`keelgate.adapters.langgraph`¶

Stability of each symbol (stable or provisional) is listed in the API contract. Narrative and semantics are in the integration contract.

LangGraph adapter: governed tools, a chat-model wrapper, and the durable checkpoint store.

keelgate[langgraph] provides the SDK. The loop's default durable checkpointer lives in keelgate.loop and is re-exported here for convenience.

KeelgateChatModel ¶

Bases: BaseChatModel

A LangChain chat model whose completions come from a Keelgate LLMClient.

LangGraphCheckpointStore ¶

Loop checkpoints on a LangGraph saver.

The stale-writer guard (save refuses a sequence that is not newer) is a read followed by a write, which LangGraph's saver API cannot make atomic by itself. Two locks make it atomic: a thread lock for threads of one process, and an optional process_lock for other processes. :meth:sqlite on a real file supplies one (a separate <path>.lock SQLite file held under BEGIN IMMEDIATE). With a saver you build yourself (for example Postgres) and no process_lock, the guard is thread-safe only; pass a lock that suits your backend, or use :class:~keelgate.loop._checkpoint.SqliteCheckpointStore.

history ¶

history(tenant_id: str, run_id: str) -> list[LoopState]

Every checkpoint of a run, oldest first.

governed_langchain_tools ¶

governed_langchain_tools(
    toolset: GovernedToolset,
) -> list[StructuredTool]

One LangChain tool per governed tool. Each call goes through the gateway.

The coroutine never raises for a refusal: a denial, a parked approval or a validation error comes back as the rendered result text, so the agent can read it and adapt, and the framework's own error handling never sees (or retries) a policy decision.

to_keelgate_messages ¶

to_keelgate_messages(
    messages: list[BaseMessage],
) -> tuple[Message, ...]

Convert LangChain messages into Keelgate Message objects.

keelgate.adapters.langgraph¶