Repository navigation
Round 7: unify model settings and detect embedding dimensions - #130
Merged
Merged
Conversation
…ection in Compose
…dentity P3-03 lets a store with no vector artifact adopt a new embedding model at startup, but the backfill only moved SKIPPED lanes to the new identity. A record admitted under the old model whose VECTOR lane was still pending, retrying or dead-lettered failed every attempt with "embedding identity mismatch", went DEAD_LETTER and never became searchable (before P3-03 the restart was refused instead). The startup backfill now aligns every unfinished VECTOR lane of a mutation without vector artifacts; mutations with vectors are still never rewritten. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Round 7 configures extraction and embeddings with MESA_LLM_MODEL and MESA_EMBEDDING_MODEL provider:model values. Removed settings fail startup with migration guidance; Ollama embedding supports automatic dimension detection and preserves the egress fence. Vector-free stores adopt the configured embedding identity, including pending projection lanes, while stores with vectors still require a rebuild. Documentation now describes model-optional writes and BM25 search.
Validation at 9a155e1: final Round 7 checker verdict READY; fresh R7-premerge full gate PASS (lane, lock, sync, lint, mypy, layer imports, 1623 local tests and 7 MCP tests). No additional code changes were required during the pre-merge review.