Skip to content

The fastest, litest AI Gateway. Rust core with Python SDK. Call 100+ LLM APIs in OpenAI (or native) format with cost tracking, guardrails, load balancing, and logging [Bedrock, Azure, OpenAI, Anthropic, OpenAI, VertexAI, vLLM, Nvidia NIM]

1.6 KB
Raw Blame

Run uv run scripts/generate_trace_types.py from the repository root to regenerate Python HTTP contracts. Run it with --check to verify the frozen Lens schemas and compare generated Python with committed files

Lens owns the Rust contracts. scripts/lens_assets/source.json pins the Lens repository revision and SHA-256 of each schema and development fixture. The gateway consumes these data files offline; it does not compile or maintain a second Rust tracing implementation

To update the pin, check out the intended Lens commit and run its export-traces-schema binary for the default, --requests and --responses modes, plus export-traces-clickhouse-schema, with the schema feature and locked dependencies. Import the consumed roots into the corresponding scripts/lens_assets/schemas group and copy fixtures from that same commit. Update the source revision and hashes together, then regenerate and run the gateway trace contract tests

The generator pins datamodel-code-generator through its inline dependency metadata. Neither it nor the schema exporter is a runtime dependency. The templates preserve frozen models, tuple conversion and bounded scalars. Request models retain each canonical schema's extra-field policy, including ignored unknown GET parameters and forbidden extra SQL body fields

The generated contracts live in litellm/tracing/generated. requests.py supplies gateway endpoint parameter models, responses.py supplies the SQL response, and queries.py validates the remote SQL envelope. Public request behavior changes need their own compatibility review