feat(embeddings): record the embedding contract and refuse a mismatched config (#189)

Stored vectors are only searchable with the backend, model, width and
prefixes that built them, and nothing enforced that: changing [embedding]
(or backend) without reembed_embeddings silently broke similarity search,
and the per-row dim check missed same-width model or prefix swaps.

- embedding_contract: one row (backend, model, dim, both prefixes). The
  server checks it once after migrations; the src/bin tools check it on
  their first embed. No row yet = record the running config.
- On a mismatch every embed (embed_many, so queries and documents) fails
  with a message naming each changed field and the fix. Mixing vector
  spaces corrupts the corpus for good; refusing only pauses semantic
  search until the config is put back or the corpus re-embedded.
- reembed_embeddings bypasses the guard and records the contract after a
  full run (all four tables, no --limit, not --dry-run).
- A DB error in the check leaves embedding allowed and is logged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Cameron CordesandClaude Opus 5.5 committed 2026-11-03 12:06:54 -05:00
1 parent 3ccce38505
commit 070f20b295
8 files changed
+374 -1

No files matched your search

@@ -0,0 +1,12 @@
-- The embedding contract the stored vectors were built under (#189): one
-- row, written on first start (or by reembed_embeddings), compared with the
-- resolved [embedding] config at startup.
CREATE TABLE embedding_contract (
id INTEGER PRIMARY KEY NOT NULL CHECK (id = 1),
backend TEXT NOT NULL,
model TEXT NOT NULL,
dim INTEGER NOT NULL,
query_prefix TEXT NOT NULL,
document_prefix TEXT NOT NULL,
recorded_at BIGINT NOT NULL
);