Summary
The Braintrust Rust SDK has no instrumentation support for Rig (rig-core), the most widely adopted Rust framework for building LLM-powered applications. Rig is not instrumented by any other Braintrust SDK either (Python, JS, Go, Ruby, Java, .NET), making this a completely uninstrumented library across the Braintrust ecosystem.
What is missing
Rig provides execution-oriented API surfaces that could be instrumented to automatically capture spans:
Agent trait — high-level interface with .prompt() and .chat() async methods for LLM interactions
CompletionModel trait — common interface for completion requests across providers (OpenAI, Anthropic, Cohere, Gemini, Mistral, and 13+ others)
EmbeddingModel trait — consistent API for embedding operations
Tool macro/trait — tool calling during agent execution
- RAG workflows — via
VectorStoreIndex trait for retrieval-augmented generation
Wrapping these traits/methods would enable automatic tracing of Rig-based applications, capturing inputs, outputs, token usage, latency, and tool call data for any provider Rig supports.
Library significance
- 472,000+ total downloads on crates.io
- 94 dependent crates (68 direct)
- Latest version: 0.34.0 (actively maintained, 52 releases)
- 18+ provider integrations built-in (OpenAI, Anthropic, Cohere, Gemini, Mistral, DeepSeek, Perplexity, etc.)
- Production users including St Jude, Coral Protocol, and others
- Official documentation at https://docs.rig.rs/
Braintrust docs status
not_found — No mention of Rig in Braintrust documentation. Rig is not listed on the SDK integrations page or any other Braintrust docs page.
Upstream sources
Local files inspected
src/extractors.rs — only OpenAI and Anthropic usage extractors; no Rig-related code
src/stream.rs — stream aggregator for OpenAI Chat Completions format only
src/types.rs — core types; no Rig integration types
src/lib.rs — public API exports; no Rig references
Cargo.toml — no rig-core dependency
- Full codebase search for "rig", "rig-core", "CompletionModel", "EmbeddingModel" — zero results
Summary
The Braintrust Rust SDK has no instrumentation support for Rig (
rig-core), the most widely adopted Rust framework for building LLM-powered applications. Rig is not instrumented by any other Braintrust SDK either (Python, JS, Go, Ruby, Java, .NET), making this a completely uninstrumented library across the Braintrust ecosystem.What is missing
Rig provides execution-oriented API surfaces that could be instrumented to automatically capture spans:
Agenttrait — high-level interface with.prompt()and.chat()async methods for LLM interactionsCompletionModeltrait — common interface for completion requests across providers (OpenAI, Anthropic, Cohere, Gemini, Mistral, and 13+ others)EmbeddingModeltrait — consistent API for embedding operationsToolmacro/trait — tool calling during agent executionVectorStoreIndextrait for retrieval-augmented generationWrapping these traits/methods would enable automatic tracing of Rig-based applications, capturing inputs, outputs, token usage, latency, and tool call data for any provider Rig supports.
Library significance
Braintrust docs status
not_found — No mention of Rig in Braintrust documentation. Rig is not listed on the SDK integrations page or any other Braintrust docs page.
Upstream sources
Local files inspected
src/extractors.rs— only OpenAI and Anthropic usage extractors; no Rig-related codesrc/stream.rs— stream aggregator for OpenAI Chat Completions format onlysrc/types.rs— core types; no Rig integration typessrc/lib.rs— public API exports; no Rig referencesCargo.toml— norig-coredependency