Qdrant RAG Skills

Agent skills for building production RAG and agentic systems with Qdrant

Skills encode deep knowledge about RAG pipeline architecture, agentic retrieval patterns, multi-tenancy, sparse vectors, reranking, and payload indexing so coding agents can make the engineering decisions that determine whether your retrieval system actually works in production.

These skills complement the official qdrant/skills repo, which covers core Qdrant operations (scaling, monitoring, deployment, SDK usage). This repo focuses on application-level patterns: how to build, debug, and scale RAG pipelines and agentic systems on top of Qdrant.

---

📦 Installation

Claude Code

/plugin marketplace add inamdarmihir/agentic-rag-skills

Cursor

Add manually via Settings > Rules > Add Rule > Remote Rule (GitHub) with inamdarmihir/agentic-rag-skills.

npx skills

npx skills add https://github.com/inamdarmihir/agentic-rag-skills

Clone / Copy

Clone and copy the skills/ folder into your agent's skill directory:

Agent	Skill Directory	Version	Docs
🤖 Claude Code	~/.claude/skills/	Latest	docs
🖱️ Cursor	.cursor/skills/	0.50+	docs
💻 OpenCode	~/.config/opencode/skill/	Latest	docs
🧠 OpenAI Codex	~/.codex/skills/	Latest	docs

---

⚡ Quick Start

Skills activate automatically when your question matches their description:

"My RAG pipeline returns relevant chunks but the answers are still wrong"
→ qdrant-rag-pipeline/context-assembly activates
  walks through ordering, deduplication, token budgeting

"I need to build an agent that can search Qdrant multiple times to answer a question"
→ qdrant-agentic-rag/iterative-retrieval activates
  covers loop termination, follow-up query generation, progressive widening

"Product codes aren't being found by our vector search"
→ qdrant-sparse-vectors activates
  covers BM25/SPLADE setup and hybrid search with Qdrant

"How do I serve 5000 customers from one Qdrant collection?"
→ qdrant-multitenancy activates
  covers payload-based isolation, tenant_id indexing, shard-per-tenant patterns

---

🧠 Skills

Skill	Useful For
🏗️ qdrant-rag-pipeline	End-to-end RAG pipeline architecture and debugging
✂️ qdrant-rag-pipeline/chunking	Document splitting strategy, chunk size, overlap, metadata
🔍 qdrant-rag-pipeline/retrieval	Query strategies, filters, hybrid search, score thresholds
📋 qdrant-rag-pipeline/context-assembly	Ordering chunks, deduplication, token budgeting, source attribution
🤖 qdrant-agentic-rag	Agentic RAG overview and pattern selection
🗺️ qdrant-agentic-rag/query-planning	Complex question decomposition, query rewriting, collection routing
🔁 qdrant-agentic-rag/iterative-retrieval	Multi-step retrieval loops, self-correction, termination conditions
💾 qdrant-agentic-rag/agent-memory	Conversation history, episodic memory, semantic memory, memory decay
🏢 qdrant-multitenancy	Collection-per-tenant vs payload isolation, preventing data leaks
🔤 qdrant-sparse-vectors	BM25 and SPLADE sparse vectors, hybrid dense+sparse search
🎯 qdrant-reranking	Cross-encoder reranking, candidate set sizing, latency management
🗂️ qdrant-payload-indexing	Index types, cardinality, compound filters, debugging slow queries

---

📦 Qdrant Version Compatibility

Component	Package	Version
🗄️ Vector DB	Qdrant	1.9.0+
🐍 Python Client	qdrant-client	^1.9.0
📦 JS/TS Client	@qdrant/js-client-rest	^1.11.0
🦀 Rust Client	qdrant-client	^1.10.0
🐹 Go Client	go-qdrant	^1.9.0

---

🤝 Pair with Official Qdrant Skills

For best coverage, use these skills alongside the official qdrant/skills repo:

If you need...	Use...
🏗️ RAG pipeline architecture	This repo: qdrant-rag-pipeline
📈 Collection scaling and sharding	qdrant/skills: qdrant-scaling
🤖 Agent memory and iterative retrieval	This repo: qdrant-agentic-rag
🔍 Search quality diagnosis	qdrant/skills: qdrant-search-quality
🏢 Tenant isolation and multi-tenancy	This repo: qdrant-multitenancy
📊 Cluster monitoring	qdrant/skills: qdrant-monitoring
🔤 Sparse vectors and hybrid search	This repo: qdrant-sparse-vectors
🛠️ SDK setup and code examples	qdrant/skills: qdrant-clients-sdk

---

🔌 MCP Servers

Pair with these Qdrant MCP servers for live retrieval context during development:

Server	Purpose
🔍 mcp-code-snippets	Search Qdrant docs and code examples across all SDKs
🗄️ mcp-server-qdrant	Store and retrieve memories, manage collections directly

---

🤝 Contributing

See CONTRIBUTING.md for skill writing conventions.

Found wrong advice? Open an issue with:

• 📛 The skill name
• 💬 The prompt you gave your agent
• ❌ What the agent said vs ✅ what it should have said

---

_{Powered by Qdrant · Works with Claude Code, Cursor, OpenCode, and Codex}