Project Aegis / A.N.I.E. (Acuity Narrative Integrity Engine)

Open-source forensic analysis platform that detects manipulation, bias, and psychological operations ("psyops") in news articles and web content.

Given a URL or body of text, the system returns a 0–100 Narrative Integrity Score along with a detailed breakdown across four forensic vectors: Reality Anchoring, Tribal Engineering, Neuro-Linguistic Intent, and Logical Integrity.

Live site: www.goanie.com

If this project is useful to you, please star the repo — it helps others discover it.

---

Table of Contents

• What's in this repo
• How it works
• Tech stack
• Quick start
  • Backend
  • Website
  • Chrome extension
  • Mobile app
• API overview
• Full documentation
• Contributing
• License

---

What's in this repo

This is a monorepo containing every client and service that powers A.N.I.E.:

Folder	Description
aegis-backend/	Python FastAPI service. Exposes /v1/scan, /v1/chat, /v1/feedback, /v1/stats. Runs the analysis engine, scraper, and cache.
anie-site/	Static HTML/CSS/JS website (deployed on Netlify).
aegis-extension-mvp/	Chrome Manifest V3 extension for in-page article scanning.
anie-mobile/	React Native / Expo mobile app (iOS + Android) with a native share-sheet extension.
DOCUMENTATION.md	Complete technical documentation (architecture, engine internals, DB schema, Synapse directive history, API reference).

---

How it works

        Chrome Extension   Mobile App   Website
                \             |             /
                 \            |            /
                  \_________  |  _________/
                            \ | /
                             \|/
                              v
                   ┌─────────────────────┐
                   │  FastAPI Backend    │
                   │  (aegis-backend/)   │
                   └──────────┬──────────┘
                              │
          ┌───────────────────┼────────────────────┐
          v                   v                    v
      Supabase            xAI Grok            Tavily
      (cache +          (analysis             (truth
      events)            engine)              layer)

Scan flow:

1. Client (extension / mobile / website) POSTs a URL (or raw text) to /v1/scan.
2. Backend checks the Supabase cache (24-hour TTL, keyed by a normalized URL hash).
3. On cache miss, the scraper fetches article text (Firecrawl → Jina.ai → direct fetch).
4. The engine extracts source claims and runs parallel Tavily searches for fact-context.
5. The xAI Grok model applies the Psyop Hunter Protocol — a 24-marker forensic checklist across four vectors — and returns a score, verdict, per-vector flags, and an overall analysis.
6. The result is cached, logged to the scan_events ledger, and returned to the client.

For the full scoring algorithm (Quote Shield, Red-Line Crash, Weakest-Link Rule, etc.), see DOCUMENTATION.md.

---

Tech stack

Component	Technology
Backend API	Python 3 + FastAPI (deployed on Render)
Analysis engine	xAI grok-4-1-fast-reasoning (2M context)
Truth layer	Tavily API
Scraping	Firecrawl (primary), Jina.ai (fallback), Archive.today, direct fetch
Database / cache	Supabase (PostgreSQL)
Website	Static HTML/CSS/JS (Netlify)
Mobile	React Native + Expo
Extension	Chrome Manifest V3

---

Quick start

Clone the repo:

git clone https://github.com/goshtasb/Anie.git
cd Anie

You will need API keys for xAI, Tavily, Firecrawl, and a Supabase project.

1. Backend (aegis-backend/)

cd aegis-backend
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Create a .env file with your API keys
cp ../.env.example .env   # if you have one, otherwise create manually

uvicorn main:app --reload --port 8000

The API will be available at http://localhost:8000. Health check: GET /.

Run the internal logic tests:

python test_logic.py

Apply the database schema to your Supabase project using aegis-backend/supabase_schema.sql.

2. Website (anie-site/)

Fully static. Serve it any way you like:

cd anie-site
python3 -m http.server 8080

Then open http://localhost:8080. Before deploying, update the API base URL inside index.html to point at your backend.

3. Chrome extension (aegis-extension-mvp/)

1. Open chrome://extensions/.
2. Enable Developer mode (top right).
3. Click Load unpacked and select the aegis-extension-mvp/ folder.
4. Pin the extension, open any article, and click Scan.

Before publishing your own build, update the backend URL in background.js. See aegis-extension-mvp/README.md for architecture details.

4. Mobile app (anie-mobile/)

cd anie-mobile
npm install
npx expo prebuild
npx expo run:ios        # or: npx expo run:android

The app supports two entry points:

• MainDashboard — open the app and paste/enter a URL.
• ShareModal — share a link from Safari / Twitter / etc. via the native share sheet.

Bundle ID: com.axiom.anie.

---

API overview

Base URL (production): https://aegis-alpha.onrender.com

Method	Path	Purpose
GET	/	Health check
POST	/v1/scan	Analyze an article (URL or text)
POST	/v1/chat	Interrogation Mode — follow-up Q&A on a scan
POST	/v1/feedback	Submit 👍 / 👎 on a scan result
GET	/v1/stats	Aggregate stats for the live ticker
DELETE	/v1/cache/clear	Debug: clear the scan cache

Full request / response schemas are in DOCUMENTATION.md.

---

Full documentation

DOCUMENTATION.md covers:

• System architecture
• Backend module-by-module breakdown (main.py, engine.py, services.py, scraper.py)
• The 24 forensic markers across all four vectors
• Scoring algorithm, Quote Attribution Protocol, Red-Line rules
• Database schema for scan_cache, scan_events, guests
• Synapse directive history (V3.0 → V6.0)
• Deployment instructions for each component
• Full API reference

---

Contributing

Contributions, issues, and feature requests are welcome.

1. Fork the repo.
2. Create a feature branch: git checkout -b feat/your-idea.
3. Make your changes. For backend changes, run python test_logic.py.
4. Commit with a clear message describing why, not just what.
5. Open a pull request against main.

Please do not commit secrets, .env files, or anything containing API keys.

For larger changes (new vectors, scoring changes, new engines), open an issue first so we can align on approach.

---

License

Released under the MIT License — you are free to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the software, as long as the original copyright notice and license are included.

---

Contact

• Website: www.goanie.com
• Repo: github.com/goshtasb/Anie