RAG Service Documentation

Quick navigation for RAG service documentation.

Contents

Document	Description
README	Overview and quick start
Architecture	System design and data flow
API Endpoints	Complete API reference
Embeddings	Ollama embedding service
Vector Store	Qdrant integration
Document Processing	Chunking and file loading
Configuration	Environment variables
Development	Local setup and testing
Deployment	Docker and production

Quick Reference

Start Service

# With Docker Compose
docker compose up -d qdrant ollama rag_service

# Initialize model
docker exec ollama ollama pull nomic-embed-text

# Local development
cd rag_service
pip install -e ".[dev]"
uvicorn rag_service.api:app --reload --port 8081

Key URLs

URL	Purpose
http://localhost:8081	API root
http://localhost:8081/health	Health check
http://localhost:8081/docs	Swagger UI
http://localhost:8081/metrics	Prometheus metrics
http://localhost:6333/dashboard	Qdrant UI

Common API Calls

# Health check
curl http://localhost:8081/health

# Semantic search
curl -X POST http://localhost:8081/search \
  -H "Content-Type: application/json" \
  -d '{"query": "fuzzy logic", "asignatura": "logica-difusa"}'

# Upload file
curl -X POST http://localhost:8081/upload \
  -F "file=@document.pdf" \
  -F 'metadata={"asignatura":"iv","tipo_documento":"teoria","auto_index":true}'

# List files
curl http://localhost:8081/files

# List subjects
curl http://localhost:8081/subjects

Configuration

Variable	Default	Description
QDRANT_HOST	qdrant	Vector DB host
OLLAMA_HOST	ollama	Embedding host
OLLAMA_MODEL	nomic-embed-text	Embedding model
TOP_K_RESULTS	5	Search results
SIMILARITY_THRESHOLD	0.5	Min score

Architecture Diagram

graph TB
    subgraph "RAG Service :8081"
        API[FastAPI API]
        FL[FileLoader]
        DP[DocumentProcessor]
        ES[EmbeddingService]
        VS[VectorStoreService]
    end
    
    subgraph "External"
        Ollama[Ollama :11434]
        Qdrant[(Qdrant :6333)]
        FS[File System]
    end
    
    API --> FL
    API --> VS
    FL --> FS
    FL --> DP
    DP --> VS
    VS --> ES
    ES --> Ollama
    VS --> Qdrant

Data Flow

1. Upload: File → FileLoader → Document
2. Process: Document → DocumentProcessor → Chunks
3. Embed: Chunks → EmbeddingService → Vectors
4. Store: Vectors → VectorStoreService → Qdrant
5. Search: Query → Embedding → Qdrant → Results

---

← Back to Services