rag-chatbot

rag-chatbot example

A tiny terminal chatbot that loads a local text file, splits it into chunks, indexes it in Qdrant, and answers questions using an agent with a RAG tool.

This example uses Qdrant over gRPC.

Start Qdrant

If you don't already have Qdrant running, one quick way is Docker:

docker run --rm -p 6333:6333 -p 6334:6334 qdrant/qdrant

Configure the LLM + embeddings

The example uses OpenAI-compatible endpoints via environment variables:

• OPENAI_API_KEY (optional for local OpenAI-compatible servers)
• OPENAI_BASE_URL (optional; if unset and no key is provided, it defaults to an Ollama-compatible base URL)
• OPENAI_MODEL (optional)

Option A: Local Ollama (OpenAI-compatible)

export OPENAI_BASE_URL=http://localhost:11434/v1
export OPENAI_MODEL=gpt-oss:20b-cloud
# OPENAI_API_KEY can be empty for Ollama

Option B: OpenAI

export OPENAI_API_KEY=...your_key...
export OPENAI_MODEL=gpt-4o-mini

# optional
# export OPENAI_BASE_URL=https://api.openai.com/v1

Run

From the repo root:

go run ./examples/rag-chatbot -file /path/to/your/file.txt

Type your questions and press Enter. Type exit to quit.

Useful flags

• -chunk-size / -chunk-overlap: controls the splitter
• -topk: how many chunks to retrieve per tool call
• -qdrant-host / -qdrant-port: Qdrant connection (defaults to localhost:6334)
• -qdrant-collection: Qdrant collection name (defaults to rag_chatbot)
• -qdrant-api-key: optional API key
• -qdrant-tls: enable TLS