Back to the room

Pass 3 of 3 · Deep dive of PHAROSPass 1 · OverviewPass 2 · How it works

Code Observatory / Recorded run

Hints turn auto into hybrid

The same question with two guessed identifiers. Auto becomes hybrid, and the function that actually does the check moves from fifth to first.

RECORDED Captured from a real run on November 2, 2026. Nothing is running now.

PHAROS at 975a38dMIT license

The same question, two rankings

Hybrid fused 50 keyword candidates with 50 dense ones. git_blob_sha ranked first for keyword and gained the exact-name boost.

RankAuto, no hints (dense)Auto, with hints (hybrid)Keyword rankDense rankScore
1codeintel.git_source.Checkout.read_blobcodeintel.source.git_blob_sha exact name150.048172
2codeintel.git_sourcecodeintel.git_source320.032002
3learning.memory.file_evidencecodeintel.git_source.Checkout.read_blob610.031545
4codeintel.git_source.read_file_at_commitcodeintel.source2110.030214
5codeintel.source.git_blob_shacodeintel.git_source.Checkout._git5320.026254

Recomputed for this page: 5 of 5 scores match their recorded ranks (1/(60+rank) per list, exact-name boost 1/61, path penalty ×0.7).

How this was recorded

Transport
REST POST /api/v1/search/code
Index
Indexed from GitHub at this commit through the production path: PostgreSQL, pgvector HNSW and ONNX embeddings.
Database
PostgreSQL 17.11 with pgvector 0.8.7
Embedding model
granite-embedding-english-r2-int8
Where
Local Mac, loopback only; not the deployed service

Claims this supports

  • PHAROS supports keyword, dense, hybrid and graph retrieval; auto prefers dense, hybrid with hints, keyword fallback with warnings.

    Source verifiedLedger

    Explicit unavailable semantic modes fail; auto can degrade with a warning. Repository scope is a filter the caller chooses; a call naming no repository searches the whole index.