fabro/Cargo.toml
Ryan Neal d5b2220ed3
feat(llm): Amazon Bedrock provider — Converse codec, SigV4 + API-key auth (#459)
Adds **Amazon Bedrock** as an opt-in built-in provider, over Bedrock's
unified **Converse / ConverseStream** API. One codec serves every
Converse-capable family — Claude, Amazon Nova, Meta Llama, Mistral,
DeepSeek, Moonshot Kimi, Z.AI GLM, MiniMax, NVIDIA Nemotron, and OpenAI
gpt-oss — because AWS translates the envelope to each model's native
dialect server-side. Auth is either **AWS SigV4** (the default
credential chain — env / profile / IMDS / IRSA / SSO, resolved per
request so sessions refresh) or a **Bedrock API key**
(`AWS_BEARER_TOKEN_BEDROCK`, bearer). Disabled by default (the Ollama /
OpenRouter opt-in pattern).

This is the redo of #459's original Claude-only `InvokeModel` adapter,
rebuilt on the gateway-refactor seams (#481–#497). @depopry's SigV4
signer, AWS event-stream frame decoder, `BedrockAuth`, the `aws_sigv4`
credential grammar, `AdapterKind::Bedrock`, region-from-base_url, and
the lean-deps decision are preserved and authored by him on the first
two commits; the per-family `BedrockCodec` trait he wrote turned out to
be the crate-wide `Codec` seam in miniature, so the refactor promoted
exactly that shape. The original Claude-only description is preserved in
a comment below.

## What's here

- **`AdapterKind::Bedrock` × `CodecKind::BedrockConverse`** on the
route, plus the `aws_sigv4` credential source (no static secret — the
adapter signs at request time; `fabro-auth` stays AWS-free).
*(@depopry)*
- **SigV4 signer + AWS event-stream `FrameDecoder`** on the lean AWS
stack (no `aws-sdk-bedrockruntime`; transport stays on `fabro-http`).
Re-targeted at Converse's direct-JSON stream frames; the signer resolves
credentials per request. *(@depopry)*
- **`bedrock_converse` codec** — Converse envelope (`system[]`, typed
content blocks, `inferenceConfig`, `toolConfig`), prompt caching via
`cachePoint`, thinking-signature round-trip through `reasoningContent`,
usage mapped onto the disjoint `TokenCounts` buckets,
`provider_options.bedrock` passthrough. Plus the adapter shell and an
event-stream byte loop beside the transport's shared SSE loop.
- **Catalog**: `bedrock.toml` (Claude incl. Fable 5, Nova 2, Llama 4,
Mistral, DeepSeek, Kimi, GLM, MiniMax, Nemotron, gpt-oss — cross-region
inference-profile ids, per-model `billing_policy` so Claude bills
Anthropic-style) and a companion **`bedrock-openai`** provider for
GPT-5.5/5.4 over the `bedrock-mantle` Responses endpoint (pure config
over the existing `openai_responses` codec, zero new code).
- Secrets registry (`AWS_BEARER_TOKEN_BEDROCK`), gitleaks rules for both
Bedrock key formats, the `docs/integrations/bedrock` guide, and live e2e
tests.

## Live verification (confirmed end-to-end against a real AWS account)

Verified on a real Bedrock account (us-east-2, SigV4 + bearer):

- **SigV4 + Converse** — multiple families (Claude, Nova, DeepSeek, …)
via the full settings → catalog → route → adapter → codec path.
- **ConverseStream** — streaming deltas through the workflow engine.
- **Multi-turn tool use** — agent loop with tool calls round-tripping
(no-arg tools included).
- **Multi-model routing** — Claude + DeepSeek pinned in one run through
the single Converse codec.
- **mantle Responses** — `openai.gpt-5.5` answered via the
`bedrock-openai` provider (bearer auth).

The exercise caught and fixed several issues that unit tests (static
creds, mocked transports) could not — see the follow-up commits below.

## Follow-up fixes from live testing (commits on top of the foundation)

1. **Worker AWS env** — the workflow worker scrubs its env to an
allowlist, so SigV4 (which re-resolves from the ambient chain per
request) couldn't work through `fabro run`. The AWS credential-chain
inputs now cross into the worker.
2. **Vault bearer key** — Bedrock was the only key-based provider
missing a `vault:` credential ref, so `fabro secret set
AWS_BEARER_TOKEN_BEDROCK` silently didn't feed it. Now resolves env →
vault → SigV4.
3. **Converse tool-encoding hardening** — a no-arg tool call's
`toolUse.input` is now a `{}` object (Bedrock rejects null), and every
tool `inputSchema` gets a top-level `type: "object"` (strict families
like DeepSeek reject a typeless schema Claude tolerates).
4. **Nova output cap** — `amazon.nova-2-lite` max_output 65536 → 65535
(Bedrock's per-request limit).

Earlier fixes already folded into the foundation commits: the
`aws-config` sleep-impl (default chain panicked) and AWS error-body
decoding (top-level `message`/`Message`/`__type` → proper messages
instead of "Unknown error").

## Manual testing & setup

See `docs/integrations/bedrock` — now documents the non-obvious account
setup that live testing surfaced: the per-Region Anthropic use-case
approval, `aws-marketplace:Subscribe` for third-party models, the Fable
5 / Mythos-class data-sharing opt-in, and the bearer-vs-SigV4 precedence
override for running Converse + mantle side by side.

## Open decision / discussion

- **Model-id naming** — Bedrock rows use dotted ids mirroring Bedrock's
native inference-profile ids (`us.anthropic.claude-sonnet-4-6`,
`openai.gpt-5.5`), which also makes them the wire `api_id`. Third scheme
alongside bare ids and OpenRouter's `vendor/model` slashes. No collision
risk (enforced at catalog build). Open to a uniform scheme if preferred.
- **`BEDROCK_API_KEY` alias** — see the comment thread; the AWS console
hands some users `export BEDROCK_API_KEY=` while the SDK-standard var is
`AWS_BEARER_TOKEN_BEDROCK`. Question of whether to accept both.

## Deferred (named follow-ups)

- **`qwen.qwen3-coder-next`** — omitted pending a verified Bedrock
model/inference-profile id (its fabro id isn't a valid Bedrock
identifier; needs an explicit `api_id`). Re-add once confirmed via `aws
bedrock list-inference-profiles`.
- **Claude Mythos 5** — Anthropic-Messages-only on `bedrock-mantle`
(limited preview).
- **Converse structured output** (`response_format` rejected with a
clear error).
- **`reasoning_effort` on Converse rows** via
`additionalModelRequestFields` (the `bedrock-openai` GPT rows already
accept effort levels).
- **CountTokens** route (`count_input_tokens` returns `None`).

## Verification

`cargo nextest run --workspace`: green except the pre-existing
environment-dependent fabro-workflow failures (identical on main).
clippy `-D warnings` + pinned-nightly fmt clean. Codec unit tests +
adapter httpmock tests + frame-decoder/signer locks.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Scott Werner <scott@sublayer.com>
Co-authored-by: Scott Werner <stwerner@vt.edu>
2026-06-16 11:46:49 -04:00

172 lines
5.7 KiB
TOML

[workspace]
members = ["lib/crates/*", "test/twin/openai", "test/twin/github"]
default-members = ["lib/crates/fabro-cli"]
resolver = "2"
[workspace.package]
edition = "2021"
version = "0.260.0-nightly.0"
license = "MIT"
[workspace.dependencies]
agent-client-protocol = { version = "0.11.1", features = ["unstable_session_usage"] }
agent-client-protocol-tokio = "0.11.1"
anyhow = "1"
axum = { version = "0.8" }
axum-extra = { version = "0.10", features = ["cookie-private", "query"] }
cookie = { version = "0.18", features = ["percent-encode", "private", "signed", "key-expansion"] }
croner = "3.0.1"
thiserror = "2"
serde = { version = "1", features = ["derive"] }
serde_json = { version = "1", features = ["preserve_order"] }
tokio = { version = "1", features = ["full"] }
reqwest = { version = "0.13", default-features = false, features = ["json", "stream", "rustls", "query", "form", "multipart"] }
sse-stream = "0.2"
ulid = "1"
uuid = { version = "1", features = ["v4", "v7", "v8"] }
rand = "0.9"
dotenvy = "0.15"
futures = "0.3"
tokio-stream = "0.1"
async-trait = "0.1"
fs2 = "0.4"
base64 = "0.22"
bytes = "1"
tokio-util = "0.7"
# AWS building blocks for the native Bedrock adapter. Lean stack: request
# signing + credential chain + event-stream decode only. Transport for the
# actual Bedrock inference calls stays on fabro-http; the full
# aws-sdk-bedrockruntime (and its parallel hyper stack) is not pulled in.
# aws-config keeps its DEFAULT features on purpose: `rt-tokio` supplies the
# TokioSleep impl the credential chain's retry requires (without it,
# resolving the default chain panics with "an async sleep implementation is
# required"), and `sso`/`credentials-process` make from_default_chain's
# documented SSO/credential-process support real. `rustls` pins the TLS
# backend for credential-resolution HTTP.
aws-config = { version = "1", features = ["behavior-version-latest", "rustls"] }
aws-credential-types = { version = "1", features = ["hardcoded-credentials"] }
aws-sigv4 = "1"
aws-smithy-eventstream = "0.60"
aws-smithy-runtime-api = "1"
aws-smithy-types = "1"
clap = { version = "4", features = ["derive", "env"] }
clap_complete = "4"
jsonschema = { version = "0.42", default-features = false }
chrono = { version = "0.4", features = ["clock", "serde"] }
dashmap = "6"
bollard = "0.18"
tar = "0.4"
cli-table = { version = "0.5", default-features = false }
console = "0.15"
dialoguer = "0.12"
git2 = { version = "0.20", default-features = false, features = ["vendored-libgit2", "vendored-openssl", "https"] }
tracing = "0.1"
tracing-subscriber = { version = "0.3", features = ["fmt", "env-filter"] }
tracing-appender = "0.2"
rmcp = { version = "1.4", default-features = false }
walkdir = "2"
regex = "1"
semver = "1"
aho-corasick = "1"
globset = "0.4"
dirs = "6"
mac_address = "1"
md5 = "0.7"
mime_guess = "2"
indicatif = "0.18"
termimad = "0.34"
toml = "0.8"
toml_edit = "0.22"
jsonwebtoken = { version = "10", features = ["aws_lc_rs"] }
hkdf = "0.12"
hmac = "0.12"
sha2 = "0.10"
hex = "0.4"
insta = "1"
fabro-test = { path = "lib/crates/fabro-test" }
twin-openai = { path = "test/twin/openai" }
twin-github = { path = "test/twin/github" }
tokio-tungstenite = { version = "0.26", features = ["rustls-tls-webpki-roots"] }
futures-util = "0.3"
daytona-sdk = { git = "https://github.com/brynary/daytona-sdk-rust", rev = "fc58e22f7f25183df6264276ee186bbc32635738", package = "daytona-sdk" }
daytona-api-client = { git = "https://github.com/brynary/daytona-sdk-rust", rev = "fc58e22f7f25183df6264276ee186bbc32635738", package = "daytona-api-client" }
sentry = { version = "0.35", default-features = false, features = ["backtrace", "contexts", "ureq", "rustls"] }
fork = "0.2"
exec = "0.3"
slatedb = { version = "0.11.2", features = ["zstd"] }
object_store = { version = "0.12.5", features = ["aws"] }
rust-embed = "8"
percent-encoding = "2"
minijinja = "=2.19.0"
miette = { version = "7.6", features = ["fancy"] }
fabro-http = { path = "lib/crates/fabro-http" }
fabro-environment = { path = "lib/crates/fabro-environment" }
fabro-options-metadata = { path = "lib/crates/fabro-options-metadata" }
fabro-redact = { path = "lib/crates/fabro-redact" }
fabro-static = { path = "lib/crates/fabro-static" }
graphviz-sys = { git = "https://github.com/fabro-sh/graphviz-sys" }
ref-cast = "1"
strum = { version = "0.28", features = ["derive"] }
url = "2"
zeroize = "1"
[workspace.lints.rust]
unsafe_code = "deny"
unreachable_pub = "warn"
[workspace.lints.clippy]
pedantic = { level = "warn", priority = -2 }
allow_attributes_without_reason = "warn"
# Allowed pedantic lints
implicit_hasher = "allow"
missing_errors_doc = "allow"
missing_panics_doc = "allow"
module_name_repetitions = "allow"
must_use_candidate = "allow"
similar_names = "allow"
struct_excessive_bools = "allow"
too_many_arguments = "allow"
too_many_lines = "allow"
cast_precision_loss = "allow"
doc_markdown = "allow"
# Disallowed restriction lints
print_stdout = "warn"
print_stderr = "warn"
dbg_macro = "warn"
empty_drop = "warn"
empty_structs_with_brackets = "warn"
disallowed_methods = "deny"
exit = "warn"
get_unwrap = "warn"
unwrap_used = "deny"
rc_buffer = "warn"
rc_mutex = "warn"
rest_pat_in_fully_bound_structs = "warn"
use_self = "warn"
# Project-specific lints
wildcard_imports = "warn"
absolute_paths = "warn"
[profile.release]
lto = "thin"
strip = true
[profile.dev]
debug = "line-tables-only"
split-debuginfo = "off"
[profile.test]
debug = "line-tables-only"
split-debuginfo = "off"
[profile.dev.package."*"]
debug = false # Disable debug info for all dependencies
opt-level = 1 # Shrinks monomorphized generics, reducing test binary size
# regex is extremely slow in debug builds (~10s to compile gitleaks patterns)
[profile.dev.package.regex]
opt-level = 2
[profile.dev.package.regex-automata]
opt-level = 2
[profile.dev.package.regex-syntax]
opt-level = 2