[bot] Add Cerebras Cloud SDK integration for chat completions and streaming instrumentation

[braintrust-bot]

Summary

The Cerebras Cloud SDK (cerebras-cloud-sdk) is the official Python client for Cerebras Inference, which provides some of the world's fastest LLM inference using Cerebras' CS-3 wafer-scale processors. Its Cerebras and AsyncCerebras clients expose chat.completions.create() with full streaming support. This repository has zero instrumentation for any Cerebras SDK surface — no integration directory, no wrapper, no patcher, no auto_instrument() support.

While Cerebras exposes an OpenAI-compatible endpoint, the native cerebras.cloud.sdk.Cerebras client is not a subclass of openai.OpenAI, so wrap_openai() cannot be used with it. The pattern is identical to Groq and Mistral — both have OpenAI-compatible APIs but received dedicated native integrations in this repo. Users who follow Cerebras' official documentation and pip install cerebras-cloud-sdk get zero Braintrust tracing.

What needs to be instrumented

The cerebras-cloud-sdk package exposes these execution surfaces via Cerebras and AsyncCerebras, none of which are instrumented:

Chat completions (highest priority)

SDK Method	Description	Streaming	Return type
client.chat.completions.create()	Chat completions with tool calling, JSON mode, structured output	stream=True returns iterator of ChatCompletionChunk	ChatCompletion
client.chat.completions.stream()	Context-manager streaming variant	Yields ChatCompletionStreamEvent	ChatCompletionStreamManager

Response shape is OpenAI-compatible: ChatCompletion has choices, usage (prompt_tokens, completion_tokens, total_tokens), model, id, time_info (wall and queue times) — standard span metrics extraction follows the OpenAI pattern, with the bonus of Cerebras-specific latency fields.

Streaming: stream=True on create() or the stream() context manager yields ChatCompletionChunk objects. The integration must handle the streaming span lifecycle (start on call, accumulate chunks, finalize on stream exhaustion).

Both sync (Cerebras) and async (AsyncCerebras) clients exist with identical method signatures.

Implementation notes

Stainless-generated SDK: The cerebras-cloud-sdk package is generated by Stainless (same toolchain as OpenAI, Groq, Mistral). The client structure (client.chat.completions.create()) mirrors the OpenAI SDK, so the patcher/wrapper pattern from the OpenAI or Mistral integration can serve as a close structural reference.

chat.completions.create() parameters relevant for span metadata: model, temperature, max_tokens, max_completion_tokens, top_p, frequency_penalty, presence_penalty, seed, tools, tool_choice, response_format, stop, n.

Extra timing fields: Cerebras responses include time_info with queue_time and total_time (floats, in seconds). These are unique to Cerebras and useful for performance monitoring spans.

No coverage in any instrumentation layer

• No integration directory (py/src/braintrust/integrations/cerebras/)
• No wrapper function (e.g. wrap_cerebras())
• No patcher in any existing integration
• No nox test session (test_cerebras)
• No version entry in py/src/braintrust/integrations/versioning.py
• No mention in py/src/braintrust/integrations/__init__.py
• No entry in [tool.braintrust.matrix] in py/pyproject.toml

A grep for cerebras across py/src/braintrust/ returns zero matches.

Braintrust docs status

not_found — Cerebras is not listed on the Braintrust integrations directory or the tracing guide.

Upstream references

• Cerebras Cloud SDK on PyPI: https://pypi.org/project/cerebras-cloud-sdk/
• Cerebras Cloud SDK on GitHub: https://github.com/Cerebras/cerebras-cloud-sdk-python
• Cerebras Inference docs: https://inference-docs.cerebras.ai/
• Cerebras API reference: https://inference-docs.cerebras.ai/api-reference/
• Cerebras Chat Completions guide: https://inference-docs.cerebras.ai/api-reference/chat-completions

Local repo files inspected

• py/src/braintrust/integrations/ — no cerebras/ directory exists on main
• py/src/braintrust/wrappers/ — no Cerebras wrapper
• py/noxfile.py — no test_cerebras session
• py/src/braintrust/integrations/__init__.py — Cerebras not listed in integration registry
• py/src/braintrust/integrations/versioning.py — no Cerebras version matrix
• py/pyproject.toml — no Cerebras entries in [tool.braintrust.matrix]
• Full repo grep for "cerebras" across py/src/braintrust/ — zero matches