Skip to content

v0.8.1 — opt-in OpenAI-compatible remote embedder

Choose a tag to compare

View all tags
@TheYonk TheYonk released this 30 May 22:29
· 30 commits to main since this release
v0.8.1
7d301a6
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

0.8.1 — 2026-05-30

Added

  • Opt-in remote embedder (type: openai). A second embedder type that calls any OpenAI-compatible /v1/embeddings endpoint instead of running locally — covers OpenAI, Azure, Voyage (Anthropic's recommended provider), Mistral, Together, and local servers (TEI / vLLM / Ollama) via base_url + model + optional api_key_env (bearer token read from an env var; keyless for local). fastembed remains the default — select the remote one per cell with embedder.type: openai. Uses stdlib HTTP only — no new runtime dependency. Batches by batch_size, reorders the response by index, retries on 429/5xx with exponential backoff, and validates the returned dim against config. Anthropic has no embeddings API (its ecosystem answer is Voyage); Cohere's distinct /v1/embed shape is not supported. See docs/reference/embedder-openai.md + docs/samples/sample-openai-embedder.yaml.

Docs: docs/reference/embedder-openai.md · sample: docs/samples/sample-openai-embedder.yaml

Assets 2
Loading