Qdrant

Build vector search with Qdrant using collections, payloads, filtering, and optimized indexing for semantic similarity.

Safety Notice

This listing is from the official public ClawHub registry. Review SKILL.md and referenced scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "Qdrant" with this command: npx skills add ivangdavila/qdrant

When to Use

User needs vector similarity search, semantic search, or recommendation systems. Agent handles collection design, point insertion, filtered queries, and index optimization.

Quick Reference

TopicFile
Query patternsqueries.md
Performance tuningperformance.md

Core Rules

1. Collection Setup

  • Set vector dimension to match embedding model (e.g., OpenAI ada-002 = 1536)
  • Choose distance metric deliberately: Cosine for normalized embeddings, Dot for raw scores, Euclid for absolute distance
  • Wrong dimension = silent failures with zero results

2. Payload Strategy

  • Store filterable metadata as payload fields
  • Index payload fields used in filters: create_payload_index
  • Don't store large blobs in payloads — use external storage + reference ID

3. Batch Operations

  • Insert points in batches of 100-1000, not one by one
  • Use upsert to handle duplicates by ID
  • Parallel uploads with wait=false then verify with collection info

4. Filtering vs Post-Filtering

WhenUse
Known constraintsFilter in query (pre-filter)
Score thresholdscore_threshold parameter
Complex logicCombine must, should, must_not
  • Pre-filtering reduces search space = faster
  • Post-filtering on results = slower, may miss relevant items

5. Search vs Scroll

NeedUse
Top-K similarsearch
All matchingscroll with filter
Paginated resultsscroll with offset
Export/backupscroll all with pagination

6. Index Optimization

  • HNSW parameters: increase m for recall, increase ef_construct for index quality
  • Default m=16, ef_construct=100 works for most cases
  • For millions of vectors: enable on_disk storage
  • Use quantization (scalar or product) to reduce memory 4-8x

7. Multi-Tenancy

  • Payload field for tenant ID + filter on every query
  • Or separate collections per tenant (simpler isolation, harder to manage)
  • Never expose one tenant's data to another

Common Traps

  • Creating collection with wrong vector size → all searches return empty
  • Forgetting wait=true on insert → querying before data indexed
  • Using scroll without limit → memory exhaustion on large collections
  • Not indexing payload fields → filter queries scan entire collection
  • Storing embeddings in payload instead of vector field → defeats purpose

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

General

Workout Readiness Check In

Create a same-day workout readiness decision card for users who planned to exercise today but feel tired, sore, stressed, underslept, ill, or uncertain. Use...

Registry SourceRecently Updated
General

URL to Markdown

Convert HTML web pages from HTTP/HTTPS URLs to clean, readable Markdown files with optional batch processing and formatting features.

Registry SourceRecently Updated
General

Appliance Manual Command Card

Assist in creating a one-page appliance command card for tracking maintenance, troubleshooting, warranty info, and repair preparation.

Registry SourceRecently Updated
General

Fireseed Writing Expert

全能小说写作专家,支持从零创作、DNA提取与克隆、风格仿写、一键发布到 fireseed.online。用户可以用自然语言描述需求,系统自动调度子技能完成科学、闭环的写作与发布流程。

Registry SourceRecently Updated