[diegosouzapw]

F4.4 / Q1 — Opt-in Qdrant scalar int8 quantization

Closes the Q1 half of the rodada4 F4.4 (vector quantization) gap — the last fully-local
implementation gap in the compression track (most of rodada4 was already done; audit
2026-06-18). Q2 (sqlite-vec int8) ships as a separate follow-up PR (recall-sensitive).

What

Adds an opt-in vector quantization mode to the Qdrant memory collection. Default is
none, so existing deployments are byte-identical — no silent behavior change.

When set to int8 (or binary):

• new collections are created with a quantization_config
  (int8: scalar { type: int8, always_ram: true, quantile: 0.99 }; binary: { always_ram: true })
  → ~4× (int8) / up to ~32× (binary) less vector RAM;
• searches send params.quantization.rescore: true so the original vectors refine the
  quantized shortlist (preserves recall).

Scope (surgical, opt-in)

• src/lib/memory/qdrant.ts — QdrantQuantization type + QdrantConfig.quantization;
  pure helpers buildQuantizationConfig / searchQuantizationParams wired into
  ensureCollection (create body) and searchSemanticMemory (search params).
• src/shared/schemas/qdrant.ts — QdrantQuantizationSchema enum in settings + update schemas.
• src/app/api/settings/qdrant/route.ts — persist + expose qdrantQuantization.

Deliberately NOT in this PR (follow-ups)

• Existing collections are left untouched — switching modes on a populated collection
  is a destructive recreate+reindex that must be an explicit UI action (never automatic,
  per the spec's "what NOT to do").
• UI select (none/int8/binary) in MemorySkillsTab + the migration endpoint.
• Q2 — sqlite-vec int8 local storage (separate PR; recall@K fixture validation).

Tests (TDD)

src/lib/memory/__tests__/qdrant-wiring.test.ts (vitest) — 9/9 green: default none;
mode parse (int8/binary/invalid→none); buildQuantizationConfig (none→undefined =
byte-identical body, int8→scalar, binary→binary); searchQuantizationParams (rescore only
when quantized). Adjacent suites green: qdrant UI card 5/5, qdrant integration 13/13.

Gates: typecheck:core / lint / check:cycles clean.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[gemini-code-assist]

Code Review

This pull request introduces support for vector quantization (none, int8, and binary) in Qdrant memory collections. It updates the API routes, configuration normalization, collection creation, and search logic to handle quantization parameters and rescoring, and adds corresponding Zod schema validation and unit tests. No review comments were provided, so there is no feedback to address.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.