#170: FTS Track 3 — offline search index builder + tokenizer parity gate#329

Merged

rdhyee merged 8 commits into

isamplesorg:mainfrom

rdhyee:feat/170-search-index-builder

Jul 10, 2026

Commits on Jul 10, 2026

#170: offline search index builder + tokenizer with JS/Python parity gate

rdhyee
and
claude
committed
#170: CI gate — tokenizer parity (Python+JS on shared JSON) + builder fixture

rdhyee
and
claude
committed
#170: hot-token isolation, keywords + ic.label fallback, prod build stats

rdhyee
and
claude
committed
#170: address all 4 Codex review findings (round 1)

rdhyee
and
claude
committed
#170: reconcile hot-token architecture with cold-bytes budget (Codex round 2)

rdhyee
and
claude
committed
#170: per-pid topk scoring + two-tier hot policy (Codex round 3)

rdhyee
and
claude
committed
#170: honest per-query budgets + shard_sizes.json (Codex round 4)

rdhyee
and
claude
committed
#170: df embedded in shipped rows — df.parquet becomes offline-only (Codex round 5)

rdhyee
and
claude
committed