feat
minor
llm
feat/phase-2g-bundle
feat(llm): decouple chat vs embeddings, optional Anthropic-first, Voyage embeds
Document the two-lane model (embedding provider for Meaning search vs chat keys for completeChat) in setup, Hub LLM docs, deploy notes, and env examples.
Chat: refactor completeChat into OpenAI/Anthropic paths; add KNOWTATION_CHAT_PREFER_ANTHROPIC to try Claude before OpenAI when both keys exist (OpenAI fallback on Claude failure). Default order unchanged when unset.
Embeddings: add Voyage provider (VOYAGE_API_KEY), EMBEDDING_PROVIDER/MODEL env overrides in config, input_type query/document at call sites, dimension heuristics, and actionable failure messages including re-index note.
Tests: prefer-anthropic behavior, Voyage embeddingDimension, config env overrides, formatEmbeddingFetchFailure for Ollama/Voyage hints.
sha256:36e31c4d81bc89fc9dc7437213e87d8c23baf764175afdaf375915206757f637
sha
sha256:f34f50bdfad511b094fc134deaa22c0ab29b79432796334dd6701eca0a824570
snapshot
0 comments
To add a comment, use the Muse CLI:
muse hub commit comment sha256:36e31c4d81bc89fc9dc7437213e87d8c23baf764175afdaf375915206757f637 --body "your comment"
No comments yet. Be the first to start the discussion.