chat_model_stream

Parse accumulated tool-chunk args into a finalized block.

Shared between the compat bridge's _finalize_block and the ChatModelStream end-of-stream sweep. Parses raw_args as JSON: on success builds the requested finalized type (tool_call or server_tool_call) with provider-specific fields (extras) preserved; on failure falls back to invalid_tool_call carrying the raw string so downstream consumers can still introspect the malformed payload.

Message from an AI.

An AIMessage is returned from a chat model as a response to a prompt.

This message represents the output of the model and consists of both the raw output as returned by the model and standardized fields (e.g., tool calls, usage metadata) added by the LangChain framework.

Sync iterable of deltas with pull-based backpressure.

Follows the same _request_more convention as langgraph's EventLog: when the cursor catches up to the buffer and the projection is not done, it calls _request_more() to pull more events from the producer.

Each call to __iter__ creates a new cursor at position 0. Multiple iterators replay all deltas from the start.

String-specialized sync projection.

Adds __str__, __bool__, __repr__ for ergonomic use with .text and .reasoning projections.

Async iterable of deltas that is also awaitable for the final value.

Uses an asyncio.Event to notify consumers of state changes. Each waiter — the awaitable (__await__) and each async iterator cursor — shares the event and re-checks its own condition on wake. The event is cleared before a waiter awaits, so stale "something happened" signals don't cause spin loops.

This is single-loop only — producers and consumers must share an event loop. If cross-thread wake is ever required, revert to a list-of-futures pattern with call_soon_threadsafe.

Synchronous per-message streaming object for a single LLM response.

Returned by BaseChatModel.stream_v2(). Content-block protocol events are fed into this object and accumulated into typed projections.

Projections (always return the same cached object):

• .text — iterable of str deltas; str() for full text
• .reasoning — same as .text for reasoning content
• .tool_calls — iterable of ToolCallChunk deltas; .get() returns list[ToolCall]
• .output — blocking property, returns assembled AIMessage

Usage info is available on .output.usage_metadata once the stream has finished.

Raw event iteration::

for event in stream:
    print(event)  # MessagesData dicts

Asynchronous per-message streaming object for a single LLM response.

Returned by BaseChatModel.astream_v2(). Content-block events are fed into this object by a background producer task.

Projections:

• .text — async iterable of text deltas; awaitable for full text
• .reasoning — async iterable of reasoning deltas; awaitable
• .tool_calls — async iterable of ToolCallChunk deltas; awaitable for list[ToolCall]
• .output — awaitable for assembled AIMessage

Usage info is available on .output.usage_metadata once the stream has finished.

The stream itself is awaitable (msg = await stream) and async-iterable (async for event in stream).