🧠 Context IQ: The Cognitive Operating System for Organizations

Tagline: From remembering what teams said — to understanding WHY, predicting WHAT HAPPENS NEXT, and feeling HOW THEY WORK TOGETHER.

---

The Problem

In any collaborative setting, context is fragmented. Critical tasks, decisions, and discussions are scattered across Slack, Zoom transcripts, documents, and tickets. This "organizational amnesia" leads to inefficiency, duplicate work, and a lack of accountability. Answering "Why did we decide this?" is nearly impossible.

The Solution

Context IQ is a living, reasoning AI platform that ingests this fragmented data to build a dynamic organizational knowledge graph. It goes beyond simple search to provide:

• Privacy-First RAG: Answer "what" questions with evidence, featuring role-based PII redaction and a full audit trail for governance.
• Predictive Analysis: Answer "what if" by using historical patterns to predict the impact (cost, timeline, sentiment, risk) of new decisions.
• Counterfactual Simulation: Explore alternate histories to understand the tradeoffs of past choices.
• Emotional & Trust Insights: Understand your team's "Context DNA" through live sentiment heatmaps and a novel Trust Graph that maps implicit agreement and influence.

---

Core Features

1. RAG with Privacy-by-Design

• Redacted View (Default): Standard users (analyst) see PII-redacted text (e.g., "...decision from [PERSON]...").
• Admin View (Privileged): admin users see the original, raw text.
• Audit Log: All admin-level access to raw data is logged as a SECURITY event in an immutable audit_trail.log.

2. Emotional Intelligence Dashboard

• Sentiment Heatmap: Live dashboard (powered by GET /api/v1/sentiment/overview) showing overall positive/neutral/negative balance.
• Channel Breakdown: Identifies sentiment hotspots by channel (e.g., #engineering vs. #product).
• Sentiment Timeline: Tracks sentiment over time (7, 30, 90 days).

3. Trust & Agreement Graph

• Graph Visualization: Renders a dynamic react-force-graph of users.
• Trust Links: Creates [:AGREES_WITH] relationships in Neo4j by analyzing messages for consensus keywords (e.g., "I agree," "good point").
• Insight: Visually identifies key influencers and disconnected teams based on agreement, not just org structure.

4. Decision Impact Predictor

• AI Foresight: A user can type a new decision (e.g., "Delay the launch").
• Impact Score: The system uses ML models (scikit-learn) and graph patterns to predict the Impact (High/Medium/Low), Confidence, and key metrics like Timeline, Cost (Simulated), and Sentiment Shift.
• Recommendations: Provides actionable advice (e.g., High risk detected, High cascade probability).

5. Counterfactual Simulator

• "What-If" Engine: A user can select any past decision and propose an alternative.
• Monte Carlo Simulation: The backend runs a simulation based on historical data to model the alternate outcome.
• Comparison UI: The frontend displays a clear side-by-side comparison (Original vs. Simulated) and provides a final recommendation (e.g., "✅ Recommend Alternate Decision" or "❌ Original Decision Preferred").

---

Tech Stack

Monorepo Structure

• /backend: FastAPI, Python, spaCy, Transformers, Neo4j/Milvus clients
• /frontend: React, Vite, TypeScript, react-force-graph-2d
• /ml: Core extraction scripts (extractor.py)
• /data: Sample data for ingestion
• /infra: docker-compose.yml for orchestration

Key Technologies

• Backend: FastAPI
• Frontend: React + Vite
• Knowledge Graph: Neo4j (for structured facts, relationships, and context)
• Vector Database: Milvus (for semantic search on redacted text)
• AI / ML:
  • spaCy: NER, PII Detection, Entity Extraction
  • sentence-transformers: Semantic Embeddings
  • transformers: LLM for RAG (e.g., FLAN-T5)
  • vaderSentiment: Sentiment Analysis
  • scikit-learn: Predictive Modeling
• Orchestration: Docker Compose

---

🚀 How to Run

Prerequisites

• Docker
• Docker Compose
• Node.js & npm (for frontend development)

1. Build & Start All Services

This command will start the backend, Neo4j, Milvus, and all other services in one go.

# From the /infra directory
docker-compose up -d --build