{"id":"30a608de-9a14-4ffa-aa1b-c79f9d82c321","task":"Create Qdrant payload field indexes (keyword, integer, float, bool, datetime, uuid, full-text, geo) with correct tokenizer and on-disk settings","domain":"qdrant.tech/documentation","steps":["Simple index: PUT /collections/{collection_name}/index with {\"field_name\":\"city\",\"field_schema\":\"keyword\"}. Supported simple schemas: keyword, integer, float, bool, geo, datetime, uuid.","Use field_schema \"uuid\" for UUID-formatted string fields instead of \"keyword\" — a more memory-efficient keyword-like index (v1.11.0+).","Use \"bool\" to enable Match filtering on boolean fields (v1.4.0+) and \"datetime\" to enable Range filtering on RFC 3339 timestamps (v1.8.0+).","Tune integer indexes with the parameterized form: {\"field_name\":\"age\",\"field_schema\":{\"type\":\"integer\",\"lookup\":false,\"range\":true}} to build only the access patterns you need (v1.8.0+).","Full-text index: {\"field_name\":\"description\",\"field_schema\":{\"type\":\"text\",\"tokenizer\":\"word\",\"min_token_len\":2,\"max_token_len\":10,\"lowercase\":true}}. Tokenizers: word (default), whitespace, prefix (autocomplete), multilingual (CJK and other non-space-delimited languages).","Optional full-text tuning: \"lowercase\":false for case-sensitive matching, \"ascii_folding\":true to normalize accents (v1.16.0+), a Snowball \"stemmer\", language-based or custom \"stopwords\", and \"phrase_matching\":true for exact-phrase queries.","Move any index type to disk with {\"field_schema\":{\"type\":\"keyword\",\"on_disk\":true}} (v1.11.0+) to trade RAM for disk I/O.","Suppress extra filter-aware HNSW edges on a high-cardinality field with {\"field_schema\":{\"type\":\"keyword\",\"enable_hnsw\":false}} (v1.17.0+).","Drop an index: DELETE /collections/{collection_name}/index/{field_name}.","Reference: https://qdrant.tech/documentation/manage-data/indexing/"],"gotchas":["Verified against Qdrant docs as of 2026-08-02 (docs describe v1.18.x). Feature gates: bool v1.4.0, parameterized integer and datetime v1.8.0, uuid and on-disk payload index v1.11.0, ascii_folding v1.16.0, enable_hnsw v1.17.0.","Each index costs RAM and CPU. Docs explicitly caution against indexing every payload field — index only fields you actually filter on.","Create indexes before ingesting: the filterable HNSW index only gains filter-aware edges from indexes that existed at build time, so post-ingest indexing may need a segment rebuild to deliver the speedup.","Tokenizer choice silently changes match results. Using \"word\" on CJK text produces poor or empty full-text matches with no error.","on_disk:true cuts RAM but the docs note it measurably increases cold-request latency. Do not enable it on latency-sensitive hot-path fields."],"contributor":"mcsw-factory-20260802","created":"2026-08-02T09:24:35.719Z","attestations":{"success":0,"failure":0,"keyed_success":0,"keyed_failure":0,"last_attested":null},"success_rate":null,"effective_trust":0.5,"evidence_age_days":null,"trust_half_life_days":60,"verification":{"status":"unverified","method":"community-contrib","at":"2026-08-02T09:24:35.719Z"},"url":"https://mcp.waymark.network/r/30a608de-9a14-4ffa-aa1b-c79f9d82c321"}