Summary
The Braintrust Go SDK instruments OpenAI, Anthropic, Google GenAI, sashabaranov/go-openai, LangChainGo, Firebase Genkit, Google ADK, CloudWeGo Eino, and AWS Bedrock — but does not instrument the Mistral AI Go SDK (github.com/mistralai/client-go). Mistral is explicitly listed as a supported provider in Braintrust's tracing documentation, yet it has no SDK-level instrumentation wrapper in this repository.
What is missing
A trace/contrib/mistral/ integration module that wraps execution calls on the Mistral AI Go client. The key execution surfaces are:
Chat / ChatStream — conversational model inference (non-streaming and streaming), including multi-turn conversations and tool calling
Embed — vector embedding generation
The Mistral Go SDK is generated with the Stainless SDK toolchain (same as openai-go and anthropic-sdk-go), and exposes option.WithMiddleware() for HTTP middleware injection, making it directly instrumentable with the same internal.Middleware(router, logger) pattern already used by the existing OpenAI and Anthropic integrations.
Braintrust docs status
supported — Braintrust explicitly lists Mistral as a supported provider for tracing:
"OpenAI, Anthropic, Gemini, AWS Bedrock, Azure, Mistral, Together, Groq, and many more"
Source: https://www.braintrust.dev/docs/guides/tracing and https://www.braintrust.dev/docs/instrument
However, there is no Go SDK-level tracing wrapper for the Mistral AI client in this repository.
Upstream sources
Braintrust docs sources
Local repo files inspected
go.mod — no mistralai/client-go dependency
trace/contrib/ — no mistral/ directory exists
trace/contrib/all/all.go — no Mistral import
examples/internal/ — no Mistral example
- Grep for
mistral across all repo files — zero matches
trace/contrib/openai/traceopenai.go — reference pattern: internal.Middleware(router, logger) with option.WithMiddleware() injection; same pattern applies to client-go since both are Stainless-generated SDKs
trace/contrib/anthropic/traceanthropic.go — reference pattern for Stainless-generated SDK middleware instrumentation
Summary
The Braintrust Go SDK instruments OpenAI, Anthropic, Google GenAI, sashabaranov/go-openai, LangChainGo, Firebase Genkit, Google ADK, CloudWeGo Eino, and AWS Bedrock — but does not instrument the Mistral AI Go SDK (
github.com/mistralai/client-go). Mistral is explicitly listed as a supported provider in Braintrust's tracing documentation, yet it has no SDK-level instrumentation wrapper in this repository.What is missing
A
trace/contrib/mistral/integration module that wraps execution calls on the Mistral AI Go client. The key execution surfaces are:Chat/ChatStream— conversational model inference (non-streaming and streaming), including multi-turn conversations and tool callingEmbed— vector embedding generationThe Mistral Go SDK is generated with the Stainless SDK toolchain (same as
openai-goandanthropic-sdk-go), and exposesoption.WithMiddleware()for HTTP middleware injection, making it directly instrumentable with the sameinternal.Middleware(router, logger)pattern already used by the existing OpenAI and Anthropic integrations.Braintrust docs status
supported — Braintrust explicitly lists Mistral as a supported provider for tracing:
Source: https://www.braintrust.dev/docs/guides/tracing and https://www.braintrust.dev/docs/instrument
However, there is no Go SDK-level tracing wrapper for the Mistral AI client in this repository.
Upstream sources
Braintrust docs sources
Local repo files inspected
go.mod— nomistralai/client-godependencytrace/contrib/— nomistral/directory existstrace/contrib/all/all.go— no Mistral importexamples/internal/— no Mistral examplemistralacross all repo files — zero matchestrace/contrib/openai/traceopenai.go— reference pattern:internal.Middleware(router, logger)withoption.WithMiddleware()injection; same pattern applies toclient-gosince both are Stainless-generated SDKstrace/contrib/anthropic/traceanthropic.go— reference pattern for Stainless-generated SDK middleware instrumentation