acdeveloper-sci/fortriage

Purpose and Entry Points

ForTriage is an ephemeral Streamlit dashboard for prioritizing modernization risk in legacy Fortran systems. It consumes curated Forti4D analysis outputs, rather than parsing or modifying source itself. The process starts at streamlit_app.py, which delegates application initialization and navigation to fortriage/app.py.

Application Shell and Session State

The app shell owns Streamlit page routing, shared initialization, and the active-analysis lifecycle. Session-scoped state keeps uploaded/generated analysis available across dashboard pages without a database or cross-session persistence. Key paths: fortriage/app.py, fortriage/state.py, fortriage/config.py.

Upload and Analysis Boundary

Users provide a project/archive; the upload layer validates and stages it, while the pipeline invokes or consumes Forti4D analysis and converts its reports into dashboard-ready structures. Forti4D is an external, independent static-analysis engine and is the boundary for Fortran parsing, metrics, dependency analysis, and source-level risk computation. Key paths: fortriage/uploads.py, fortriage/pipeline.py, test_pipeline_manual.py.

Prioritized Analysis Model

The pipeline’s curated report corpus drives the application’s risk-ranked unit views: complexity, reachability/dead code, clones, coupling, dependencies, and score breakdowns. This is the central data boundary: UI code should consume normalized analysis results, not Forti4D internal diagnostics such as per-file audit/block dumps. Key paths: fortriage/pipeline.py, fortriage/state.py, fortriage/config.py.

Dashboard Exploration Views

Streamlit views present the same session analysis through complementary decision workflows: system-level prioritization, per-unit inspection, management-facing synthesis, and raw curated-report exploration. Navigation and selections should hand off through shared state rather than recomputing analysis. Key paths: fortriage/views/overview.py, fortriage/views/drilldown.py, fortriage/views/executive_summary.py, fortriage/views/report_explorer.py.

Visualization and Reusable UI

Shared components provide charts, risk/status badges, metric presentation, and unit-detail modal interactions; they are presentation-layer consumers of normalized analysis data. Altair is the material visualization runtime, embedded within Streamlit. Key paths: fortriage/components/charts.py, fortriage/components/metric_bar.py, fortriage/components/unit_modal.py.

AI Insights and External Service Boundary

AI Insights builds constrained context from aggregated reports and executive/prioritization findings, assembles prompt assets, and calls Google Gemini only when configured. It must not send raw Fortran source; non-AI dashboard functionality remains offline. Gemini credentials come from Streamlit secrets, and the feature degrades when no API key is available. Key paths: fortriage/views/ai_insights.py, fortriage/insights_context.py, fortriage/gemini_client.py, fortriage/prompts.py.

Runtime and Infrastructure

The runtime is Python 3.10+ with Streamlit, launched through uv run streamlit run streamlit_app.py; dependency resolution is defined by pyproject.toml and uv.lock. Deployment configuration is Streamlit-local in .streamlit/config.toml. No database, persistent project store, or cloud analysis backend is present; the only runtime external service is optional Google Gemini, while Forti4D remains an external analysis dependency.