Personal News & World-Events Intelligence Dashboard

A self-hosted, agent-driven research and bibliography system that consolidates news and public discourse across mainstream, regional, and social sources into a single intelligence engine — optimized for depth, objectivity, and second-order significance.

Complete specification and functional details are documented in news-intelligence-dashboard-prd.md.

---

Repository Structure

.
├── news-intelligence-dashboard-prd.md  # Comprehensive Product Requirements Document
├── README.md                           # Main repository overview (this file)
├── .gitignore                          # Git ignore rules for Python, DBs, and local data
└── geonewsdesk/                        # Core Python application workspace
    ├── README.md                       # Submodule progress log & feature guide
    ├── requirements.txt                # Python dependencies
    ├── config.py                       # System configuration & model role allocation
    ├── model_client.py                 # LM Studio OpenAI-compatible client & offline fallback
    ├── agents/                         # Autonomous agents (Retrieval, Extraction, Significance, Editor, Threading, Update)
    ├── checkpoints/                    # Working state checkpoint manager & MasterTaskTracker
    ├── db/                             # SQLAlchemy models, sessions, and seed scripts
    ├── orchestration/                  # Daily loop & Weekly summary orchestration engines
    ├── tests/                          # Automated unit and integration test suite
    └── utils/                          # Security & prompt-injection safeguards

---

Key Features

• Multi-Source Retrieval: Pluggable source adapters for RSS wire/press feeds and social feeds (Reddit JSON API).
• Rubric-Driven Significance Filter: Scores stories across 5 dimensions (Scale, Causal Reach, Rarity, Persistence, Standing Interest), leveraging an infrastructure Dependency Graph.
• Content Security & Prompt Safeguards: Enforces strict data/instruction separation and logs prompt-injection anomalies.
• Media Authenticity Verification: Evaluates media using multi-check scoring (EXIF, reverse search, landmark matching, synthetic media detection).
• Editor Agent & Narrative Structuring: Categorizes research into Main Line, Branching, and Alternate narratives with 5 selectable Social Source Inclusion strategies.
• Latent Story Threading: Discovers non-obvious cross-regional/cross-industry connections across stories.
• Continuity & Evolution Tracking: Multi-signal story matching with 3-zone confidence routing (High, Middle, Low).
• Daily & Weekly Consolidation Loops: Automated daily loops and 7-day weekly synthesis passes with plain-file JSON backups under /data/.

---

Quick Start Guide

1. Setup Virtual Environment

Navigate to the geonewsdesk directory and set up the isolated Python virtual environment:

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

2. Initialize Database & Seed Infrastructure Graph

.venv/bin/python -m db.session
.venv/bin/python -m db.seed_graph

3. Run Daily & Weekly Consolidation Loops

# Run Daily Loop (Retrieval -> Security -> Verification -> Scoring -> Editor Synthesis -> Backup export)
.venv/bin/python -m orchestration.daily_loop

# Run Weekly Loop (7-day synthesis pass -> Story Threading -> Evolution Tracking -> Weekly Summary)
.venv/bin/python -m orchestration.weekly_loop

4. Run Automated Test Suite

.venv/bin/python -m unittest discover -s tests -p "test_*.py"

---

Configuration & LLM Orchestration

• LM Studio Integration: By default, model_client.py connects to LM Studio on http://localhost:1234/v1.
• Offline / Mock Fallback Mode: If LM Studio is not running, the system automatically falls back to an offline rule-based mock generator so testing and demonstration run seamlessly without errors. You can also explicitly set NEWSDESK_MOCK_LLM=1.

---

License

Personal Research & Bibliography Intelligence Project.