VPE Hackathon: AI Agents for Life Sciences

Please visit the project page website for a detailed overview of the Hackathon tasks and results

⚠️ Warning ⚠️

• The results of the Hackathon are provided as with only minimal changes by the judges made during the evaluation and which were made to reproduce the participants final demos given at the end of the Hackathon.
• The Hackathon code is currently specific to the Code Ocean capsule environment which was used to boost participant productivity and ensure reproducibility of results.
• We are in the process of migrating the Talk2Biomodels code to a stand-alone repository following our internal code- and dev-ops to facilitate community engagement (be on the lookout 👀 for when we announce the completion of the migration).

Recommended environment setup options

1. From a code ocean capsule
2. From a Docker image locally or on the cloud
3. From a python virtual environment

Running the projects

Assumed folder structure

code/... # where the code lives
data/... # where data assets live
environment/... # where docker files and other environment setup scripts live
results/... # where results are written to

From the command line

Run the demo README
streamlit run /code/demo/streamlit_app.py
Run the Talk2Biomodels README
streamlit run /code/talk2biomodels/streamlit_app.py
Run the Talk2Cells README
streamlit run /code/talk2cells/streamlit_app.py
Run the Talk2KnowledgeGraphs README
streamlit run /code/talk2knowledgegraphs/streamlit_app.py

Caveats

• Talk2Biomodels is currently not integrated with Ollama, and instead uses OpenAI which requires setting the environmental variable OPENAI_API_KEY
• Demo, Talk2Cells, and Talk2KnowledgeGraphs require external data assets that were supplied during the Hackathon, which are explained in the setup and project specific READMEs