WattIQ · full-stack request tracer
Two phases. Background (offline): how the EPM least-cost planning model is cloned from GitHub and its output — plus the country's policy documents — are ingested into the grounding corpora that every query references. Live (per query): what happens the instant a user types a question and clicks Analyze. Each stage lights up; the right-hand window shows the real code, with the decisive line highlighted to lead your eye.
Next lets you manually click through all steps in the program from the user query. On each click, watch the code window to see what is happening.
Analyze automatically traces all steps in the program from the user query. As it runs, watch the code window on the right to see exactly what is happening.
EPM/tools/sync.ps1, scripts/azure_ingest_country_doc.py, scripts/azure_create_search_index.py, poc/app/datasource.py, poc/app/data_tools.py, poc/app/agent.py, poc/app/main.py, poc/app/context_retriever.py, poc/app/azure_search_retriever.py, poc/app/provenance.py, poc/app/insights.py, poc/app/evidence_graph.py, poc/web/app.js, poc/web/turn.js. Timings are representative of a warm request, not live measurements.