ActuaryGPT Triangle Diagnostic Workbench

Production-oriented Streamlit app for actuarial triangle diagnostics, reserving review, and AI commentary grounded on deterministic outputs.

Run locally

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
streamlit run app.py

Set OPENAI_API_KEY for live AI responses; otherwise deterministic fallback commentary is used.

Test

pytest -q

Architecture

• app.py: Streamlit entrypoint.
• actuarygpt_workbench/pages/main_workspace.py: seven-section professional workspace.
• actuarygpt_workbench/services/*: ingestion, triangle, samples, export.
• actuarygpt_workbench/actuarial/*: diagnostics + reserving engines.
• actuarygpt_workbench/ai/copilot.py: structured AI service and prompt templates.
• tests/*: parsing, triangles, diagnostics, exports, sample model smoke.