[szabosteve]

What this PR does

This PR brings the skills in line with the 1.20 docs.

Skill	Summary of changes	Relevant PRs
qdrant-search-quality/search-strategies/hybrid-search/search-types	Recommends BM25 on a full-text index (scoring: true, text query) as the 1.20+ default and keeps sparse-vector BM25 as the fallback for 1.19 and older. Adds full-text caveats (language config, corpus statistics, nested scoring) and a "Migrate from sparse BM25" section. Marks the avg_len guidance sparse-only. Multi-field advice now uses one BM25 index per field. Adds "What NOT to Do" items.	#2981, #2982, #2983, #2985, #2987
qdrant-search-quality/search-strategies/hybrid-search (hub)	Per-tenant idf now covers BM25 text queries on 1.20+, including average document length. Anchor renamed to per-tenant-scoring-statistics. Adds a trigger phrase.	#2985
qdrant-search-quality/search-strategies/hybrid-search/combining-searches	Adds the max/min formula expressions (1.19.1) and the missing-prefetch defaults pitfall. Adds acosh hyperbolic rescoring.	#2965, #2980
qdrant-search-quality/diagnosis	Ranking text now points to a scoring full-text index on 1.20+, not only sparse vectors. A full-text index without scoring only filters.	#2981, #2983
qdrant-multitenancy	Adds a bullet on per-tenant scoring statistics via idf, now covering BM25 text queries.	#2985
qdrant-model-migration	On 1.20+, adding BM25 to a dense collection needs no vector backfill: create a scoring text index. The sparse-vector path stays for 1.19 and older.	#2981, #2983, #2986
qdrant-edge	Clarifies that Edge BM25 is compatible with the server's sparse-vector BM25 model.	#2982
qdrant-performance-optimization/memory-usage-optimization	Adds the ID tracker memory tier, the inline-storage single-copy layout and hnsw_config.memory, and the turbo8/turbo16 datatypes. Notes that uint8 suits only pre-quantized vectors. Adds 8-bit TurboQuant (bits8) as the recommended quantization on 1.20+.	#2955, #3001, #3003, #3015
qdrant-sizing	Adds bytes per dimension for the turbo datatypes, the 1–65,536 dense dimension range, and BM25 scoring index overhead. Adds ~4 bytes per point for per-element indexes, the ID tracker tier, and the inline-storage tier rules.	#2955, #2965, #2981, #3001, #3003, #3014
qdrant-scaling/scaling-data-volume/vertical-scaling	Replaces the turbo4-only line with the full TurboQuant datatype set. Adds the ID tracker tier as a way to free RAM.	#2955, #3003
qdrant-performance-optimization/search-speed-optimization	Nested filters now recommend per-element indexes. Warns that substring match is expensive and rejected under strict mode, the Cloud default.	#2965, #3014
qdrant-monitoring/setup	Audit logs can go to stdout or stderr via audit.output (1.20+), which removes the sidecar and Persistent Volume workaround. Notes the "kind": "audit" field and that /audit/logs is empty for stream-only nodes.	#2965

Type

• new skill
• skill improvement
• bug fix
• repo hygiene

[netlify]

✅ Deploy Preview for qdrant-skills ready!

Name	Link
🔨 Latest commit	88b7718
🔍 Latest deploy log	https://app.netlify.com/projects/qdrant-skills/deploys/6ada059306f137f9e74041ca
😎 Deploy Preview	https://deploy-preview-190--qdrant-skills.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.