waynen12/ai-knowledge-manager repository overview

⁠AI Knowledge Manager

Open-source RAG (Retrieval-Augmented Generation) system for uploading documents, vector-indexing them, and chatting over that knowledge with multiple LLM providers.

⁠🚀 Quick Start

Get a complete AI knowledge management system running in under 2 minutes:

# Download the Docker Compose file
curl -O https://raw.githubusercontent.com/waynen12/ChatComplete/main/docker-compose.dockerhub.yml

# Start the full stack (Knowledge Manager + Qdrant + Ollama)
docker-compose -f docker-compose.dockerhub.yml up -d

# Access the application
open http://localhost:8080

⁠🏗️ Architecture

• Frontend: React + Vite with shadcn/ui components
• Backend: ASP.NET 8 Minimal APIs with Semantic Kernel
• Vector Database: Qdrant for embeddings storage
• LLM Support: OpenAI, Anthropic, Google Gemini, Ollama

⁠⚡ Features

• Drag & Drop Upload: Upload documents, PDFs, markdown files
• Smart Chunking: Intelligent document splitting and embedding
• Multi-LLM Support: Switch between OpenAI, Anthropic, Gemini, or Ollama per request
• Persistent Conversations: Chat history maintained across sessions
• Vector Search: Semantic similarity search through your knowledge base
• RESTful API: Full API access for integrations

⁠🔧 Configuration

⁠Environment Variables

Create a .env file or set via command line:

# Required for external LLM providers
OPENAI_API_KEY=your_openai_key
ANTHROPIC_API_KEY=your_anthropic_key
GEMINI_API_KEY=your_gemini_key

# Vector store configuration (pre-configured in compose)
ChatCompleteSettings__VectorStore__Provider=Qdrant
ChatCompleteSettings__VectorStore__Qdrant__Host=qdrant
ChatCompleteSettings__VectorStore__Qdrant__Port=6334  # gRPC port

# Ollama configuration (pre-configured in compose)
ChatCompleteSettings__OllamaBaseUrl=http://ollama:11434

Note: The docker-compose.dockerhub.yml file already includes these configurations. You only need to set the API keys.

⁠API Keys Setup

1. OpenAI: Get key from OpenAI Platform⁠
2. Anthropic: Get key from Anthropic Console⁠
3. Google Gemini: Get key from Google AI Studio⁠
4. Ollama: No API key needed - runs locally

⁠📚 Usage

⁠Upload Documents

curl -X POST http://localhost:8080/api/knowledge \
  -F "[email protected]" \
  -F "knowledgeId=my-docs"

⁠Chat with Your Knowledge

curl -X POST http://localhost:8080/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "knowledgeId": "my-docs",
    "message": "What is this document about?",
    "provider": "OpenAi"
  }'

⁠🐳 Deployment Options

⁠Full Stack (Recommended)

Includes Knowledge Manager, Qdrant, and Ollama:

curl -O https://raw.githubusercontent.com/waynen12/ChatComplete/main/docker-compose.dockerhub.yml
docker-compose -f docker-compose.dockerhub.yml up -d

⁠Minimal Setup

Just the Knowledge Manager (bring your own vector DB):

docker run -d -p 8080:7040 \
  -e OPENAI_API_KEY=your_key \
  -v ai-knowledge-data:/app/data \
  waynen12/ai-knowledge-manager:latest

⁠🔍 Health Monitoring

Check service status:

# Main application
curl http://localhost:8080/api/ping
# Expected: {"status":"healthy"}

# Qdrant vector database (REST API)
curl http://localhost:6333/health
# Expected: {"status":"ok"}

# Ollama LLM service
curl http://localhost:11434/api/version
# Expected: {"version":"..."}

# Check all services via Docker
docker-compose -f docker-compose.dockerhub.yml ps

Port Reference:

• 8080: AI Knowledge Manager web UI and API
• 6333: Qdrant REST API (health checks)
• 6334: Qdrant gRPC API (data operations - used by app)
• 11434: Ollama API

⁠📊 Supported File Types

• Documents: PDF, DOCX, TXT, MD
• Code: All major programming languages
• Structured: JSON, XML, CSV

⁠🔒 Security

• Runs as non-root user
• No sensitive data logged
• API keys managed via environment variables
• Container isolation and network security

⁠🏷️ Tags

• latest - Latest stable release
• v1.0.0 - Specific version releases
• main - Latest from main branch

⁠💾 Data Persistence

Database Location:

• Container: /app/data/knowledge.db (mounted to Docker volume)
• Self-hosted: /opt/knowledge-api/data/knowledge.db (outside /out directory)

Critical: The database MUST be outside the /out directory to survive deployments. The /out folder is wiped during updates.

Volumes:

• ai-knowledge-data: Application database, configuration, chat history
• qdrant-data: Vector embeddings storage
• ollama-data: Downloaded LLM models

⁠📖 Documentation

• Full Documentation & README⁠
• CLAUDE.md - Development Guide⁠
• Swagger API Docs⁠ (when running)

⁠🤝 Contributing

Visit the GitHub repository⁠ to contribute, report issues, or request features.

⁠📄 License

Open source under the MIT License.