[bot] Instrument Pinecone Inference API (pinecone-sdk) (2,500 weekly downloads)

[braintrust-bot]

Library and missing instrumentation

The official Pinecone Rust client crate, pinecone-sdk (maintained at pinecone-io/pinecone-rust-client), exposes Pinecone's Inference API as a first-class execution surface: PineconeClient::embed(...) (generate embeddings from hosted models such as multilingual-e5-large) and reranking. These are model-execution calls (analogous in kind to the OpenAI/Anthropic completion calls this SDK already extracts usage from), not vector-store CRUD.

This repository currently has no extractor, usage-metrics parser, or span helper for Pinecone Inference responses. src/extractors.rs only implements extract_openai_usage and extract_anthropic_usage; there is no Pinecone-shaped equivalent, and Cargo.toml has no Pinecone dependency, confirming zero existing support.

Weekly downloads: 2,500 (as of 2026-09-07; computed by summing the official crates.io per-day download API for 2026-08-31 through 2026-09-06 — source: https://crates.io/api/v1/crates/pinecone-sdk/downloads)

Braintrust docs status: not_found. Braintrust's existing Pinecone-related documentation (RAG cookbook, vector database article, integrations index) only covers using Pinecone as a vector store for externally-generated (e.g. OpenAI) embeddings. None of it documents instrumenting Pinecone's own Inference API (embed/rerank) execution calls, and no braintrust SDK appears to wrap or extract usage from Pinecone Inference responses.

Upstream sources

• https://github.com/pinecone-io/pinecone-rust-client (official crate source)
• https://docs.rs/pinecone-sdk (API docs — embed/rerank on PineconeClient)
• https://crates.io/crates/pinecone-sdk
• https://docs.pinecone.io/guides/inference/understanding-inference (Inference API docs)
• https://www.pinecone.io/blog/integrated-inference/ (feature announcement)

Braintrust docs sources checked

• https://www.braintrust.dev/docs/cookbook/recipes/ToolRAG
• https://www.braintrust.dev/docs/integrations
• https://www.braintrust.dev/articles/best-vector-databases-for-rag-2026

Local repo files inspected

• src/extractors.rs (only extract_openai_usage, extract_anthropic_usage)
• src/lib.rs (crate exports)
• Cargo.toml (dependency list — no Pinecone dependency)
• src/stream.rs (generic streaming helper, provider-agnostic)
• README.md