Update src/mosaic/inference/paladin.py

Add GitHub Actions workflows and comprehensive test suite

Fix remaining hardcoded data paths after rebase

Add comprehensive logging for batch processing verification

Fix segment_tissue unpacking in batch analysis

Add batch processing optimization for slide analysis

Fix user directory creation and hide HF Spaces logs locally

Fix circular import by moving get_data_directory to separate module

Fix model file location to use HuggingFace cache directory

Define DEFAULT_TISSUE_SITE_IDX constant for "Not Applicable" tissue site

Merge pull request #12 from pathology-data-mining/copilot/sub-pr-7-one-more-time

Add error handling for missing tissue site CSV file

Merge pull request #10 from pathology-data-mining/copilot/sub-pr-7-another-one

Merge pull request #9 from pathology-data-mining/copilot/sub-pr-7-again

Add error handling for missing CSV files in get_tissue_site_map and get_sex_map

Fix hardcoded path for target_dict.tsv using robust path construction

Update src/mosaic/inference/data.py

Update src/mosaic/analysis.py

Remove unused global cancer type mappings from aeon.py

Add Aeon model test suite and reproducibility scripts

Add comprehensive sex and tissue site parameter support

Complete implementation of sex and tissue site parameters

Add sex and tissue site parameters to Aeon inference

Fix user detection by decoding JWT token from referer

Add debug logging for request object

Improve user detection: check username and auth headers

Add request parameter to analyze_slide function signature

Add request parameter to analyze_slides function signature

Pass request parameter through UI to analyze_slide

Add gr.Request parameter to analyze_slide signature

Add dynamic GPU duration: 60s for anonymous, 300s for logged-in users

Fix T4 detection by checking actual GPU name

Add T4 GPU detection and optimized settings

Actually apply simplification: remove individual decorators, add single pipeline decorator

Simplify: single GPU decorator on pipeline instead of 4 separate calls

Reduce GPU durations to fit 300s total limit per request

Remove chunking completely - doesn't work with ZeroGPU

Increase chunk sizes for better performance

Implement proper multi-GPU-call chunking for ZeroGPU

Drastically reduce chunk sizes for ZeroGPU reliability

Make all GPU memory stats collection optional with try-except

Fix CUDA device error by checking availability before reset

Implement chunked processing for ZeroGPU to prevent token expiry

Fix ZeroGPU detection to use correct env var

Optimize batch sizes for H100 ZeroGPU and reduce Optimus duration

Fix ZeroGPU detection and increase workers for non-ZeroGPU environments

Increase Optimus GPU duration to 600s (max)

Separate GPU calls for each operation in HF Spaces

Import spaces before CUDA packages to fix initialization error

fix: increase zerogpu duration