qdrant-ultimate-rag

Ultimate Qdrant RAG Integration Test

This example demonstrates the "ultimate" usage of the goframe library, integrating several core components:

1. Repository Ingestion: Uses GitLoader to load the entire source code of this repo.
2. Code-Aware Splitting: Uses CodeAwareTextSplitter to split files into logical chunks (functions, classes) while maintaining metadata.
3. Modern Qdrant Features: Uses the refactored Qdrant store with Streaming Integration, Binary Quantization (RAM saving), and Payload Indexing (fast metadata filtering).
4. Local AI Infrastructure: Uses Docker to run Qdrant v1.16.0 and Ollama with Qwen3 models (qwen3-embedding:0.6b and qwen3-coder).

Prerequisites

• Docker and Docker Compose
• Go 1.25.6+

How to Run

1. Start Infrastructure:

   docker compose up -d
   

   This will start Qdrant and Ollama, and automatically pull the required models (may take a few minutes depending on your internet connection).

2. Run the Integration Test:

   go run examples/qdrant-ultimate-rag/main.go
   

What This Test Validates

• Memory Safety: The streaming pipeline in AddDocumentsBatch ensures that even a large repository won't cause OOM errors.
• Quantization: Verifies that the library correctly configures Qdrant's binary quantization.
• RAG Accuracy: Demonstrates that the code-aware splitting and retrieval work in harmony to provide high-quality context to the LLM.
• Ollama Integration: Validates the end-to-end flow with state-of-the-art Qwen3 models.

Clean Up

To stop and remove the containers:

docker compose down