Dont directly eval tool call

Make auth safer

Merge branch 'main' into feature/benchmark

fixes

tool changes

other modifications

useful scripts

revamped benchmarking suite

updated benchmarks

Merge branch 'feature/benchmark' of github.com:bowang-lab/MedRAX2 into feature/benchmark

Add security to gradio

CLI/env/SLURM-aware resolver

Configurable MedGemma API URL and auto-pick free loopback IP/port

Fix merge conflicts

Improve style

Improve style

Clean up API

Support customizing systemp prompt and API endpoint

setup api

Setup argparse

hard code a fixed medgemma ip [WIP]

Fix merge conflicts

GPT-5

Fix interface bugs

Fix bugs of interface

Fix bugs of interface

Add GPT-4.1

Add parallel tool calling

Merge pull request #26 from bowang-lab/victor/benchmarking

updates

fixed rexvqa benchmark and added handling for image norm for tools

MedGemma fixes

Fix env loading of MedGemma

adding scripts

partially fixed rexvqa benchmark

added concurrency to benchmarking

Add env loading for HF

Remove top_p

Update langgraph and langchain version requiremenets

Fix merge conflicts

Setup MedRAX2 for test

Merge pull request #23 from bowang-lab/victor/benchmarking

fixed medgemma

fixed up some things and corrected agent init logic

cleared merge issues

yes:wq

adding medgemma

Merge pull request #24 from bowang-lab/emily/medgemma

Merge branch 'main' into emily/medgemma

updates system prompt