QDrant Loader






📝 [Changelog v1.0.4](./CHANGELOG.md) - Latest improvements and bug fixes

<div align="left">

A comprehensive toolkit for loading data into Qdrant vector database with advanced MCP server support for AI-powered development workflows.

</div>

🎯 What is QDrant Loader?

QDrant Loader is a data ingestion and retrieval system that collects content from multiple sources, processes and vectorizes it, then provides intelligent search capabilities through a Model Context Protocol (MCP) server for AI development tools.

Perfect for:

• 🤖 AI-powered development with Cursor, Windsurf, and other MCP-compatible tools
• 📚 Knowledge base creation from technical documentation
• 🔍 Intelligent code assistance with contextual information
• 🏢 Enterprise content integration from multiple data sources

📦 Packages

This monorepo contains three complementary packages:

🔄 QDrant Loader

Data ingestion and processing engine

Collects and vectorizes content from multiple sources into QDrant vector database.

Key Features:

• Multi-source connectors: Git, Confluence (Cloud & Data Center), JIRA (Cloud & Data Center), Public Docs, Local Files
• File conversion: PDF, Office docs (Word, Excel, PowerPoint), images, audio, EPUB, ZIP, and more using MarkItDown
• Smart chunking: Modular chunking strategies with intelligent document processing and hierarchical context
• Incremental updates: Change detection and efficient synchronization
• Multi-project support: Organize sources into projects with shared collections
• Provider-agnostic LLM: OpenAI, Azure OpenAI, Ollama, and custom endpoints with unified configuration

⚙️ QDrant Loader Core

Core library and LLM abstraction layer

Provides the foundational components and provider-agnostic LLM interface used by other packages.

Key Features:

• LLM Provider Abstraction: Unified interface for OpenAI, Azure OpenAI, Ollama, and custom endpoints
• Configuration Management: Centralized settings and validation for LLM providers
• Rate Limiting: Built-in rate limiting and request management
• Error Handling: Robust error handling and retry mechanisms
• Logging: Structured logging with configurable levels

🔌 QDrant Loader MCP Server

AI development integration layer

Model Context Protocol server providing search capabilities to AI development tools.

Key Features:

• MCP Protocol 2025-06-18: Latest protocol compliance with dual transport support (stdio + HTTP)
• Advanced search tools: Semantic search, hierarchy-aware search, attachment discovery, and conflict detection
• Cross-document intelligence: Document similarity, clustering, relationship analysis, and knowledge graphs
• Streaming capabilities: Server-Sent Events (SSE) for real-time search results
• Production-ready: HTTP transport with security, session management, and health checks

🚀 Quick Start

Installation

# Install both packages
pip install qdrant-loader qdrant-loader-mcp-server

# Or install individually
pip install qdrant-loader          # Data ingestion only
pip install qdrant-loader-mcp-server  # MCP server only

5-Minute Setup

1. Create a workspace

```bash

mkdir my-workspace && cd my-workspace

```

1. Initialize workspace with templates

```bash

qdrant-loader init --workspace .

```

1. Configure your environment (edit .env)

```bash

# Qdrant connection

QDRANT_URL=http://localhost:6333

QDRANTCOLLECTIONNAME=my_docs

# LLM provider (new unified configuration)

OPENAIAPIKEY=youropenaikey

LLM_PROVIDER=openai

LLMBASEURL=https://api.openai.com/v1

LLMEMBEDDINGMODEL=text-embedding-3-small

LLMCHATMODEL=gpt-4o-mini

```

1. Configure data sources (edit config.yaml)

```yaml

global:

qdrant:

url: "http://localhost:6333"

collectionname: "mydocs"

llm:

provider: "openai"

base_url: "https://api.openai.com/v1"

apikey: "${OPENAIAPI_KEY}"

models:

embeddings: "text-embedding-3-small"

chat: "gpt-4o-mini"

embeddings:

vector_size: 1536

projects:

my-project:

project_id: "my-project"

sources:

git:

docs-repo:

base_url: "https://github.com/your-org/your-repo.git"

branch: "main"

file_types: [".md", ".rst"]

```

1. Load your data

```bash

qdrant-loader ingest --workspace .

```

1. Start the MCP server

```bash

mcp-qdrant-loader --env /path/tp/your/.env

```

🔧 MCP-Compatible IDE Setup

QDrant Loader works with any IDE/tool that supports MCP, including Cursor, Windsurf, and Claude Desktop.

Minimal MCP server entry (adapt path/format to your tool):

{
  "mcpServers": {
    "qdrant-loader": {
      "command": "/path/to/venv/bin/mcp-qdrant-loader",
      "env": {
        "QDRANT_URL": "http://localhost:6333",
        "QDRANT_COLLECTION_NAME": "my_docs",
        "OPENAI_API_KEY": "your_key"
      }
    }
  }
}

Alternative: Use configuration file (recommended for complex setups):

{
  "mcpServers": {
    "qdrant-loader": {
      "command": "/path/to/venv/bin/mcp-qdrant-loader",
      "args": [
        "--config",
        "/path/to/your/config.yaml",
        "--env",
        "/path/to/your/.env"
      ]
    }
  }
}

For tool-specific setup and exact config format:

• [MCP Setup and Integration](./docs/users/detailed-guides/mcp-server/setup-and-integration.md) - Full guide
• [Cursor Setup](./docs/users/detailed-guides/mcp-server/setup-and-integration.md#-cursor-ide)
• [Windsurf Setup](./docs/users/detailed-guides/mcp-server/setup-and-integration.md#-windsurf)
• [Claude Desktop Setup](./docs/users/detailed-guides/mcp-server/setup-and-integration.md#-claude-desktop)

Example queries in AI tools:

• "Find documentation about authentication in our API"
• "Show me examples of error handling patterns"
• "What are the deployment requirements for this service?"
• "Find all attachments related to database schema"

📚 Documentation

Getting Started

• [Getting Started](./docs/getting-started/) - Quick start and core concepts
• [Installation Guide](./docs/getting-started/installation.md) - Complete setup instructions
• [Quick Start](./docs/getting-started/quick-start.md) - Step-by-step tutorial
• [Core Concepts](./docs/getting-started/README.md#-core-concepts) - Understand the core architecture: workspace model, projects and sources, ingestion pipeline, and MCP search flow

User Guides

• [User Guides](./docs/users/) - Detailed usage instructions
• [Configuration](./docs/users/configuration/) - Complete configuration reference
• [Data Sources](./docs/users/detailed-guides/data-sources/) - Git, Confluence, JIRA setup
• [File Conversion](./docs/users/detailed-guides/file-conversion/) - File processing capabilities
• [MCP Server](./docs/users/detailed-guides/mcp-server/) - AI tool integration

🛠️ Developer Resources

• [Developer hub](./docs/developers) - Developer guides for architecture, testing, deployment, and contribution workflows.
• [Architecture](./docs/developers/architecture/) - System design overview
• [Testing](./docs/developers/testing/) - Testing guide and best practices

🆘 Support

• [Issues](https://github.com/martin-papy/qdrant-loader/issues) - Bug reports and feature requests
• [Discussions](https://github.com/martin-papy/qdrant-loader/discussions) - Community Q&A

🤝 Contributing

We welcome contributions! See our Contributing Guide for:

• Development environment setup
• Code style and standards
• Pull request process

Quick Development Setup

# Clone and setup
git clone https://github.com/martin-papy/qdrant-loader.git
cd qdrant-loader

# Sync workspace environment (recommended)
uv sync --all-packages --all-extras

# Add a new dependency during development
uv add fastapi
uv sync

📄 License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Ready to get started? Check out our Quick Start Guide or browse the complete documentation.