qdrant-memory

Cache LLM responses and retrieve relevant context with Qdrant vector search.

Updated Jan 22, 2026

Search Tags:#rag #vector database #memory #token optimization #qdrant #llm context #semantic cache

Ready to run

npx skills add https://github.com/techwavedev/skillsets --skill qdrant-memory-techwavedev

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: qdrant-memory
Source: https://github.com/techwavedev/skillsets/tree/main/skills/qdrant-memory
Command: npx skills add https://github.com/techwavedev/skillsets --skill qdrant-memory-techwavedev

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires requests, boto3, sentence-transformers, torch, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill tackles the challenge of ever-increasing LLM context windows and associated costs by intelligently managing and retrieving only the most relevant information, drastically reducing token usage.

Core Features & Use Cases

Semantic Cache: Avoids LLM calls entirely for semantically similar queries, achieving 100% token savings.
Long-Term Memory: Retrieves only relevant context chunks instead of full conversation history, reducing context by 80-95%.
Hybrid Search: Combines vector similarity with keyword filtering for precise technical queries.
Use Case: When an agent needs to recall past decisions or code snippets, this skill retrieves only the pertinent details, making LLM prompts more efficient and cost-effective.

Quick Start

Use the qdrant-memory skill to store the current conversation context for future retrieval.

Frequently Asked Questions about qdrant-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I reduce LLM token consumption using a vector database?▼

Reduce LLM token consumption by using a vector database like Qdrant to implement semantic caching and retrieve only relevant context chunks. This avoids redundant LLM calls and shrinks context windows by up to 95%.

What is semantic caching for LLM context optimization?▼

Semantic caching for LLM context optimization stores vector embeddings of previous queries in Qdrant to match semantically similar requests. It bypasses LLM generation entirely for similar queries, achieving 100% token savings on cached responses.

How do I retrieve specific code patterns from long-term memory without full conversation history?▼

Retrieve specific code patterns from long-term memory by storing conversational data as vector embeddings in Qdrant. The system fetches only pertinent context chunks instead of full history, reducing context size by 80-95%.

Does hybrid search work with Qdrant for technical queries?▼

Hybrid search works with Qdrant by combining vector similarity with keyword filtering. This approach precisely retrieves relevant technical decisions and code snippets from stored memory.

Do I need sentence-transformers to use Qdrant for semantic caching?▼

You need sentence-transformers to generate vector embeddings for Qdrant semantic caching. The dependency converts text data into vector representations for efficient similarity matching and retrieval.