feat(groq): native content-block streaming events

[LennyMalcolm0]

Stacked on #124 (core BlockStreamTracker + Anthropic), which it branches from. Depends only on the core primitive — sibling of the OpenAI chain. Retargets to master once #124 lands.

Gives ChatGroq native _stream_chat_model_events / _astream_chat_model_events hooks so stream_events(version="v3") maps groq's stream directly to content-block events.

How it works

A bespoke converter (convert_groq_stream / async twin) builds text, reasoning, and tool-call blocks directly from groq's raw OpenAI-shaped delta, feeding the shared BlockStreamTracker. groq streams tool-call arguments incrementally (id+name on the first delta, fragments on later deltas at the same index), so the tracker accumulates them and finalizes to a parsed dict. Usage is read from x_groq.usage; the arguments == "null" quirk is normalized to "{}".

It deliberately does not depend on langchain-openai — it reuses only groq's own _create_usage_metadata.

The win over the bridge

groq's existing chunk parser puts tool_calls and reasoning in additional_kwargs, so the compat bridge surfaces neither as content blocks. The native converter surfaces both — tool calls as tool_call blocks and reasoning as reasoning blocks.

Worth careful review

• Tool-arg accumulation across deltas (keyed by tool:{index}); a test covers parallel tool calls at distinct indices.
• message_id is keyword-only on the hooks, kept out of the request kwargs; message-start carries core's LangChain run id.
• Scope cut: groq server-tool results (executed_tools) are not yet surfaced as blocks (noted in-code); the bridge still carried them in additional_kwargs.
• Sync/async are faithful twins.

---

Cloned from langchain-ai#38029 for DevAsign stress testing.

[devasign-test-app]

AI review: changes requested

The pull request successfully implements native content-block streaming events for Groq by adding sync and async hooks to ChatGroq. It introduces convert_groq_stream and its async twin to parse raw OpenAI-shaped deltas directly into text, reasoning, and tool-call blocks using the shared BlockStreamTracker. All acceptance criteria are fully met with comprehensive unit tests covering parallel tool calls and argument accumulation.

Acceptance criteria

Met

• ChatGroq implements both _stream_chat_model_events and _astream_chat_model_events hooks to natively support stream_events(version="v3").
• Implement convert_groq_stream and its asynchronous counterpart to build text, reasoning, and tool-call blocks directly from Groq's raw OpenAI-shaped delta, feeding the shared BlockStreamTracker.
• Accumulate tool-call arguments incrementally across deltas using index-based keying (e.g., tool:{index}) and finalize them to a parsed dictionary.
• Normalize the Groq quirk where tool arguments are returned as the string "null" to an empty JSON object string "{}".
• Extract usage information from x_groq.usage and reuse Groq's own _create_usage_metadata helper without introducing a dependency on langchain-openai.
• Ensure message_id is a keyword-only argument on both _stream_chat_model_events and _astream_chat_model_events and is excluded from the request kwargs sent to the Groq API.
• Ensure the message-start event carries the core LangChain run ID.
• Include unit tests verifying parallel tool calls at distinct indices and tool-argument accumulation across deltas.

Additional findings

Blockers

• Critical error (libs/partners/groq/langchain_groq/chat_models.py)
  In libs/partners/groq/langchain_groq/chat_models.py, the native streaming event hooks call self.client.create(...) (line 788) and self.async_client.create(...) (line 812). However, the Groq SDK client (groq.Groq / groq.AsyncGroq) does not expose a create method directly on the client instance; chat completions must be created via client.chat.completions.create. This will cause an immediate AttributeError and crash at runtime when native streaming events are requested.

Warnings

• Deferred / incomplete work (libs/partners/groq/langchain_groq/_stream_events.py)
  Incidental: In libs/partners/groq/langchain_groq/_stream_events.py, the code notes that "groq server-tool results (delta["executed_tools"]) are not surfaced as blocks yet; tracked for a follow-up."