feat(openrouter): surface cost and cost_details in response_metadata

[Hamza Kyamanywa (untilhamza)]

Description

OpenRouter returns cost and cost_details in its API response usage object, providing the actual cost of each API call. Currently, _create_usage_metadata() only extracts token counts and drops these cost fields.

This PR surfaces both cost and cost_details in response_metadata for both non-streaming and streaming paths, allowing users to access actual API costs directly from the response without manual estimation from token counts.

Example response from OpenRouter:

{
  "usage": {
    "prompt_tokens": 100,
    "completion_tokens": 50,
    "cost": 0.000075,
    "cost_details": {
      "upstream_inference_cost": 0.00007745,
      "upstream_inference_prompt_cost": 0.00000895,
      "upstream_inference_completions_cost": 0.0000685
    }
  }
}

After this change:

result = chat.invoke("hello")
result.response_metadata["cost"]          # 0.000075
result.response_metadata["cost_details"]  # {...}

Changes

• _create_chat_result: Surface cost and cost_details from token_usage into response_metadata (non-streaming)
• _convert_chunk_to_message_chunk: Same for streaming AIMessageChunk
• Added PLR0912 to noqa comments (new branches pushed count over threshold)
• Added two unit tests: one verifying cost fields are present when returned, one verifying they're absent when not in usage

Issue

N/A — discovered while integrating OpenRouter in a production pipeline. The cost data is already returned by the API but was being silently dropped.

Dependencies

None.

Twitter handle

@hamza_kyamanywa

[Mason Daugherty (mdrxy)]

I do see one blocker in the streaming path:

1. In _stream and _astream, final-chunk metadata currently replaces message_chunk.response_metadata:
   • message_chunk = message_chunk.model_copy(update={"response_metadata": generation_info})
2. Since cost fields are added earlier in _convert_chunk_to_message_chunk, they can be dropped if finish_reason metadata is set on the same chunk.

Could you merge metadata instead of replacing it (e.g. merge existing message_chunk.response_metadata with generation_info)?

Also, please add streaming regression tests (stream + astream) asserting cost/cost_details survive on the final chunk when usage includes them.

[Hamza Kyamanywa (untilhamza)]

Mason Daugherty (@mdrxy) Good catch — thanks for the review!

Fixed in c79c26b:

1. Metadata merge: Both _stream and _astream now merge existing response_metadata with generation_info instead of replacing it:

   update={"response_metadata": {**message_chunk.response_metadata, **generation_info}}

2. Regression tests: Added test_stream_cost_survives_final_chunk and test_astream_cost_survives_final_chunk — both assert that cost/cost_details survive alongside finish_reason on the final chunk.

All 194 unit tests pass.