CapstoneProject is an end-to-end system for public-opinion analysis. It keeps separate evidence, coordination, reporting, and UI components; the current Coordinator implementation uses an EvidenceGraph-centered intelligence layer to gather observable public evidence, model quality and repeated coverage, audit claims against cited excerpts, and turn cited insights into editable, exportable reports through Signal Studio.

Positioning

Core strengths: multi-agent evidence analysis, inspectable reasoning, structured report IR, Signal Studio UI, OpenAPI contracts, and acceptance-ready docs. See Contribution And Novelty.

Product Flow

Architecture At A Glance

Interface Preview

More screenshots: Signal Studio Screenshots.

Quick Start

Use Python 3.11 or newer. If python is not on PATH, use Conda or the uv commands shown in the full Setup Guide.

python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip
pip install -r requirements.txt
python -m playwright install chromium
Copy-Item .env.example .env

Configure provider keys in .env, then build the final UI:

cd frontend
npm install
npm run build
cd ..

Start the Flask runtime:

python app.py

uv equivalent:

uv run --with-requirements requirements.txt python app.py

Open:

http://127.0.0.1:5000

For the complete setup path, provider profile, artifact review path, and verification checklist, use the Setup Guide.

Core API Surface

References:

• API Reference
• OpenAPI YAML
• Coordinator Output Schema
• Report IR

Coordinator Intelligence Runtime

POST /api/coordinator/run keeps the public endpoint name for compatibility and executes AgentCoordinator.run(). Internally, the Coordinator uses its intelligence layer, then writes one fused coordinator_output_latest.json artifact with schema 2.1-coordinator-intelligence:

• coordinator_intelligence: the internal evidence ledger with EvidenceGraph, quality_summary, freshness_summary, provider_diagnostics, audit_summary, and cited insights.
• Compatibility fields such as synthesis, source_data, divergence_matrix, fact_opinion_separation, and coordinator_trace; these are evidence-derived views, not a separate second artifact.
• Explicit provider diagnostics for missing or failing Jina semantic quality, QueryEngine LLM routing, source acquisition providers (Tavily, Bocha, Anspire), and local replay fallback routes.

If optional external search or rerank providers are not configured, the engine records not_configured diagnostics and continues the Coordinator flow. Local replay is disabled by default; it is intended for demos/tests only and is used only when COORDINATOR_ALLOW_REPLAY_FALLBACK=true, with local_fixture:used recorded in the artifact.

Documentation

Repository Map

CapstoneProject/
|-- app.py                    # Flask orchestrator and final runtime API gateway
|-- frontend/                 # Signal Studio React/Vite source
|-- AgentCoordinator/         # Coordinator entry point, internal intelligence layer, artifact cache
|   `-- intelligence/         # EvidenceGraph contracts, source gateway, quality pipeline, audit, synthesis
|-- QueryEngine/              # Search, trust, stance, social enrichment
|-- MediaEngine/              # Media-side research
|-- ReportEngine/             # Report generation graph, IR, renderers, exports
|-- ForumEngine/              # Forum-style coordination utilities
|-- SentimentAnalysisModel/   # Sentiment and topic model assets
|-- api_evaluation/           # Provider benchmark harness and results
|-- tests/                    # Regression and contract tests
`-- docs/                     # Authoritative documentation library

Quality And Operations

License

This repository is distributed under the GNU General Public License v2.0.

Project Status

The main application path is the Flask-served Signal Studio runtime. The authoritative project documentation is in docs/, with runtime asset boundaries documented in Runtime Assets.