gbrain

admin/gbrain

Fork 0

mirror of https://github.com/garrytan/gbrain.git synced 2026-07-28 14:59:47 +00:00

Files

T

History

cf2deedfc6 fix(embed): label content_chunks.model with the model that produced the vector (#1717 )

The embed paths (embedPage, embedAll, embedAllStale) and the inline
import/sync embed paths built ChunkInput[] without a model field, so the
engines' upsertChunks defaulted content_chunks.model to the hardcoded
DEFAULT_EMBEDDING_MODEL instead of the gateway-configured model that
actually produced the vector.

- New core helper resolveEmbeddingModelLabel() in src/core/embedding.ts
  (returns the resolved gateway model, undefined when unconfigured).
- embed.ts: stamp the label on (re)embedded chunks in all three paths;
  chunks preserved from a prior embed keep their existing model so a
  mixed-model page isn't relabeled wholesale.
- import-file.ts: stamp the label on inline-embedded markdown chunks and
  re-embedded code chunks; reused (incremental) code-chunk embeddings
  carry their existing model label forward.

Takeover of PR #1803 (rebased onto master over the pace-mode changes;
helper moved into core so import-file.ts can share it).

Co-authored-by: harjothkhara <harjothkhara@users.noreply.github.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

2026-07-21 14:42:17 -07:00

assets/wasm

v0.40.9.0 feat(chunker): .sql indexing via tree-sitter + code-def on SQL DDL (#1173 ) (#1350 )

2026-05-24 09:30:57 -07:00

commands

fix(embed): label content_chunks.model with the model that produced the vector (#1717 )

2026-07-21 14:42:17 -07:00

core

fix(embed): label content_chunks.model with the model that produced the vector (#1717 )

2026-07-21 14:42:17 -07:00

eval

v0.42.56.0 feat(chronicle): Life Chronicle — temporal timeline + thought diary + bi-temporal per-entity ontology (#2390 ) (#2533 )