Tool calling, human-in-the-loop approval, RAG and Langfuse tracing on a local Qwen model (Ollama).
170
Customer-support AI agent PoC built with LangChain and LangGraph, running on a local Qwen 2.5 model via Ollama. Source code and full README: https://github.com/bikashshah-lgtm/langchain-agent-poc
1.0 – LangChain tool-calling agent (agent.py)1.1 – adds LangGraph agent with human-in-the-loop approval (graph_agent.py)1.2 – adds Langfuse tracing (agent_langfuse.py)Requires Ollama on the host with qwen2.5:7b pulled.
docker run --rm -e OLLAMA_BASE_URL=http://host.docker.internal:11434 b1kash/langchain-agent-poc:1.2
Human-in-the-loop version (interactive):
docker run -it --rm -e OLLAMA_BASE_URL=http://host.docker.internal:11434 b1kash/langchain-agent-poc:1.2 python graph_agent.py
Optional tracing: pass LangSmith/Langfuse keys with --env-file .env (no quotes around values).
Note: built on Apple Silicon (ARM64).
Content type
Image
Digest
sha256:bc185900e…
Size
66.2 MB
Last updated
14 days ago
docker pull b1kash/langchain-agent-poc:1.2