valPDB-Agent

A LangChain-based agent for protein structure evaluation. The agent uses large language models to orchestrate protein analysis tools, supports local Ollama models, and runs in a Docker-based sandbox.

The system provides an interactive command-line interface for protein structure assessment (for example, Ramachandran plot generation).

Usage

1. Build the Docker image

docker pull swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/continuumio/anaconda3:latest
docker tag  swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/continuumio/anaconda3:latest  docker.io/continuumio/anaconda3:latest

docker build -t protein-agent:latest .

2. Prepare environment variables

Create a .env file in the project root, for example:

LLM_PROVIDER=ollama
LLM_MODEL=modelscope.cn/Qwen/Qwen3-4B-GGUF:latest
LLM_BASE_URL=http://<host-ip>:11434
LLM_API_KEY=dummy

Note:
If Ollama is running on the host machine, LLM_BASE_URL must point to a host-accessible address. Do not use localhost inside the container.

3. Run the agent (interactive mode)

docker run -it \
  --network=host \
  --env-file .env \
  -v /your/data/path:/data \
  --name protein-agent \
  protein-agent:latest

The agent will start in interactive mode. If required information (such as file paths or output locations) is missing, the agent will ask for clarification before executing any tool.

4. Retrieve outputs

To copy results from the container to the host:

docker cp protein-agent:/app/output ./output

This workflow does not require mounting any host directories.

5. Cleanup

docker rm -f protein-agent

Status

This project is under active development. Additional protein evaluation and generation capabilities will be added.