* Optimize _get_model_cost_key to avoid expensive scans
- Remove expensive O(n) scan fallback that was causing 42.87% CPU overhead
- Only scan when size mismatch detected (O(1) check)
- Add warning in docstring: Only O(1) lookup operations are acceptable
- Clean up comments to be more concise
- Keep stale entry rebuild for pop() case (only triggers when stale entry found)
This fixes the performance issue where the scan was being triggered on every
failed lookup, causing severe CPU overhead during router operations.
* Add code quality check to enforce O(1) operations in _get_model_cost_key
- Add check_get_model_cost_key_performance.py to statically analyze _get_model_cost_key
- Detects O(n) operations (loops, comprehensions, problematic function calls)
- Recursively checks called functions to find nested O(n) operations
- Allows conditional O(n) rebuilds in helper functions (_rebuild_model_cost_lowercase_map, _handle_stale_map_entry_rebuild, _handle_new_key_with_scan)
* Integrate _get_model_cost_key performance check into CI pipeline
- Add check_get_model_cost_key_performance.py to check_code_and_doc_quality job
- Ensures O(1) requirement is enforced in CI to prevent performance regressions
* Remove unused performance test and clean up utils.py
- Remove test_get_model_info_performance.py (no longer needed)
- Remove extra blank line in utils.py
* Document allowed helper functions and exception process in _get_model_cost_key
- Add documentation listing allowed helper functions with O(n) operations
- Explain why these are acceptable (conditionally called)
- Add instructions for adding new exceptions to check_get_model_cost_key_performance.py
* Fix docstring detection and type checker error in performance check
- Add proper docstring tracking to skip docstring content (fixes false positive for 'map' in docstring)
- Add None check for docstring_quote to fix type checker error
- Restore _handle_new_key_with_scan to allowed_helpers list
* Remove check_get_model_cost_key_performance from CI pipeline
- Temporarily remove the performance check from CI to avoid blocking builds
* Restore performance check and remove memory leak tests from CI
- Add back check_get_model_cost_key_performance.py to CI pipeline
- Remove memory_leak_tests job that was causing port conflicts
* Remove extra blank line in CI config
* do not fallback to token counter if disable_token_counter is enabled, and return errors instead
* add exceptions and exception utils to map the same as /v1/chat/completions
* use safe_json_loads
* feat(gemini): add opt-in support for responseJsonSchema
Add support for Gemini's native responseJsonSchema parameter which uses
standard JSON Schema format instead of OpenAPI-style responseSchema.
Benefits of responseJsonSchema (Gemini 2.0+ only):
- Standard JSON Schema format (lowercase types)
- Supports additionalProperties for stricter validation
- Better compatibility with Pydantic's model_json_schema()
- No propertyOrdering required
Usage:
```python
response_format={
"type": "json_schema",
"json_schema": {"schema": {...}},
"use_json_schema": True # opt-in
}
```
This is backwards compatible - existing code continues to use
responseSchema by default.
Closes#16340
* docs: add documentation for use_json_schema parameter
Document the new use_json_schema option for Gemini 2.0+ models
in the JSON Mode documentation.
* refactor(gemini): use responseJsonSchema by default for Gemini 2.0+
Remove opt-in flag `use_json_schema` and automatically detect model version:
- Gemini 2.0+: uses responseJsonSchema (standard JSON Schema, supports additionalProperties)
- Gemini 1.5: uses responseSchema (OpenAPI format, legacy)
This follows LiteLLM's philosophy of abstracting provider differences -
users write the same code regardless of model version.
When extract_and_raise_litellm_exception tries to raise a LiteLLM exception
from an error string, it was always passing the response parameter. However,
some exceptions like APIConnectionError don't accept this parameter, causing
a TypeError.
This fix tries to raise the exception with the response parameter first,
and falls back to raising without it if a TypeError occurs.
This fixes the error:
TypeError: APIConnectionError.__init__() got an unexpected keyword argument 'response'
Which was occurring when Gemini returned UNEXPECTED_TOOL_CALL finish reason
and LiteLLM tried to convert the error to an APIConnectionError.
Fixes: cascading error when Gemini uses thinking feature (__thought__ tool calls)
* Update CLAUDE.md with qwen3 tool_calls bug fix instructions (#18922)
* fix(ollama): set finish_reason to "tool_calls" when tool_calls present
When qwen3 models return tool_calls through Ollama, the finish_reason
was incorrectly left as "stop" instead of being set to "tool_calls".
This caused clients to miss the tool_calls in the response.
Added _get_finish_reason helper method following OpenAI provider's
pattern, and fixed both streaming and non-streaming response paths.
Fixes: https://github.com/BerriAI/litellm/issues/18922
* fix(ollama): pass tools directly without model capability check
The previous code tried to check model capability via get_model_info()
which made network calls to localhost:11434. When Ollama is remote,
this fails and falls back to JSON format, breaking tool calling.
Ollama 0.4+ supports native tool calling - let Ollama handle
model capability detection instead of LiteLLM.
Fixes#18922
* fix(ollama): transform tool_calls response to OpenAI format
Ollama returns tool_calls with arguments as dict, but OpenAI format
requires arguments to be a JSON string. Also ensures 'type': 'function'
field is present.
Completes the fix for #18922
* fix(ollama): set finish_reason to "tool_calls" when tool_calls present
Fixes#18922
Two issues addressed:
1. Remove broken model capability check
- get_model_info() fails when Ollama runs on remote server
- Broken fallback triggered JSON prompt injection
- Now passes tools directly - Ollama 0.4+ handles detection
2. Set finish_reason correctly
- Was hardcoded to "stop" even with tool_calls present
- Clients use this to know how to process the response
- Now returns "tool_calls" when tool_calls are in response
Both streaming and non-streaming responses are fixed.
Tests:
- All 14 existing Ollama tests pass
- Added 3 focused tests for the fixes
Adds 'servers' field to OpenAPI schema when server_root_path is set, ensuring correct Swagger UI execute path for reverse proxies and subpath deployments. Includes tests to verify correct server URL handling for various root path formats.
- Add `should_wrap_with_default_message` parameter to GuardrailRaisedException
- Update Generic Guardrail API to use clean error messages without wrapper
- When should_wrap_with_default_message=False, exception shows the original
blocked_reason directly (e.g., "pii detected") instead of verbose format
- Update test to verify GuardrailRaisedException is raised with clean message
