Your LangChain pipeline, now a REST API whether it likes it or not

What it does

LangCorn is a thin deployment wrapper that takes a LangChain chain—or any custom Python function—and exposes it as a FastAPI endpoint. You point the create_service helper at a module path like examples.ex1:chain and it serves a POST route, complete with auto-generated OpenAPI docs. It handles the boring part: request parsing, async invocation, and wiring up Uvicorn so you don’t have to.

The interesting bit

The project treats your chain like a pluggable endpoint factory. Pass several chains to the create_service helper and it spins up parallel routes for each one. It also lets callers override LLM parameters per request via HTTP headers and thread conversational memory through a JSON payload, which is a pragmatic (if slightly manual) way to handle state.

Key highlights

• Generates REST endpoints from LangChain chains or plain Python functions with minimal configuration.
• Built-in static token auth via the auth_token flag or constructor argument.
• Per-request LLM overrides (temperature, max tokens, model name) through custom HTTP headers.
• Conversational memory support by passing history and memory arrays in the request body.
• Available as both a standalone CLI and a programmable FastAPI factory.

Caveats

• Auth is limited to a single static token; there is no user management or OAuth.
• Memory management requires manual round-tripping of JSON arrays; the client must maintain and resend the full conversation state.

Verdict

Best for developers who need to slap a REST API onto a LangChain chain or custom function without boilerplate. If you need multi-user auth or client-managed conversation state feels like a burden, you will likely want something heavier.