await;
|
||||
let mut request = wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({"req_format":"native"}),
|
||||
);
|
||||
request
|
||||
.transport
|
||||
.extra_headers
|
||||
.push(("X-Trace".into(), "initial-only".into()));
|
||||
|
||||
let result = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(
|
||||
result.provider_native_response.as_ref(),
|
||||
operation.as_object()
|
||||
);
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 3);
|
||||
assert!(requests[0].to_ascii_lowercase().contains("x-trace:"));
|
||||
for poll in &requests[1..] {
|
||||
assert!(!poll.to_ascii_lowercase().contains("x-trace:"));
|
||||
assert!(
|
||||
poll.to_ascii_lowercase()
|
||||
.contains("ocp-apim-subscription-key: test-key")
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn accepted_response_emits_response_received_for_submission_and_completed_poll() {
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse {
|
||||
status: 202,
|
||||
headers: vec![("Operation-Location", "{base}/operation".into())],
|
||||
body: json!({"submitted": true}),
|
||||
},
|
||||
MockResponse::json(json!({"status":"succeeded"})),
|
||||
])
|
||||
.await;
|
||||
let responses_received = Arc::new(Mutex::new(Vec::new()));
|
||||
let request_count = seen.clone();
|
||||
let observed = responses_received.clone();
|
||||
let host = LocalOcrHost::new(wire_request(
|
||||
"azure_ai/doc-intelligence/prebuilt-read",
|
||||
&base,
|
||||
json!({}),
|
||||
))
|
||||
.with_observer(move |event| {
|
||||
if let CallEvent::Machine(MachineEvent::ResponseReceived { raw }) = event {
|
||||
observed
|
||||
.lock()
|
||||
.unwrap()
|
||||
.push((request_count.lock().unwrap().len(), raw.body.clone()));
|
||||
}
|
||||
});
|
||||
|
||||
perform_ocr_with(host).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(seen.lock().unwrap().len(), 2);
|
||||
assert_eq!(
|
||||
*responses_received.lock().unwrap(),
|
||||
[
|
||||
(1, r#"{"submitted":true}"#.to_string()),
|
||||
(2, r#"{"status":"succeeded"}"#.to_string()),
|
||||
]
|
||||
);
|
||||
}
|
||||
}
|
||||
|
|
@ -1,136 +0,0 @@
|
|||
mod transformation {
|
||||
use litellm_llms::{
|
||||
base_llm::ocr::{
|
||||
error::Error,
|
||||
transformation::{BaseOcrConfig, OcrDocument, OcrResponseFormat},
|
||||
},
|
||||
cohere::ocr::transformation::*,
|
||||
};
|
||||
use rstest::rstest;
|
||||
use serde_json::{Value, json};
|
||||
|
||||
#[tokio::test]
|
||||
async fn composed_body_preserves_native_document_fields_and_untyped_overrides() {
|
||||
let request = crate::ocr::test_support::wire_request(
|
||||
"cohere/parse",
|
||||
"https://example.com",
|
||||
json!({
|
||||
"output_format":"markdown", "timeout":30,
|
||||
"extra_body":{
|
||||
"output_format": {"future":true},
|
||||
"document":{"type":"image_url","image_url":"https://example.com/a.png",
|
||||
"provider_options":{"nested":[false,0,null]}}
|
||||
}
|
||||
}),
|
||||
);
|
||||
let request = request.with_document(
|
||||
serde_json::from_value(json!({
|
||||
"type":"image_url","image_url":"https://example.com/original.png"
|
||||
}))
|
||||
.unwrap(),
|
||||
);
|
||||
let request = crate::ocr::prepare::prepare_request_for_test(request);
|
||||
let http = CohereParseConfig
|
||||
.prepare_request(
|
||||
&request,
|
||||
&crate::ocr::test_support::ocr_client(),
|
||||
&crate::ocr::test_support::NoHooks,
|
||||
)
|
||||
.await
|
||||
.unwrap();
|
||||
let body: Value = serde_json::from_slice(http.body()).unwrap();
|
||||
assert_eq!(
|
||||
body,
|
||||
json!({
|
||||
"model":"parse", "output_format":{"future":true},
|
||||
"document":{"type":"image_url","image_url":"https://example.com/a.png",
|
||||
"provider_options":{"nested":[false,0,null]}}
|
||||
})
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn explicit_null_options_use_defaults_before_http() {
|
||||
let request = crate::ocr::test_support::wire_request(
|
||||
"cohere/parse",
|
||||
"https://example.com",
|
||||
json!({"output_format":null,"req_format":null}),
|
||||
);
|
||||
let request = request.with_document(
|
||||
serde_json::from_value(
|
||||
json!({"type":"image_url","image_url":"https://example.com/a.png"}),
|
||||
)
|
||||
.unwrap(),
|
||||
);
|
||||
assert_eq!(
|
||||
request.response_format().unwrap(),
|
||||
OcrResponseFormat::Litellm
|
||||
);
|
||||
let request = crate::ocr::prepare::prepare_request_for_test(request);
|
||||
let http = CohereParseConfig
|
||||
.prepare_request(
|
||||
&request,
|
||||
&crate::ocr::test_support::ocr_client(),
|
||||
&crate::ocr::test_support::NoHooks,
|
||||
)
|
||||
.await
|
||||
.unwrap();
|
||||
let body: Value = serde_json::from_slice(http.body()).unwrap();
|
||||
assert_eq!(body["output_format"], "markdown");
|
||||
assert!(body.get("req_format").is_none());
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::cohere("cohere/parse-v5.0", "POST /v2/parse ")]
|
||||
#[case::azure_ai("azure_ai/Cohere-parse-v5.0", "POST /providers/cohere/v2/parse ")]
|
||||
#[tokio::test]
|
||||
async fn route_sends_image_to_its_parse_endpoint_with_the_bearer_key(
|
||||
#[case] model: &str,
|
||||
#[case] request_line: &str,
|
||||
) {
|
||||
use crate::ocr::test_support::{MockResponse, header, mock_server, perform_ocr};
|
||||
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({"pages":[]}))]).await;
|
||||
let request = crate::ocr::test_support::wire_request(model, &base, json!({}))
|
||||
.with_document(
|
||||
serde_json::from_value::<OcrDocument>(
|
||||
json!({"type":"image_url","image_url":"data:image/png;base64,YWJj"}),
|
||||
)
|
||||
.unwrap()
|
||||
.into(),
|
||||
);
|
||||
|
||||
perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 1);
|
||||
assert!(requests[0].starts_with(request_line), "{}", requests[0]);
|
||||
assert_eq!(
|
||||
header(&requests[0], "authorization"),
|
||||
Some("Bearer test-key")
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[tokio::test]
|
||||
async fn route_rejects_non_image_document_without_a_request(
|
||||
#[values("cohere/parse-v5.0", "azure_ai/Cohere-parse-v5.0")] model: &str,
|
||||
) {
|
||||
use crate::ocr::test_support::{MockResponse, mock_server, perform_ocr};
|
||||
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({"pages":[]}))]).await;
|
||||
|
||||
let error = perform_ocr(crate::ocr::test_support::wire_request(
|
||||
model,
|
||||
&base,
|
||||
json!({}),
|
||||
))
|
||||
.await
|
||||
.unwrap_err();
|
||||
server.abort();
|
||||
|
||||
assert!(matches!(error, Error::CohereImageOnly), "{error:?}");
|
||||
assert!(seen.lock().unwrap().is_empty());
|
||||
}
|
||||
}
|
||||
|
|
@ -1,133 +0,0 @@
|
|||
use litellm_llms::{
|
||||
base_llm::ocr::transformation::{BaseOcrConfig, OcrDocument},
|
||||
vertex_ai::ocr::deepseek_transformation::{
|
||||
DeepSeekOcrParams, DeepSeekOcrResponse, VertexAIDeepSeekOCRConfig,
|
||||
normalize_response as transform_ocr_response,
|
||||
},
|
||||
};
|
||||
use rstest::rstest;
|
||||
use serde_json::{Value, json};
|
||||
|
||||
fn document() -> OcrDocument {
|
||||
serde_json::from_value(json!({"type":"image_url","image_url":"gs://bucket/a.png"})).unwrap()
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case("stream", json!(true))]
|
||||
#[case("temperature", json!(0.1))]
|
||||
#[case("max_tokens", json!(1024))]
|
||||
#[case("top_p", json!(0.9))]
|
||||
#[case("n", json!(2))]
|
||||
#[case("stop", json!("done"))]
|
||||
#[case("stop", json!(["done", "stop"]))]
|
||||
fn request_mapping_matches_python(#[case] name: &str, #[case] value: Value) {
|
||||
let params: DeepSeekOcrParams =
|
||||
serde_json::from_value(json!({name: value.clone(), "ignored": true})).unwrap();
|
||||
let result = serde_json::to_value(
|
||||
VertexAIDeepSeekOCRConfig
|
||||
.transform_ocr_request("deepseek-ai/deepseek-ocr-maas", document(), ¶ms, &[])
|
||||
.unwrap(),
|
||||
)
|
||||
.unwrap();
|
||||
assert_eq!(result["model"], "deepseek-ai/deepseek-ocr-maas");
|
||||
assert_eq!(
|
||||
result["messages"][0]["content"][0],
|
||||
json!({"type":"image_url","image_url":"gs://bucket/a.png"})
|
||||
);
|
||||
assert_eq!(result[name], value);
|
||||
assert!(result.get("ignored").is_none());
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case(json!({"type":"image_url","image_url":"data:image/png;base64,AA=="}))]
|
||||
#[case(json!({"type":"document_url","document_url":"data:application/pdf;base64,AA=="}))]
|
||||
fn request_maps_both_document_types_to_image_content(#[case] document: Value) {
|
||||
let source = document
|
||||
.get("image_url")
|
||||
.or_else(|| document.get("document_url"))
|
||||
.unwrap()
|
||||
.clone();
|
||||
let request = VertexAIDeepSeekOCRConfig
|
||||
.transform_ocr_request(
|
||||
"deepseek-ai/deepseek-ocr-maas",
|
||||
serde_json::from_value(document).unwrap(),
|
||||
&DeepSeekOcrParams::default(),
|
||||
&[],
|
||||
)
|
||||
.unwrap();
|
||||
let result = serde_json::to_value(request).unwrap();
|
||||
assert_eq!(
|
||||
result["messages"][0]["content"][0],
|
||||
json!({"type":"image_url","image_url":source})
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case(json!("# hello"), "# hello")]
|
||||
#[case(json!("{broken"), "{broken")]
|
||||
#[case(json!(" {\"pages\":[]} "), " {\"pages\":[]} ")]
|
||||
#[case(json!({"pages":[]}), "")]
|
||||
#[case(json!("[]"), "[]")]
|
||||
#[case(json!("{\"pages\":[{\"markdown\":\"json text\"}]}"), "json text")]
|
||||
#[case(json!({"pages":[{"markdown":"object"}]}), "object")]
|
||||
fn response_codec_handles_text_json_and_objects(#[case] content: Value, #[case] expected: &str) {
|
||||
let structured = content
|
||||
.as_object()
|
||||
.is_some_and(|object| object.contains_key("pages"))
|
||||
|| content
|
||||
.as_str()
|
||||
.is_some_and(|text| text.contains("\"pages\""));
|
||||
let response: DeepSeekOcrResponse = serde_json::from_value(
|
||||
json!({"choices":[{"message":{"content":content}}],"usage":{"prompt_tokens":1}}),
|
||||
)
|
||||
.unwrap();
|
||||
let result = transform_ocr_response("model", response)
|
||||
.unwrap()
|
||||
.into_json();
|
||||
assert_eq!(result["pages"][0]["markdown"], expected);
|
||||
assert_eq!(result["pages"][0]["index"], 0);
|
||||
if structured {
|
||||
assert!(result["usage_info"].is_null());
|
||||
} else {
|
||||
assert_eq!(result["usage_info"]["prompt_tokens"], 1);
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn structured_result_maps_pages_usage_model_and_annotation() {
|
||||
let response: DeepSeekOcrResponse = serde_json::from_value(json!({
|
||||
"choices":[{"message":{"content":{
|
||||
"pages":[{"index":2,"markdown":"page","images":[{"id":"one"}],"dimensions":{"width":10}}],
|
||||
"model":"provider-model",
|
||||
"usage_info":{"pages_processed":1},
|
||||
"document_annotation":{"language":"en"},
|
||||
"future":"kept"
|
||||
}}}]
|
||||
}))
|
||||
.unwrap();
|
||||
let result = transform_ocr_response("requested", response)
|
||||
.unwrap()
|
||||
.into_json();
|
||||
assert_eq!(result["pages"][0]["index"], 2);
|
||||
assert_eq!(result["pages"][0]["images"][0]["id"], "one");
|
||||
assert_eq!(result["model"], "provider-model");
|
||||
assert_eq!(result["usage_info"]["pages_processed"], 1);
|
||||
assert_eq!(result["document_annotation"]["language"], "en");
|
||||
assert_eq!(result["future"], "kept");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn response_codec_rejects_missing_empty_and_malformed_content() {
|
||||
for value in [
|
||||
json!({"choices":[{"message":{"content":{}}}]}),
|
||||
json!({"choices":[]}),
|
||||
json!({"choices":[{"message":{"content":""}}]}),
|
||||
json!({"choices":[{"message":{"content":"{\"pages\":[{\"markdown\":42}]}"}}]}),
|
||||
json!({"choices":[{"message":{"content":{"pages":[{"markdown":42}]}}}]}),
|
||||
] {
|
||||
let result = serde_json::from_value::<DeepSeekOcrResponse>(value)
|
||||
.map_err(|_| ())
|
||||
.and_then(|response| transform_ocr_response("model", response).map_err(|_| ()));
|
||||
assert!(result.is_err());
|
||||
}
|
||||
}
|
||||
|
|
@ -1,7 +1,11 @@
|
|||
use std::{sync::Arc, time::Duration};
|
||||
|
||||
use futures_util::future::BoxFuture;
|
||||
use litellm_http::request::{has_bearer_auth, has_header};
|
||||
use litellm_core::messages::{
|
||||
Error, messages,
|
||||
route::{LocalMessagesHost, MessagesCall, messages_machine},
|
||||
types::{MessagesRequest, MessagesShaping},
|
||||
};
|
||||
use litellm_secrets::{SecretValue, source::SecretSource};
|
||||
use serde_json::{Map, Value, json};
|
||||
use tokio::{
|
||||
|
|
@ -9,14 +13,6 @@ use tokio::{
|
|||
net::{TcpListener, TcpStream},
|
||||
};
|
||||
|
||||
use super::{
|
||||
Error,
|
||||
common_utils::{messages_provider_config, string_headers, truncate_error_body},
|
||||
messages,
|
||||
route::{LocalMessagesHost, MessagesCall, MessagesOutput, messages_machine},
|
||||
};
|
||||
use crate::messages::types::{MessagesRequest, MessagesShaping};
|
||||
|
||||
struct RecordingSecrets {
|
||||
values: Vec<(&'static str, String)>,
|
||||
fails: bool,
|
||||
|
|
@ -73,55 +69,6 @@ fn secrets_call() -> MessagesCall {
|
|||
}
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn route_reads_the_provider_credential_and_base_from_the_secret_source() {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
|
||||
let addr = listener.local_addr().expect("addr");
|
||||
let server = tokio::spawn(async move {
|
||||
let (mut socket, _) = listener.accept().await.expect("accepts request");
|
||||
let request = read_http_request(&mut socket).await;
|
||||
let response_body = r#"{"id":"msg_1","type":"message","role":"assistant","content":[],"model":"claude-sonnet-4-5","stop_reason":"end_turn","usage":{"input_tokens":1,"output_tokens":1}}"#;
|
||||
socket
|
||||
.write_all(write_response(response_body).as_bytes())
|
||||
.await
|
||||
.expect("writes response");
|
||||
request
|
||||
});
|
||||
let secrets = Arc::new(RecordingSecrets::new(
|
||||
vec![
|
||||
("ANTHROPIC_API_KEY", "sk-from-manager".to_string()),
|
||||
("ANTHROPIC_BASE_URL", format!("http://{addr}")),
|
||||
],
|
||||
false,
|
||||
));
|
||||
|
||||
let output = litellm_host::run::run(
|
||||
messages_machine(secrets.clone()),
|
||||
&LocalMessagesHost::new(secrets_call()),
|
||||
)
|
||||
.await
|
||||
.expect("messages request succeeds");
|
||||
|
||||
assert!(matches!(output, MessagesOutput::Message(_)));
|
||||
let request = server.await.expect("server task completes");
|
||||
assert!(
|
||||
request
|
||||
.to_ascii_lowercase()
|
||||
.contains("x-api-key: sk-from-manager"),
|
||||
"{request}"
|
||||
);
|
||||
let requested = secrets.requested.lock().unwrap().clone();
|
||||
assert_eq!(
|
||||
requested,
|
||||
messages_provider_config("anthropic")
|
||||
.unwrap()
|
||||
.secret_names()
|
||||
.iter()
|
||||
.map(ToString::to_string)
|
||||
.collect::<Vec<_>>()
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn route_surfaces_a_secret_manager_failure_before_the_call() {
|
||||
let Err(error) = litellm_host::run::run(
|
||||
|
|
@ -179,75 +126,6 @@ fn write_response(body: &str) -> String {
|
|||
)
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn provider_config_resolves_anthropic_and_azure_ai() {
|
||||
assert!(messages_provider_config("anthropic").is_some());
|
||||
assert!(messages_provider_config("azure_ai").is_some());
|
||||
assert!(messages_provider_config("openai").is_none());
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn truncate_error_body_caps_long_payloads() {
|
||||
let body = "x".repeat(400);
|
||||
let truncated = truncate_error_body(&body);
|
||||
assert!(truncated.ends_with("... (truncated)"));
|
||||
let prefix_chars = truncated
|
||||
.strip_suffix("... (truncated)")
|
||||
.expect("truncated marker present")
|
||||
.chars()
|
||||
.count();
|
||||
assert_eq!(prefix_chars, 256);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn string_headers_rejects_non_string_values() {
|
||||
let headers = json!({"x-count": 3}).as_object().unwrap().clone();
|
||||
let err = string_headers(Some(headers)).expect_err("non-string header rejected");
|
||||
assert_eq!(
|
||||
err,
|
||||
Error::Headers(litellm_http::request::HeaderError {
|
||||
context: "messages",
|
||||
name: "x-count".to_string(),
|
||||
actual: "number",
|
||||
})
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn has_header_is_case_insensitive() {
|
||||
let headers = vec![("X-Api-Key".to_string(), "secret".to_string())];
|
||||
assert!(has_header(&headers, "x-api-key"));
|
||||
assert!(!has_header(&headers, "authorization"));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn has_bearer_auth_requires_a_nonempty_bearer_token() {
|
||||
assert!(has_bearer_auth(&[(
|
||||
"Authorization".to_string(),
|
||||
"Bearer tok".to_string()
|
||||
)]));
|
||||
assert!(has_bearer_auth(&[(
|
||||
"authorization".to_string(),
|
||||
"bearer tok".to_string()
|
||||
)]));
|
||||
assert!(!has_bearer_auth(&[(
|
||||
"authorization".to_string(),
|
||||
"Bearer ".to_string()
|
||||
)]));
|
||||
assert!(!has_bearer_auth(&[(
|
||||
"authorization".to_string(),
|
||||
String::new()
|
||||
)]));
|
||||
assert!(!has_bearer_auth(&[(
|
||||
"authorization".to_string(),
|
||||
"Basic abc".to_string()
|
||||
)]));
|
||||
assert!(!has_bearer_auth(&[(
|
||||
"x-api-key".to_string(),
|
||||
"sk".to_string()
|
||||
)]));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn messages_round_trip_builds_azure_request_and_passes_response_through() {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.expect("binds");
|
||||
File diff suppressed because it is too large
Load diff
|
|
@ -1,152 +0,0 @@
|
|||
use litellm_host::event::WireRequest;
|
||||
use litellm_llms::base_llm::ocr::error::Error;
|
||||
use rstest::rstest;
|
||||
use serde_json::{Value, json};
|
||||
|
||||
use super::test_support::{
|
||||
MockResponse, SERVED_DOCUMENT, document_server, mock_server, perform_ocr_with, request_body,
|
||||
wire_request_with_document,
|
||||
};
|
||||
use crate::ocr::route::LocalOcrHost;
|
||||
|
||||
#[derive(Clone, Copy, Debug)]
|
||||
enum Route {
|
||||
Mistral,
|
||||
AzureAi,
|
||||
VertexMistral,
|
||||
AzureCohereParse,
|
||||
Cohere,
|
||||
}
|
||||
|
||||
impl Route {
|
||||
fn model(self) -> &'static str {
|
||||
match self {
|
||||
Self::Mistral => "mistral/model",
|
||||
Self::AzureAi => "azure_ai/model",
|
||||
Self::VertexMistral => "vertex_ai/mistral-ocr-maas",
|
||||
Self::AzureCohereParse => "azure_ai/cohere-parse",
|
||||
Self::Cohere => "cohere/model",
|
||||
}
|
||||
}
|
||||
|
||||
fn document_type(self) -> &'static str {
|
||||
match self {
|
||||
Self::Mistral | Self::AzureAi | Self::VertexMistral => "document_url",
|
||||
Self::AzureCohereParse | Self::Cohere => "image_url",
|
||||
}
|
||||
}
|
||||
|
||||
fn options(self) -> Value {
|
||||
match self {
|
||||
Self::Mistral | Self::AzureAi => json!({"pages": [0]}),
|
||||
Self::VertexMistral => json!({"pages": [0], "vertex_project": "project-1"}),
|
||||
Self::AzureCohereParse | Self::Cohere => json!({"output_format": "markdown"}),
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// What the host does to the wire request in `before_send`.
|
||||
#[derive(Clone, Copy, Debug)]
|
||||
enum Host {
|
||||
Detached,
|
||||
ReplacesDocument,
|
||||
}
|
||||
|
||||
const REPLACED_DOCUMENT: &str = "data:image/png;base64,cmVwbGFjZWQ=";
|
||||
|
||||
impl Host {
|
||||
fn before_send(self, wire: WireRequest) -> WireRequest {
|
||||
let Value::Object(fields) = wire.body else {
|
||||
return wire;
|
||||
};
|
||||
let body = fields
|
||||
.into_iter()
|
||||
.map(|(name, value)| match self {
|
||||
Self::Detached => (name, value),
|
||||
Self::ReplacesDocument if name == "document" => {
|
||||
let document_type = value["type"].clone();
|
||||
let key = document_type.as_str().unwrap_or_default().to_string();
|
||||
(name, json!({"type": document_type, key: REPLACED_DOCUMENT}))
|
||||
}
|
||||
Self::ReplacesDocument => (name, value),
|
||||
})
|
||||
.collect();
|
||||
WireRequest {
|
||||
body: Value::Object(body),
|
||||
..wire
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
struct Sent {
|
||||
result: Result<(), Error>,
|
||||
provider_body: Option<Value>,
|
||||
}
|
||||
|
||||
async fn send(route: Route, host: Host, document_base: &str) -> Sent {
|
||||
let (base, seen, provider) = mock_server(vec![MockResponse::json(json!({"pages": []}))]).await;
|
||||
let document_type = route.document_type();
|
||||
let document =
|
||||
json!({"type": document_type, document_type: format!("{document_base}/scan.png")});
|
||||
let request = wire_request_with_document(route.model(), &base, document, route.options());
|
||||
let local =
|
||||
LocalOcrHost::new(request).with_before_send(move |wire, _| Ok(host.before_send(wire)));
|
||||
let result = perform_ocr_with(local).await.map(|_| ());
|
||||
match result {
|
||||
Ok(()) => provider.await.unwrap(),
|
||||
Err(_) => provider.abort(),
|
||||
}
|
||||
let provider_body = seen
|
||||
.lock()
|
||||
.unwrap()
|
||||
.first()
|
||||
.map(|request| request_body(request));
|
||||
Sent {
|
||||
result,
|
||||
provider_body,
|
||||
}
|
||||
}
|
||||
|
||||
fn served_document_uri() -> String {
|
||||
use base64::Engine;
|
||||
format!(
|
||||
"data:image/png;base64,{}",
|
||||
base64::engine::general_purpose::STANDARD.encode(SERVED_DOCUMENT)
|
||||
)
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::azure_ai(Route::AzureAi)]
|
||||
#[case::vertex_mistral(Route::VertexMistral)]
|
||||
#[case::azure_cohere_parse(Route::AzureCohereParse)]
|
||||
#[tokio::test]
|
||||
async fn inlining_routes_send_the_downloaded_document(#[case] route: Route) {
|
||||
let (document_base, _documents) = document_server().await;
|
||||
let sent = send(route, Host::Detached, &document_base).await;
|
||||
sent.result.unwrap();
|
||||
assert_eq!(
|
||||
sent.provider_body.unwrap()["document"][route.document_type()],
|
||||
json!(served_document_uri())
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[tokio::test]
|
||||
async fn document_replaced_by_the_host_reaches_the_provider(
|
||||
#[values(
|
||||
Route::Mistral,
|
||||
Route::AzureAi,
|
||||
Route::VertexMistral,
|
||||
Route::AzureCohereParse,
|
||||
Route::Cohere
|
||||
)]
|
||||
route: Route,
|
||||
) {
|
||||
let (document_base, _documents) = document_server().await;
|
||||
let sent = send(route, Host::ReplacesDocument, &document_base).await;
|
||||
sent.result.unwrap();
|
||||
assert_eq!(
|
||||
sent.provider_body.unwrap()["document"][route.document_type()],
|
||||
json!(REPLACED_DOCUMENT)
|
||||
);
|
||||
}
|
||||
|
|
@ -1,203 +0,0 @@
|
|||
use std::sync::{Arc, Mutex};
|
||||
|
||||
use futures_util::future::BoxFuture;
|
||||
use litellm_host::event::WireRequest;
|
||||
use litellm_llms::base_llm::ocr::{
|
||||
error::Error,
|
||||
handler::{CallHooks, OcrClient},
|
||||
transformation::LiteLLMOcrResponse,
|
||||
};
|
||||
use serde_json::{Value, json};
|
||||
use tokio::{
|
||||
io::{AsyncReadExt, AsyncWriteExt},
|
||||
net::TcpListener,
|
||||
};
|
||||
|
||||
use crate::ocr::{
|
||||
route::{LocalOcrHost, ocr_machine},
|
||||
types::LiteLLMOcrRequest,
|
||||
wire::{OcrWireRequest, decode_request},
|
||||
};
|
||||
|
||||
/// Stands in for a host with no hooks registered: the wire request goes out unchanged
|
||||
/// and response events go nowhere.
|
||||
pub(crate) struct NoHooks;
|
||||
|
||||
impl CallHooks<Error> for NoHooks {
|
||||
fn before_send(&self, wire: WireRequest) -> BoxFuture<'_, Result<WireRequest, Error>> {
|
||||
Box::pin(async move { Ok(wire) })
|
||||
}
|
||||
|
||||
fn response_received<'a>(&'a self, _body: &'a [u8]) -> BoxFuture<'a, Result<(), Error>> {
|
||||
Box::pin(async { Ok(()) })
|
||||
}
|
||||
}
|
||||
|
||||
pub(crate) fn ocr_client() -> OcrClient {
|
||||
let document_http = reqwest::Client::builder()
|
||||
.redirect(reqwest::redirect::Policy::none())
|
||||
.build()
|
||||
.expect("test document client builds");
|
||||
OcrClient::for_test(reqwest::Client::new(), document_http)
|
||||
}
|
||||
|
||||
pub(crate) async fn perform_ocr(request: LiteLLMOcrRequest) -> Result<LiteLLMOcrResponse, Error> {
|
||||
crate::ocr::client::perform(&ocr_client(), request).await
|
||||
}
|
||||
|
||||
pub(crate) async fn perform_ocr_with(host: LocalOcrHost) -> Result<LiteLLMOcrResponse, Error> {
|
||||
litellm_host::run::run(ocr_machine(ocr_client()), &host).await
|
||||
}
|
||||
|
||||
pub(crate) fn wire_request(model: &str, base: &str, options: Value) -> LiteLLMOcrRequest {
|
||||
wire_request_with_document(
|
||||
model,
|
||||
base,
|
||||
json!({"type":"document_url","document_url":"data:application/pdf;base64,YWJj"}),
|
||||
options,
|
||||
)
|
||||
}
|
||||
|
||||
pub(crate) fn wire_request_with_document(
|
||||
model: &str,
|
||||
base: &str,
|
||||
document: Value,
|
||||
options: Value,
|
||||
) -> LiteLLMOcrRequest {
|
||||
decode_request(OcrWireRequest {
|
||||
model: model.into(),
|
||||
document,
|
||||
api_key: Some(litellm_auth::SecretValue::new("test-key")),
|
||||
api_base: Some(base.into()),
|
||||
custom_llm_provider: None,
|
||||
extra_headers: None,
|
||||
optional_params: options.as_object().unwrap().clone(),
|
||||
input_sources: Default::default(),
|
||||
timeout_seconds: Some(2.0),
|
||||
})
|
||||
.unwrap()
|
||||
}
|
||||
|
||||
pub(crate) fn resolved_request(
|
||||
request: LiteLLMOcrRequest,
|
||||
) -> crate::ocr::types::ResolvedOcrRequest {
|
||||
request
|
||||
.map_document(crate::ocr::document::prepare_document)
|
||||
.unwrap()
|
||||
}
|
||||
|
||||
pub(crate) fn with_source(request: LiteLLMOcrRequest, source: &str) -> LiteLLMOcrRequest {
|
||||
let request = resolved_request(request);
|
||||
let document = request.document.clone().with_source(source.into());
|
||||
request.with_document(document.into())
|
||||
}
|
||||
|
||||
pub(crate) fn request_body(request: &str) -> Value {
|
||||
serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap()
|
||||
}
|
||||
|
||||
pub(crate) const SERVED_DOCUMENT: &[u8] = b"\x89PNG served document";
|
||||
|
||||
/// Serves [`SERVED_DOCUMENT`] as `image/png` to every connection until aborted.
|
||||
pub(crate) async fn document_server() -> (String, tokio::task::JoinHandle<()>) {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
|
||||
let base = format!("http://{}", listener.local_addr().unwrap());
|
||||
let task = tokio::spawn(async move {
|
||||
loop {
|
||||
let (mut socket, _) = listener.accept().await.unwrap();
|
||||
let mut buffer = [0u8; 4096];
|
||||
let _ = socket.read(&mut buffer).await.unwrap();
|
||||
let head = format!(
|
||||
"HTTP/1.1 200 OK\r\nContent-Type: image/png\r\nContent-Length: {}\r\nConnection: close\r\n\r\n",
|
||||
SERVED_DOCUMENT.len()
|
||||
);
|
||||
socket.write_all(head.as_bytes()).await.unwrap();
|
||||
socket.write_all(SERVED_DOCUMENT).await.unwrap();
|
||||
}
|
||||
});
|
||||
(base, task)
|
||||
}
|
||||
|
||||
pub(crate) struct MockResponse {
|
||||
pub status: u16,
|
||||
pub headers: Vec<(&'static str, String)>,
|
||||
pub body: Value,
|
||||
}
|
||||
|
||||
impl MockResponse {
|
||||
pub fn json(body: Value) -> Self {
|
||||
Self {
|
||||
status: 200,
|
||||
headers: vec![],
|
||||
body,
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
pub(crate) async fn mock_server(
|
||||
responses: Vec<MockResponse>,
|
||||
) -> (String, Arc<Mutex<Vec<String>>>, tokio::task::JoinHandle<()>) {
|
||||
let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
|
||||
let base = format!("http://{}", listener.local_addr().unwrap());
|
||||
let requests = Arc::new(Mutex::new(Vec::new()));
|
||||
let seen = requests.clone();
|
||||
let server_base = base.clone();
|
||||
let task = tokio::spawn(async move {
|
||||
for response in responses {
|
||||
let (mut socket, _) = listener.accept().await.unwrap();
|
||||
let mut bytes = Vec::new();
|
||||
let mut buffer = [0u8; 4096];
|
||||
let header_end = loop {
|
||||
let n = socket.read(&mut buffer).await.unwrap();
|
||||
assert!(n > 0);
|
||||
bytes.extend_from_slice(&buffer[..n]);
|
||||
if let Some(index) = bytes.windows(4).position(|s| s == b"\r\n\r\n") {
|
||||
break index + 4;
|
||||
}
|
||||
};
|
||||
let length = String::from_utf8_lossy(&bytes[..header_end])
|
||||
.lines()
|
||||
.find_map(|line| {
|
||||
let (name, value) = line.split_once(':')?;
|
||||
name.eq_ignore_ascii_case("content-length")
|
||||
.then(|| value.trim().parse::<usize>().unwrap())
|
||||
})
|
||||
.unwrap_or(0);
|
||||
while bytes.len() < header_end + length {
|
||||
let n = socket.read(&mut buffer).await.unwrap();
|
||||
assert!(n > 0);
|
||||
bytes.extend_from_slice(&buffer[..n]);
|
||||
}
|
||||
seen.lock()
|
||||
.unwrap()
|
||||
.push(String::from_utf8_lossy(&bytes).into_owned());
|
||||
let body = serde_json::to_vec(&response.body).unwrap();
|
||||
let headers = response
|
||||
.headers
|
||||
.into_iter()
|
||||
.map(|(name, value)| {
|
||||
format!("{name}: {}\r\n", value.replace("{base}", &server_base))
|
||||
})
|
||||
.collect::<String>();
|
||||
let head = format!(
|
||||
"HTTP/1.1 {} OK\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n{}\r\n",
|
||||
response.status,
|
||||
body.len(),
|
||||
headers
|
||||
);
|
||||
socket.write_all(head.as_bytes()).await.unwrap();
|
||||
socket.write_all(&body).await.unwrap();
|
||||
}
|
||||
});
|
||||
(base, requests, task)
|
||||
}
|
||||
|
||||
pub(crate) fn header<'a>(request: &'a str, name: &str) -> Option<&'a str> {
|
||||
request
|
||||
.lines()
|
||||
.take_while(|line| !line.is_empty())
|
||||
.find_map(|line| {
|
||||
let (key, value) = line.split_once(':')?;
|
||||
key.eq_ignore_ascii_case(name).then(|| value.trim())
|
||||
})
|
||||
}
|
||||
|
|
@ -1,584 +0,0 @@
|
|||
use litellm_host::event::{CallEvent, MachineEvent, WireRequest};
|
||||
use litellm_llms::base_llm::ocr::{error::Error, transformation::OcrDocument};
|
||||
use rstest::rstest;
|
||||
use serde_json::{Value, json};
|
||||
|
||||
use super::test_support::{MockResponse, mock_server, perform_ocr, perform_ocr_with, wire_request};
|
||||
use crate::ocr::route::LocalOcrHost;
|
||||
|
||||
fn request_body(request: &str) -> Value {
|
||||
serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap()
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case(
|
||||
"reducto/parse-v3",
|
||||
json!({
|
||||
"formatting":{"table_output_format":"html"},
|
||||
"retrieval":{"chunk_mode":"section"},
|
||||
"settings":{"ocr_system":"standard"},
|
||||
"future_ocr_option":true,
|
||||
"extra_body":{"provider_option":"value"}
|
||||
}),
|
||||
"reducto://already.pdf",
|
||||
json!({
|
||||
"input":"reducto://already.pdf",
|
||||
"formatting":{"table_output_format":"html"},
|
||||
"retrieval":{"chunk_mode":"section"},
|
||||
"settings":{"ocr_system":"standard"},
|
||||
"future_ocr_option":true,
|
||||
"provider_option":"value"
|
||||
})
|
||||
)]
|
||||
#[case(
|
||||
"reducto/parse-legacy",
|
||||
json!({
|
||||
"enhance":{"agentic":[{"type":"table"}]},
|
||||
"future_ocr_option":true,
|
||||
"extra_body":{"provider_option":"value"}
|
||||
}),
|
||||
"reducto://legacy.pdf",
|
||||
json!({
|
||||
"document_url":"reducto://legacy.pdf",
|
||||
"options":{"enhance":{"agentic":[{"type":"table"}]}},
|
||||
"future_ocr_option":true,
|
||||
"provider_option":"value"
|
||||
})
|
||||
)]
|
||||
#[tokio::test]
|
||||
async fn request_mapping_matches_python(
|
||||
#[case] model: &str,
|
||||
#[case] options: Value,
|
||||
#[case] source: &str,
|
||||
#[case] expected: Value,
|
||||
) {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"result":{"chunks":[]}
|
||||
}))])
|
||||
.await;
|
||||
let request = super::test_support::with_source(wire_request(model, &base, options), source);
|
||||
|
||||
perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 1);
|
||||
assert!(requests[0].starts_with("POST /parse "));
|
||||
assert_eq!(request_body(&requests[0]), expected);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case("parse-v3")]
|
||||
#[case("parse-legacy")]
|
||||
#[tokio::test]
|
||||
async fn data_uri_upload_preserves_multipart_headers(
|
||||
#[case] model: &str,
|
||||
#[values("application/pdf", "image/png")] mime_type: &str,
|
||||
) {
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse::json(json!({"file_id":"reducto://uploaded.pdf"})),
|
||||
MockResponse::json(json!({"result":{"chunks":[{"content":"hello"}]}})),
|
||||
])
|
||||
.await;
|
||||
let document = if mime_type.starts_with("image/") {
|
||||
json!({"type":"image_url","image_url":format!("data:{mime_type};base64,YWJj")})
|
||||
} else {
|
||||
json!({"type":"document_url","document_url":format!("data:{mime_type};base64,YWJj")})
|
||||
};
|
||||
let mut request = crate::ocr::types::LiteLLMOcrRequest {
|
||||
document: serde_json::from_value::<OcrDocument>(document)
|
||||
.unwrap()
|
||||
.into(),
|
||||
..wire_request(&format!("reducto/{model}"), &base, json!({}))
|
||||
};
|
||||
request.transport.extra_headers = vec![
|
||||
("Content-Type".into(), "application/json".into()),
|
||||
("X-Trace".into(), "upload-test".into()),
|
||||
];
|
||||
|
||||
let response = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(response.pages[0].markdown, "hello");
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 2);
|
||||
assert!(requests[0].starts_with("POST /upload "));
|
||||
assert!(
|
||||
requests[0]
|
||||
.to_ascii_lowercase()
|
||||
.contains("content-type: multipart/form-data; boundary=")
|
||||
);
|
||||
assert!(requests[0].contains("x-trace: upload-test"));
|
||||
let multipart = requests[0].split_once("\r\n\r\n").unwrap().1;
|
||||
assert!(multipart.contains(&format!("Content-Type: {mime_type}\r\n")));
|
||||
assert!(multipart.contains("\r\n\r\nabc\r\n--"));
|
||||
assert!(requests[1].starts_with("POST /parse "));
|
||||
let source_field = if model == "parse-legacy" {
|
||||
"document_url"
|
||||
} else {
|
||||
"input"
|
||||
};
|
||||
assert_eq!(
|
||||
request_body(&requests[1]),
|
||||
json!({source_field:"reducto://uploaded.pdf"})
|
||||
);
|
||||
for request in requests.iter() {
|
||||
assert!(
|
||||
request
|
||||
.to_ascii_lowercase()
|
||||
.contains("authorization: bearer test-key\r\n")
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn response_received_stays_after_reducto_upload_and_parse() {
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse::json(json!({"file_id":"reducto://uploaded.pdf"})),
|
||||
MockResponse::json(json!({"result":{"chunks":[]}})),
|
||||
])
|
||||
.await;
|
||||
let request_count = seen.clone();
|
||||
let host = LocalOcrHost::new(wire_request("reducto/parse-v3", &base, json!({}))).with_observer(
|
||||
move |event| {
|
||||
if let CallEvent::Machine(MachineEvent::ResponseReceived { raw }) = event {
|
||||
assert_eq!(request_count.lock().unwrap().len(), 2);
|
||||
assert_eq!(raw.body, r#"{"result":{"chunks":[]}}"#);
|
||||
}
|
||||
},
|
||||
);
|
||||
|
||||
perform_ocr_with(host).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(seen.lock().unwrap().len(), 2);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case(json!({"file_id":""}))]
|
||||
#[case(json!({}))]
|
||||
#[case(json!({"file_id":null}))]
|
||||
#[tokio::test]
|
||||
async fn invalid_upload_ids_stop_before_parse(#[case] response: Value) {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(response)]).await;
|
||||
let error = perform_ocr(wire_request("reducto/parse-v3", &base, json!({})))
|
||||
.await
|
||||
.unwrap_err();
|
||||
server.await.unwrap();
|
||||
assert!(error.to_string().contains("file_id"));
|
||||
assert_eq!(seen.lock().unwrap().len(), 1);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn upload_failure_stops_before_parse() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse {
|
||||
status: 503,
|
||||
headers: vec![],
|
||||
body: json!({"error":"unavailable"}),
|
||||
}])
|
||||
.await;
|
||||
assert!(
|
||||
perform_ocr(wire_request("reducto/parse-v3", &base, json!({})))
|
||||
.await
|
||||
.is_err()
|
||||
);
|
||||
server.await.unwrap();
|
||||
assert_eq!(seen.lock().unwrap().len(), 1);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case("https://example.com/a.pdf", Error::ReductoSource)]
|
||||
#[case("reducto://", Error::RequestField { path: "document file id".into() })]
|
||||
#[case("data:application/pdf;base64", Error::InvalidDataUri)]
|
||||
#[case("data:application/pdf;base64,INVALID!", Error::InvalidDataUri)]
|
||||
#[tokio::test]
|
||||
async fn rejects_invalid_document_sources_before_network(
|
||||
#[case] source: &str,
|
||||
#[case] expected: Error,
|
||||
) {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({}))]).await;
|
||||
let request = super::test_support::with_source(
|
||||
wire_request("reducto/parse-v3", &base, json!({})),
|
||||
source,
|
||||
);
|
||||
let result = perform_ocr(request).await;
|
||||
server.abort();
|
||||
let _ = server.await;
|
||||
assert!(
|
||||
seen.lock().unwrap().is_empty(),
|
||||
"sent invalid source: {source}"
|
||||
);
|
||||
let error = result.unwrap_err();
|
||||
assert_eq!(
|
||||
std::mem::discriminant(&error),
|
||||
std::mem::discriminant(&expected)
|
||||
);
|
||||
assert_eq!(error.http_status_code(), Some(400));
|
||||
assert_eq!(error.to_string(), expected.to_string());
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn response_normalization_groups_blocks_and_distinguishes_null_result() {
|
||||
use litellm_llms::reducto::ocr::transformation::{
|
||||
ReductoResponse, normalize_response as transform_ocr_response,
|
||||
};
|
||||
|
||||
let raw = json!({"usage":{"num_pages":"2","credits":"3"},"result":{"type":"full","chunks":[
|
||||
{"blocks":[{
|
||||
"type":"Table",
|
||||
"content":"B",
|
||||
"bbox":{"left":0.1,"top":0.2,"width":0.8,"height":0.3,"page":2,"original_page":4},
|
||||
"confidence":"high",
|
||||
"granular_confidence":{"parse_confidence":0.95,"extract_confidence":null},
|
||||
"image_url":null
|
||||
}]},
|
||||
{"blocks":[{"content":"A","bbox":{"page":1},"type":"Text"},{"content":"C","bbox":{"page":1}}]}
|
||||
]}});
|
||||
let response: ReductoResponse = serde_json::from_value(raw).unwrap();
|
||||
let normalized = transform_ocr_response("parse-v3", response)
|
||||
.unwrap()
|
||||
.into_json();
|
||||
assert_eq!(normalized["pages"][0]["markdown"], "A\n\nC");
|
||||
assert_eq!(normalized["pages"][1]["markdown"], "B");
|
||||
assert_eq!(normalized["pages"][1]["blocks"][0]["type"], "Table");
|
||||
assert_eq!(
|
||||
normalized["pages"][1]["blocks"][0]["bbox"],
|
||||
json!({"left":0.1,"top":0.2,"width":0.8,"height":0.3,"page":2,"original_page":4})
|
||||
);
|
||||
assert_eq!(normalized["pages"][1]["blocks"][0]["confidence"], "high");
|
||||
assert_eq!(
|
||||
normalized["pages"][1]["blocks"][0]["granular_confidence"]["parse_confidence"],
|
||||
0.95
|
||||
);
|
||||
assert!(normalized["pages"][1]["blocks"][0]["image_url"].is_null());
|
||||
assert_eq!(normalized["usage_info"]["pages_processed"], 2);
|
||||
assert_eq!(normalized["usage_info"]["credits"], 3.0);
|
||||
|
||||
let missing: ReductoResponse =
|
||||
serde_json::from_value(json!({"chunks":[{"content":"text"}]})).unwrap();
|
||||
let missing = transform_ocr_response("parse-v3", missing).unwrap();
|
||||
assert_eq!(missing.pages[0].markdown, "text");
|
||||
let null: ReductoResponse = serde_json::from_value(
|
||||
json!({"result":null,"chunks":[{"content":"ignored"}],"usage":null}),
|
||||
)
|
||||
.unwrap();
|
||||
let null = transform_ocr_response("parse-v3", null).unwrap();
|
||||
assert!(null.pages.is_empty());
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn facade_omits_native_response_by_default_and_preserves_auth_priority() {
|
||||
let raw = json!({"job_id":"job-1","result":{"chunks":[]}});
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(raw)]).await;
|
||||
let mut request = super::test_support::with_source(
|
||||
wire_request("reducto/parse-v3", &base, json!({})),
|
||||
"reducto://ready.pdf",
|
||||
);
|
||||
request.transport.extra_headers = vec![("authorization".into(), "Bearer existing".into())];
|
||||
|
||||
let response = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(response.provider_native_response, None);
|
||||
assert!(
|
||||
seen.lock().unwrap()[0]
|
||||
.to_ascii_lowercase()
|
||||
.contains("authorization: bearer existing")
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn native_format_retains_the_provider_response() {
|
||||
let raw = json!({
|
||||
"result":{"chunks":[{"content":"native OCR response"}]},
|
||||
"usage":{"num_pages":1}
|
||||
});
|
||||
let (base, _, server) = mock_server(vec![MockResponse::json(raw.clone())]).await;
|
||||
let request = super::test_support::with_source(
|
||||
wire_request("reducto/parse-v3", &base, json!({"req_format":"native"})),
|
||||
"reducto://ready.pdf",
|
||||
);
|
||||
|
||||
let response = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
assert_eq!(response.pages[0].markdown, "native OCR response");
|
||||
assert_eq!(response.provider_native_response.as_ref(), raw.as_object());
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn unknown_model_reaches_parse_and_keeps_its_name() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"result":{"chunks":[{"content":"future model response"}]}
|
||||
}))])
|
||||
.await;
|
||||
let request = super::test_support::with_source(
|
||||
wire_request("reducto/future-parse-model", &base, json!({})),
|
||||
"reducto://ready.pdf",
|
||||
);
|
||||
|
||||
let response = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
|
||||
assert_eq!(response.model, "future-parse-model");
|
||||
assert_eq!(response.pages[0].markdown, "future model response");
|
||||
let requests = seen.lock().unwrap();
|
||||
assert!(requests[0].starts_with("POST /parse "));
|
||||
assert_eq!(
|
||||
request_body(&requests[0]),
|
||||
json!({"input":"reducto://ready.pdf"})
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn guardrail_rewrites_document_before_upload() {
|
||||
let (base, seen, server) =
|
||||
mock_server(vec![MockResponse::json(json!({"result":{"chunks":[]}}))]).await;
|
||||
let host = LocalOcrHost::new(wire_request("reducto/parse-v3", &base, json!({})))
|
||||
.with_before_send(|wire, _| {
|
||||
assert_eq!(
|
||||
wire.body["document_url"],
|
||||
"data:application/pdf;base64,YWJj"
|
||||
);
|
||||
Ok(WireRequest {
|
||||
body: json!({"type":"document_url","document_url":"reducto://guarded.pdf"}),
|
||||
..wire
|
||||
})
|
||||
});
|
||||
|
||||
perform_ocr_with(host).await.unwrap();
|
||||
server.await.unwrap();
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 1);
|
||||
assert!(requests[0].starts_with("POST /parse "));
|
||||
assert!(requests[0].contains("reducto://guarded.pdf"));
|
||||
}
|
||||
|
||||
mod transformation {
|
||||
use litellm_host::event::{CallEvent, MachineEvent, WireRequest};
|
||||
use litellm_llms::{
|
||||
base_llm::ocr::transformation::{BaseOcrConfig, OcrConnection, OcrRequestContext},
|
||||
reducto::ocr::transformation::*,
|
||||
};
|
||||
use rstest::rstest;
|
||||
|
||||
use super::*;
|
||||
use crate::ocr::{
|
||||
route::LocalOcrHost,
|
||||
test_support::{MockResponse, mock_server, perform_ocr, perform_ocr_with, wire_request},
|
||||
};
|
||||
|
||||
#[tokio::test]
|
||||
async fn v3_options_preserve_explicit_null() {
|
||||
let overrides =
|
||||
serde_json::from_value(json!({"formatting":null,"settings":{},"unknown":true}))
|
||||
.unwrap();
|
||||
let params = ReductoParseV3Config
|
||||
.map_ocr_params(&overrides, "parse-v3")
|
||||
.unwrap();
|
||||
let client = crate::ocr::test_support::ocr_client();
|
||||
let connection = OcrConnection::default();
|
||||
let document = serde_json::from_value(
|
||||
json!({"type":"document_url","document_url":"reducto://ready.pdf"}),
|
||||
)
|
||||
.unwrap();
|
||||
let body = ReductoParseV3Config
|
||||
.async_transform_ocr_request(
|
||||
"parse-v3",
|
||||
document,
|
||||
¶ms,
|
||||
&[],
|
||||
OcrRequestContext {
|
||||
client: &client,
|
||||
connection: &connection,
|
||||
},
|
||||
)
|
||||
.await
|
||||
.unwrap();
|
||||
assert_eq!(
|
||||
serde_json::to_value(body).unwrap(),
|
||||
json!({
|
||||
"input":"reducto://ready.pdf", "formatting":null, "settings":{}
|
||||
})
|
||||
);
|
||||
let absent = ReductoParseV3Config
|
||||
.map_ocr_params(
|
||||
&litellm_core_utils::call_arguments::CallArguments::default(),
|
||||
"parse-v3",
|
||||
)
|
||||
.unwrap();
|
||||
assert_eq!(serde_json::to_value(absent).unwrap(), json!({}));
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case(
|
||||
"reducto/parse-v3",
|
||||
json!({
|
||||
"formatting":{"table_output_format":"html"},
|
||||
"retrieval":{"chunk_mode":"section"},
|
||||
"settings":{"ocr_system":"standard"},
|
||||
"future_ocr_option":true,
|
||||
"extra_body":{"provider_option":"value"}
|
||||
}),
|
||||
"reducto://already.pdf",
|
||||
json!({
|
||||
"input":"reducto://already.pdf",
|
||||
"formatting":{"table_output_format":"html"},
|
||||
"retrieval":{"chunk_mode":"section"},
|
||||
"settings":{"ocr_system":"standard"},
|
||||
"future_ocr_option":true,
|
||||
"provider_option":"value"
|
||||
})
|
||||
)]
|
||||
#[case(
|
||||
"reducto/parse-legacy",
|
||||
json!({
|
||||
"enhance":{"agentic":[{"type":"table"}]},
|
||||
"future_ocr_option":true,
|
||||
"extra_body":{"provider_option":"value"}
|
||||
}),
|
||||
"reducto://legacy.pdf",
|
||||
json!({
|
||||
"document_url":"reducto://legacy.pdf",
|
||||
"options":{"enhance":{"agentic":[{"type":"table"}]}},
|
||||
"future_ocr_option":true,
|
||||
"provider_option":"value"
|
||||
})
|
||||
)]
|
||||
#[tokio::test]
|
||||
async fn request_mapping_matches_python(
|
||||
#[case] model: &str,
|
||||
#[case] options: Value,
|
||||
#[case] source: &str,
|
||||
#[case] expected: Value,
|
||||
) {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"result":{"chunks":[]}
|
||||
}))])
|
||||
.await;
|
||||
let request =
|
||||
crate::ocr::test_support::with_source(wire_request(model, &base, options), source);
|
||||
|
||||
perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 1);
|
||||
assert!(requests[0].starts_with("POST /parse "));
|
||||
assert_eq!(request_body(&requests[0]), expected);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case("parse-v3")]
|
||||
#[case("parse-legacy")]
|
||||
#[tokio::test]
|
||||
async fn data_uri_upload_preserves_multipart_headers(#[case] model: &str) {
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse::json(json!({"file_id":"reducto://uploaded.pdf"})),
|
||||
MockResponse::json(json!({"result":{"chunks":[{"content":"hello"}]}})),
|
||||
])
|
||||
.await;
|
||||
let mut request = wire_request(&format!("reducto/{model}"), &base, json!({}));
|
||||
request.transport.extra_headers = vec![
|
||||
("Content-Type".into(), "application/json".into()),
|
||||
("X-Trace".into(), "upload-test".into()),
|
||||
];
|
||||
|
||||
let response = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(response.pages[0].markdown, "hello");
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 2);
|
||||
assert!(requests[0].starts_with("POST /upload "));
|
||||
assert!(
|
||||
requests[0]
|
||||
.to_ascii_lowercase()
|
||||
.contains("content-type: multipart/form-data; boundary=")
|
||||
);
|
||||
assert!(requests[0].contains("x-trace: upload-test"));
|
||||
assert!(requests[0].contains("application/pdf"));
|
||||
assert!(requests[0].contains("abc"));
|
||||
assert!(requests[1].starts_with("POST /parse "));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn response_received_stays_after_reducto_upload_and_parse() {
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse::json(json!({"file_id":"reducto://uploaded.pdf"})),
|
||||
MockResponse::json(json!({"result":{"chunks":[]}})),
|
||||
])
|
||||
.await;
|
||||
let request_count = seen.clone();
|
||||
let host = LocalOcrHost::new(wire_request("reducto/parse-v3", &base, json!({})))
|
||||
.with_observer(move |event| {
|
||||
if let CallEvent::Machine(MachineEvent::ResponseReceived { raw }) = event {
|
||||
assert_eq!(request_count.lock().unwrap().len(), 2);
|
||||
assert_eq!(raw.body, r#"{"result":{"chunks":[]}}"#);
|
||||
}
|
||||
});
|
||||
|
||||
perform_ocr_with(host).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(seen.lock().unwrap().len(), 2);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case("https://example.com/a.pdf")]
|
||||
#[case("reducto://")]
|
||||
#[case("data:application/pdf;base64")]
|
||||
#[case("data:application/pdf;base64,INVALID!")]
|
||||
#[tokio::test]
|
||||
async fn rejects_invalid_document_sources_before_network(#[case] source: &str) {
|
||||
let request = crate::ocr::test_support::with_source(
|
||||
wire_request("reducto/parse-v3", "http://127.0.0.1:1", json!({})),
|
||||
source,
|
||||
);
|
||||
assert!(perform_ocr(request).await.is_err());
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn facade_omits_native_response_by_default_and_preserves_auth_priority() {
|
||||
let raw = json!({"job_id":"job-1","result":{"chunks":[]}});
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(raw)]).await;
|
||||
let mut request = crate::ocr::test_support::with_source(
|
||||
wire_request("reducto/parse-v3", &base, json!({})),
|
||||
"reducto://ready.pdf",
|
||||
);
|
||||
request.transport.extra_headers = vec![("authorization".into(), "Bearer existing".into())];
|
||||
|
||||
let response = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(response.provider_native_response, None);
|
||||
assert!(
|
||||
seen.lock().unwrap()[0]
|
||||
.to_ascii_lowercase()
|
||||
.contains("authorization: bearer existing")
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case("reducto/parse-v3")]
|
||||
#[case("reducto/parse-legacy")]
|
||||
#[tokio::test]
|
||||
async fn guardrail_headers_reach_upload_and_parse(#[case] model: &str) {
|
||||
let (base, seen, server) = mock_server(vec![
|
||||
MockResponse::json(json!({"file_id":"reducto://uploaded.pdf"})),
|
||||
MockResponse::json(json!({"result":{"chunks":[]}})),
|
||||
])
|
||||
.await;
|
||||
let mut request = wire_request(model, &base, json!({}));
|
||||
request.transport.extra_headers = vec![("authorization".into(), "Bearer original".into())];
|
||||
let host = LocalOcrHost::new(request).with_before_send(|wire, _| {
|
||||
Ok(WireRequest {
|
||||
headers: vec![("authorization".into(), "Bearer guarded".into())],
|
||||
..wire
|
||||
})
|
||||
});
|
||||
|
||||
perform_ocr_with(host).await.unwrap();
|
||||
server.await.unwrap();
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 2);
|
||||
assert!(requests[0].starts_with("POST /upload "));
|
||||
assert!(requests[1].starts_with("POST /parse "));
|
||||
for request in requests.iter() {
|
||||
assert!(request.contains("authorization: Bearer guarded"));
|
||||
assert!(!request.contains("Bearer original"));
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
@ -1,143 +0,0 @@
|
|||
use litellm_auth::InputSource;
|
||||
use serde_json::{Value, json};
|
||||
|
||||
use super::test_support::{MockResponse, mock_server, perform_ocr, wire_request};
|
||||
|
||||
fn request_body(request: &str) -> Value {
|
||||
serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap()
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn facade_executes_vertex_deepseek_at_the_openai_endpoint() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"choices":[{"message":{"content":"recognized"}}],
|
||||
"usage":{"prompt_tokens":1}
|
||||
}))])
|
||||
.await;
|
||||
let request = wire_request(
|
||||
"vertex_ai/deepseek-ocr-maas",
|
||||
&base,
|
||||
json!({
|
||||
"vertex_project":"project-1",
|
||||
"vertex_location":"europe-west4",
|
||||
"temperature":0.1,
|
||||
"future_ocr_option":true,
|
||||
"extra_body":{"provider_option":"value"}
|
||||
}),
|
||||
);
|
||||
let request = super::test_support::with_source(request, "gs://bucket/document.pdf");
|
||||
|
||||
let response = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(response.pages[0].markdown, "recognized");
|
||||
assert_eq!(
|
||||
response.usage_info.unwrap().extra_fields["prompt_tokens"],
|
||||
1
|
||||
);
|
||||
let requests = seen.lock().unwrap();
|
||||
assert!(requests[0].starts_with(
|
||||
"POST /v1/projects/project-1/locations/europe-west4/endpoints/openapi/chat/completions "
|
||||
));
|
||||
assert!(
|
||||
requests[0]
|
||||
.to_ascii_lowercase()
|
||||
.contains("authorization: bearer test-key")
|
||||
);
|
||||
let body = request_body(&requests[0]);
|
||||
assert_eq!(body["model"], "deepseek-ai/deepseek-ocr-maas");
|
||||
assert_eq!(body["temperature"], 0.1);
|
||||
assert_eq!(body["future_ocr_option"], true);
|
||||
assert!(body.get("extra_body").is_none());
|
||||
assert_eq!(
|
||||
body["messages"][0]["content"][0],
|
||||
json!({"type":"image_url","image_url":"gs://bucket/document.pdf"})
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn host_registration_selects_deepseek_without_affecting_mistral() {
|
||||
assert!(crate::ocr::arguments::is_supported_request(
|
||||
"deepseek-ocr-maas",
|
||||
Some("vertex_ai")
|
||||
));
|
||||
assert!(crate::ocr::arguments::is_supported_request(
|
||||
"mistral-ocr-maas",
|
||||
Some("vertex_ai")
|
||||
));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn request_controlled_api_base_is_rejected_before_vertex_auth() {
|
||||
let mut request = wire_request(
|
||||
"vertex_ai/deepseek-ocr-maas",
|
||||
"https://caller.example",
|
||||
json!({"vertex_project":"project-1"}),
|
||||
);
|
||||
request.credentials.api_base = Some(litellm_auth::Sourced::new(
|
||||
"https://caller.example".into(),
|
||||
InputSource::Request,
|
||||
));
|
||||
|
||||
let error = perform_ocr(request).await.unwrap_err();
|
||||
assert!(
|
||||
error
|
||||
.to_string()
|
||||
.contains("request-controlled Vertex AI endpoint")
|
||||
);
|
||||
}
|
||||
|
||||
mod deepseek_transformation {
|
||||
use serde_json::json;
|
||||
|
||||
use super::*;
|
||||
use crate::ocr::test_support::{MockResponse, mock_server, perform_ocr, wire_request};
|
||||
|
||||
#[tokio::test]
|
||||
async fn facade_executes_vertex_deepseek_at_the_openai_endpoint() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"choices":[{"message":{"content":"recognized"}}],
|
||||
"usage":{"prompt_tokens":1}
|
||||
}))])
|
||||
.await;
|
||||
let request = wire_request(
|
||||
"vertex_ai/deepseek-ocr-maas",
|
||||
&base,
|
||||
json!({
|
||||
"vertex_project":"project-1",
|
||||
"vertex_location":"europe-west4",
|
||||
"temperature":0.1,
|
||||
"future_ocr_option":true,
|
||||
"extra_body":{"provider_option":"value"}
|
||||
}),
|
||||
);
|
||||
let request = crate::ocr::test_support::with_source(request, "gs://bucket/document.pdf");
|
||||
|
||||
let response = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(response.pages[0].markdown, "recognized");
|
||||
assert_eq!(
|
||||
response.usage_info.unwrap().extra_fields["prompt_tokens"],
|
||||
1
|
||||
);
|
||||
let requests = seen.lock().unwrap();
|
||||
assert!(requests[0].starts_with(
|
||||
"POST /v1/projects/project-1/locations/europe-west4/endpoints/openapi/chat/completions "
|
||||
));
|
||||
assert!(
|
||||
requests[0]
|
||||
.to_ascii_lowercase()
|
||||
.contains("authorization: bearer test-key")
|
||||
);
|
||||
let body = request_body(&requests[0]);
|
||||
assert_eq!(body["model"], "deepseek-ai/deepseek-ocr-maas");
|
||||
assert_eq!(body["temperature"], 0.1);
|
||||
assert_eq!(body["future_ocr_option"], true);
|
||||
assert_eq!(body["provider_option"], "value");
|
||||
assert!(body.get("vertex_project").is_none());
|
||||
assert!(body.get("extra_body").is_none());
|
||||
assert_eq!(
|
||||
body["messages"][0]["content"][0],
|
||||
json!({"type":"image_url","image_url":"gs://bucket/document.pdf"})
|
||||
);
|
||||
}
|
||||
}
|
||||
|
|
@ -1,293 +0,0 @@
|
|||
use litellm_auth::InputSource;
|
||||
use litellm_llms::base_llm::ocr::{settings::OcrSettings, transformation::OcrResponseFormat};
|
||||
use serde_json::{Value, json};
|
||||
|
||||
use super::test_support::{MockResponse, mock_server, ocr_client, perform_ocr, wire_request};
|
||||
|
||||
fn request_body(request: &str) -> Value {
|
||||
serde_json::from_str(request.split_once("\r\n\r\n").unwrap().1).unwrap()
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn facade_executes_vertex_mistral_with_resolved_project_and_location() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({
|
||||
"pages":[{"index":0,"markdown":"hello"}],
|
||||
"usage_info":{"pages_processed":1}
|
||||
}))])
|
||||
.await;
|
||||
let request = wire_request(
|
||||
"vertex_ai/mistral-ocr-maas",
|
||||
&base,
|
||||
json!({
|
||||
"vertex_project":"project-1",
|
||||
"vertex_location":"europe-west4",
|
||||
"extract_footer":true
|
||||
}),
|
||||
);
|
||||
|
||||
let response = perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert_eq!(response.pages[0].markdown, "hello");
|
||||
let requests = seen.lock().unwrap();
|
||||
assert_eq!(requests.len(), 1);
|
||||
assert!(requests[0].starts_with(
|
||||
"POST /v1/projects/project-1/locations/europe-west4/publishers/mistralai/models/mistral-ocr-maas:rawPredict "
|
||||
));
|
||||
assert!(
|
||||
requests[0]
|
||||
.to_ascii_lowercase()
|
||||
.contains("authorization: bearer test-key")
|
||||
);
|
||||
assert_eq!(
|
||||
request_body(&requests[0]),
|
||||
json!({
|
||||
"model":"mistral-ocr-maas",
|
||||
"document":{"type":"document_url","document_url":"data:application/pdf;base64,YWJj"},
|
||||
"extract_footer":true
|
||||
})
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn configured_project_and_location_apply_when_the_call_sets_neither() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({"pages":[]}))]).await;
|
||||
let client = ocr_client().with_settings(OcrSettings {
|
||||
vertex_project: Some("configured-project".into()),
|
||||
vertex_location: Some("europe-west4".into()),
|
||||
..OcrSettings::default()
|
||||
});
|
||||
|
||||
crate::ocr::client::perform(
|
||||
&client,
|
||||
wire_request("vertex_ai/mistral-ocr-maas", &base, json!({})),
|
||||
)
|
||||
.await
|
||||
.unwrap();
|
||||
server.await.unwrap();
|
||||
assert!(seen.lock().unwrap()[0].starts_with(
|
||||
"POST /v1/projects/configured-project/locations/europe-west4/publishers/mistralai/models/mistral-ocr-maas:rawPredict "
|
||||
));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn supplied_authorization_is_forwarded_without_a_static_token() {
|
||||
let (base, seen, server) = mock_server(vec![MockResponse::json(json!({"pages":[]}))]).await;
|
||||
let mut request = wire_request(
|
||||
"vertex_ai/model",
|
||||
&base,
|
||||
json!({"vertex_project":"project-1"}),
|
||||
);
|
||||
request.credentials.api_key = None;
|
||||
request.transport.extra_headers = vec![("authorization".into(), "Bearer supplied".into())];
|
||||
|
||||
perform_ocr(request).await.unwrap();
|
||||
server.await.unwrap();
|
||||
assert!(
|
||||
seen.lock().unwrap()[0]
|
||||
.to_ascii_lowercase()
|
||||
.contains("authorization: bearer supplied")
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn invalid_credentials_fail_before_provider_http() {
|
||||
let request = wire_request(
|
||||
"vertex_ai/model",
|
||||
"http://127.0.0.1:1",
|
||||
json!({"vertex_credentials": true}),
|
||||
);
|
||||
let error = perform_ocr(request).await.unwrap_err();
|
||||
assert!(error.to_string().contains("vertex_credentials"));
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn request_controlled_api_base_is_rejected_before_vertex_auth() {
|
||||
let mut request = wire_request(
|
||||
"vertex_ai/mistral-ocr-maas",
|
||||
"https://caller.example",
|
||||
json!({"vertex_project":"project-1"}),
|
||||
);
|
||||
request.credentials.api_base = Some(litellm_auth::Sourced::new(
|
||||
"https://caller.example".into(),
|
||||
InputSource::Request,
|
||||
));
|
||||
|
||||
let error = perform_ocr(request).await.unwrap_err();
|
||||
assert!(
|
||||
error
|
||||
.to_string()
|
||||
.contains("request-controlled Vertex AI endpoint")
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn adapters_build_complete_requests_and_share_mistral_normalization() {
|
||||
use std::time::Duration;
|
||||
|
||||
use litellm_llms::{
|
||||
base_llm::ocr::transformation::BaseOcrConfig,
|
||||
mistral::ocr::transformation::MistralOcrConfig,
|
||||
vertex_ai::ocr::transformation::VertexAiOcrConfig,
|
||||
};
|
||||
|
||||
use crate::ocr::test_support::ocr_client;
|
||||
|
||||
let client = ocr_client();
|
||||
let options = json!({
|
||||
"pages": [0, 2],
|
||||
"include_image_base64": true,
|
||||
"vertex_project": "project-1",
|
||||
"vertex_location": "us-central1",
|
||||
"unknown": "ignored"
|
||||
});
|
||||
let direct = wire_request(
|
||||
"mistral/mistral-ocr-maas",
|
||||
"https://mistral.test",
|
||||
options.clone(),
|
||||
);
|
||||
let vertex = wire_request("vertex_ai/mistral-ocr-maas", "https://vertex.test", options);
|
||||
let direct = crate::ocr::prepare::prepare_request_for_test(
|
||||
super::test_support::resolved_request(direct),
|
||||
);
|
||||
let vertex = crate::ocr::prepare::prepare_request_for_test(
|
||||
super::test_support::resolved_request(vertex),
|
||||
);
|
||||
let direct_http = MistralOcrConfig
|
||||
.prepare_request(&direct, &client, &crate::ocr::test_support::NoHooks)
|
||||
.await
|
||||
.unwrap();
|
||||
let vertex_http = VertexAiOcrConfig
|
||||
.prepare_request(&vertex, &client, &crate::ocr::test_support::NoHooks)
|
||||
.await
|
||||
.unwrap();
|
||||
assert_eq!(direct_http.url(), "https://mistral.test/v1/ocr");
|
||||
assert_eq!(
|
||||
vertex_http.url(),
|
||||
"https://vertex.test/v1/projects/project-1/locations/us-central1/publishers/mistralai/models/mistral-ocr-maas:rawPredict"
|
||||
);
|
||||
for http in [&direct_http, &vertex_http] {
|
||||
assert_eq!(http.header("authorization").unwrap(), "Bearer test-key");
|
||||
assert_eq!(http.header("content-type").unwrap(), "application/json");
|
||||
assert_eq!(http.timeout(), Some(Duration::from_secs(2)));
|
||||
let body: Value = serde_json::from_slice(http.body()).unwrap();
|
||||
assert_eq!(
|
||||
body,
|
||||
json!({
|
||||
"model": "mistral-ocr-maas",
|
||||
"document": {"type": "document_url", "document_url": "data:application/pdf;base64,YWJj"},
|
||||
"pages": [0, 2],
|
||||
"include_image_base64": true,
|
||||
"unknown": "ignored"
|
||||
})
|
||||
);
|
||||
}
|
||||
let payload = json!({"pages": [{"index": 0, "markdown": "hello"}], "extra": "preserved"});
|
||||
let raw = serde_json::to_vec(&payload).unwrap();
|
||||
let direct_response = MistralOcrConfig
|
||||
.transform_ocr_response(&direct.model, &raw, OcrResponseFormat::Litellm)
|
||||
.unwrap()
|
||||
.into_json();
|
||||
let vertex_response = VertexAiOcrConfig
|
||||
.transform_ocr_response(&vertex.model, &raw, OcrResponseFormat::Litellm)
|
||||
.unwrap()
|
||||
.into_json();
|
||||
assert_eq!(direct_response, vertex_response);
|
||||
assert_eq!(direct_response["model"], "mistral-ocr-maas");
|
||||
assert_eq!(direct_response["object"], "ocr");
|
||||
assert_eq!(direct_response["extra"], "preserved");
|
||||
}
|
||||
|
||||
mod transformation {
|
||||
|
||||
use rstest::rstest;
|
||||
use serde_json::{Value, json};
|
||||
|
||||
use crate::ocr::test_support::wire_request;
|
||||
|
||||
#[rstest]
|
||||
#[case::mistral(false)]
|
||||
#[case::vertex(true)]
|
||||
#[tokio::test]
|
||||
async fn configs_build_complete_requests_and_share_mistral_normalization(
|
||||
#[case] use_vertex: bool,
|
||||
) {
|
||||
use std::time::Duration;
|
||||
|
||||
use litellm_llms::{
|
||||
base_llm::ocr::transformation::BaseOcrConfig,
|
||||
mistral::ocr::transformation::MistralOcrConfig,
|
||||
vertex_ai::ocr::transformation::VertexAiOcrConfig,
|
||||
};
|
||||
|
||||
use crate::ocr::test_support::ocr_client;
|
||||
|
||||
let client = ocr_client();
|
||||
let options = json!({
|
||||
"pages": [0, 2],
|
||||
"include_image_base64": true,
|
||||
"vertex_project": "project-1",
|
||||
"vertex_location": "us-central1",
|
||||
"unknown": "preserved"
|
||||
});
|
||||
let direct = wire_request(
|
||||
"mistral/mistral-ocr-maas",
|
||||
"https://mistral.test",
|
||||
options.clone(),
|
||||
);
|
||||
let vertex = wire_request("vertex_ai/mistral-ocr-maas", "https://vertex.test", options);
|
||||
let direct = crate::ocr::prepare::prepare_request_for_test(
|
||||
crate::ocr::test_support::resolved_request(direct),
|
||||
);
|
||||
let vertex = crate::ocr::prepare::prepare_request_for_test(
|
||||
crate::ocr::test_support::resolved_request(vertex),
|
||||
);
|
||||
let direct_http = MistralOcrConfig
|
||||
.prepare_request(&direct, &client, &crate::ocr::test_support::NoHooks)
|
||||
.await
|
||||
.unwrap();
|
||||
let vertex_http = VertexAiOcrConfig
|
||||
.prepare_request(&vertex, &client, &crate::ocr::test_support::NoHooks)
|
||||
.await
|
||||
.unwrap();
|
||||
assert_eq!(direct_http.url(), "https://mistral.test/v1/ocr");
|
||||
assert_eq!(
|
||||
vertex_http.url(),
|
||||
"https://vertex.test/v1/projects/project-1/locations/us-central1/publishers/mistralai/models/mistral-ocr-maas:rawPredict"
|
||||
);
|
||||
let http = if use_vertex {
|
||||
&vertex_http
|
||||
} else {
|
||||
&direct_http
|
||||
};
|
||||
assert_eq!(http.header("authorization").unwrap(), "Bearer test-key");
|
||||
assert_eq!(http.header("content-type").unwrap(), "application/json");
|
||||
assert_eq!(http.timeout(), Some(Duration::from_secs(2)));
|
||||
let body: Value = serde_json::from_slice(http.body()).unwrap();
|
||||
assert_eq!(
|
||||
body,
|
||||
json!({
|
||||
"model": "mistral-ocr-maas",
|
||||
"document": {"type": "document_url", "document_url": "data:application/pdf;base64,YWJj"},
|
||||
"pages": [0, 2],
|
||||
"include_image_base64": true,
|
||||
"unknown": "preserved"
|
||||
})
|
||||
);
|
||||
let payload = serde_json::to_vec(
|
||||
&json!({"pages": [{"index": 0, "markdown": "hello"}], "extra": "preserved"}),
|
||||
)
|
||||
.unwrap();
|
||||
let direct_response = MistralOcrConfig
|
||||
.transform_ocr_response(&direct.model, &payload, Default::default())
|
||||
.unwrap()
|
||||
.into_json();
|
||||
let vertex_response = VertexAiOcrConfig
|
||||
.transform_ocr_response(&vertex.model, &payload, Default::default())
|
||||
.unwrap()
|
||||
.into_json();
|
||||
assert_eq!(direct_response, vertex_response);
|
||||
assert_eq!(direct_response["model"], "mistral-ocr-maas");
|
||||
assert_eq!(direct_response["object"], "ocr");
|
||||
assert_eq!(direct_response["extra"], "preserved");
|
||||
}
|
||||
}
|
||||
|
|
@ -217,13 +217,25 @@ mod tests {
|
|||
"Authorization".to_string(),
|
||||
"Bearer abc".to_string()
|
||||
)]));
|
||||
assert!(has_bearer_auth(&[(
|
||||
"authorization".to_string(),
|
||||
"bearer abc".to_string()
|
||||
)]));
|
||||
assert!(!has_bearer_auth(&[(
|
||||
"Authorization".to_string(),
|
||||
"Bearer ".to_string()
|
||||
)]));
|
||||
assert!(!has_bearer_auth(&[(
|
||||
"authorization".to_string(),
|
||||
String::new()
|
||||
)]));
|
||||
assert!(!has_bearer_auth(&[(
|
||||
"Authorization".to_string(),
|
||||
"Basic abc".to_string()
|
||||
)]));
|
||||
assert!(!has_bearer_auth(&[(
|
||||
"x-api-key".to_string(),
|
||||
"abc".to_string()
|
||||
)]));
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -218,7 +218,3 @@ fn anthropic_body(
|
|||
);
|
||||
Value::Object(body)
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
#[path = "tests.rs"]
|
||||
mod tests;
|
||||
|
|
|
|||
|
|
@ -302,7 +302,3 @@ fn has_blank_text(message: &ChatMessage) -> bool {
|
|||
}),
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
#[path = "tests.rs"]
|
||||
mod tests;
|
||||
|
|
|
|||
|
|
@ -1,7 +1,11 @@
|
|||
use serde_json::json;
|
||||
|
||||
use super::*;
|
||||
use crate::base_llm::chat::transformation::Error;
|
||||
use litellm_llms::{
|
||||
anthropic::chat::transformation::ANTHROPIC_CHAT_COMPLETIONS_CONFIG,
|
||||
base_llm::chat::transformation::{
|
||||
BaseConfig, Error, ProviderChatResponseData, RequestAuth, Unsupported,
|
||||
},
|
||||
};
|
||||
use litellm_types::{llms::openai::ChatMessage, utils::ChatCompletionsResponse};
|
||||
use serde_json::{Map, Value, json};
|
||||
|
||||
fn messages(value: Value) -> Vec<ChatMessage> {
|
||||
serde_json::from_value(value).expect("valid messages")
|
||||
|
|
@ -205,7 +209,8 @@ fn declines_tool_calls_tool_results_and_multimodal_content() {
|
|||
);
|
||||
assert_eq!(
|
||||
reason(
|
||||
json!([{"role": "user", "content": [
|
||||
json!([
|
||||
{"role": "user", "content": [
|
||||
{"type": "image_url", "image_url": {"url": "https://x/y.png"}}
|
||||
]}]),
|
||||
json!({})
|
||||
|
|
@ -214,7 +219,8 @@ fn declines_tool_calls_tool_results_and_multimodal_content() {
|
|||
);
|
||||
assert_eq!(
|
||||
reason(
|
||||
json!([{"role": "user", "content": [
|
||||
json!([
|
||||
{"role": "user", "content": [
|
||||
{"type": "text", "text": "hi", "cache_control": {"type": "ephemeral"}}
|
||||
]}]),
|
||||
json!({})
|
||||
|
|
@ -1,7 +1,11 @@
|
|||
use serde_json::json;
|
||||
|
||||
use super::*;
|
||||
use crate::base_llm::chat::transformation::Error;
|
||||
use litellm_llms::{
|
||||
base_llm::chat::transformation::{
|
||||
BaseConfig, Error, ProviderChatResponseData, RequestAuth, Unsupported,
|
||||
},
|
||||
bedrock::chat::converse_transformation::BEDROCK_CHAT_COMPLETIONS_CONFIG,
|
||||
};
|
||||
use litellm_types::{llms::openai::ChatMessage, utils::ChatCompletionsResponse};
|
||||
use serde_json::{Map, Value, json};
|
||||
|
||||
fn messages(value: Value) -> Vec<ChatMessage> {
|
||||
serde_json::from_value(value).expect("valid messages")
|
||||
|
|
@ -56,7 +56,7 @@ pub struct ChatCompletionsChoice {
|
|||
///
|
||||
/// There is deliberately no `id`: Python mints the `chatcmpl-…` id on the
|
||||
/// `ModelResponse` it already created, and echoing the provider's own id here
|
||||
/// would change it. Pinned by `response_carries_no_id` in `tests.rs`.
|
||||
/// would change it. Pinned by `response_carries_no_id` in the Anthropic chat transformation tests.
|
||||
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
|
||||
pub struct ChatCompletionsResponse {
|
||||
pub created: u64,
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue