searchi/docs/RETRIEVAL-ROUTING.md

Mixed query routing audit

Read-only implementation audit, 2026-10-13, source revision 5d9cfc4. These opening findings describe that pre-fix snapshot; they are not a retrieval-quality measurement or an implemented ranking change.

Pre-fix behavior at 5d9cfc4

HybridSearch::query_vector selects Google's CodeRetrieval task only when SearchOptions::code_only is true. All-files searches use the general task, including natural-language queries asking for an implementation. The bounded query-vector LRU distinguishes tasks, so vectors cannot be reused across them accidentally. See crates/searchi-service/src/lib.rs.

The audio preference in crates/searchi-service/src/routing.rs recognizes words such as speech, sound and music. Document and image cues veto that preference; code cues did not at this snapshot. If the preferred audio search returns any qualifying candidate, it becomes the exclusive semantic lane. Consequently a query such as "code that detects speech" can exclude semantic code candidates in All files. Lexical candidates still participate in hybrid fusion, and explicit Code scope bypasses this audio preference. This is a risk to evaluate, not proof that every such query fails.

Cost of adding lanes

crates/searchi-service/src/store.rs retains one eligibility-cache entry. Its key includes the generation, user filters, source type and code scope. Changing a filter replaces the entry and scans/hashes matching passage metadata again. Alternating general and code lanes would therefore perform repeated eligibility work even when query vectors and the ANN graph are warm. The existing empty-audio fallback can also alternate audio and general filters.

Incremental dirty-path updates apply when the cached filter matches and the revision is still within the dirty-history floor. A replacement cache design must preserve those freshness checks, generation invalidation, strict filtering before candidate limits, and the total allocation budget.

Next experiment

First complete the unchanged frozen mixed baseline described in MIXED-QUALITY.md. Keep its labels and source bytes unchanged. Then compare soft audio preference and task-specific rank fusion against that baseline, retaining exact filename/symbol priority. Different task scores must not be treated as calibrated probabilities or compared blindly.

Before introducing multiple lanes, measure a bounded multi-entry eligibility cache against a shared validated snapshot with derived code/origin views. Bound memory across all entries, rather than giving each the full chunk budget. Reuse the existing worker and prepared ANN graph.

Report All-files and Code scope separately: top-1, recall@5, hard negatives and no-answer behavior, plus query-vector cache misses, inference time, eligibility scans/hash calls, memory and warm p95. Include mutation/deletion regressions. The existing semantic.eligibility and semantic.eligibility_filtered_scan_and_hash traces expose eligibility cost. Use an independently reviewed held-out corpus for acceptance; the frozen diagnostic alone must not become the optimization target.

Narrow suppression fix and actual regression

After the frozen baseline confirmed the English speech-related code failure, explicit code/function/implementation cues now veto automatic audio preference, alongside the existing document/image cues. English and German cues are covered by the routing unit test. This changes candidate eligibility through routing; it does not add CodeRetrieval to All files or implement soft multi-lane fusion.

The ignored integration routing_google_cpu passed on Windows CPU in 38.40 s using the verified normal worker and Google model, unchanged frozen routing source bytes and the licensed real roaring-creature sound. The same remembered code query produced qualifying audio candidates through the explicit Audio diagnostic lane. Updated default semantic routing also returned the source implementation, proving the exclusive suppression is removed in this fixture. The test asserts candidate preservation, not top-1 superiority or broad quality. The full mixed baseline above remains the unchanged pre-fix measurement. The completed 0.1.4 full regression now confirms that this known failed query returns the implementation first. Code intents in All files improved from 10/14 to 11/14 top-1 and 13/14 to 14/14 recall@5, without scored regression elsewhere. Independent relevance and no-answer gates stay open.

With this fix and the verified-asset callback, the Windows CPU workspace suite passed 130 tests (23 explicitly ignored integration/resource checks), desktop Rust passed 28, and both strict all-target Clippy checks and formatting passed. The ignored actual routing regression was executed separately as described above.