🛡️ SRISHTI — AI Criminal & Paedophile Investigation Platform (ACPIA)

Government-Grade AI Criminal Investigation & Cybercrime Intelligence Platform
Built for Kerala Police, Cyberdome & International Cyber Crime Branch — Hac'KP 2026

---

🌟 Overview

SRISHTI (System for Real-time Investigation & Synthetic Threat Intelligence) is an enterprise-grade AI Criminal & Paedophile Investigation Assistant. It transforms raw digital evidence (images, audio voicemails, chat exports, syslogs) into structured, court-admissible criminal intelligence.

Upon evidence intake, SRISHTI orchestrates a 16-Agent Autonomous DAG Pipeline that extracts EXIF metadata, verifies SHA-256 chain-of-custody, detects grooming & extortion patterns, maps communication networks, reconstructs timelines, performs vector semantic search, and generates FRE 901 / Section 65B compliant legal reports.

---

🤖 16 Investigation Agents Matrix

Each agent is equipped with a designated Primary AI Model and Supporting Tool suite:

Agent #	Agent Name	Primary AI Model	Supporting Tools / Engines
01	Evidence Intake	No LLM (WebCrypto)	ExifTool, MinIO S3, SHA-256 Hashing
02	Content Analysis	Google Gemini Flash / GPT-4o Vision	PaddleOCR, Whisper AI, FFmpeg
03	Threat Identification	Google Gemini Flash / GPT-4o	Rule Engine (Grooming, Blackmail, Extortion)
04	Context Extraction	Google Gemini Flash / Florence-2	Florence-2, Visual Landmark & GPS Engine
05	Activity Pattern	Google Gemini Flash / NetworkX	NetworkX, Python Centrality Engine
06	Metadata Mapping	No LLM	ExifTool, Apache Tika Metadata Parser
07	Synthetic Detection	Deepfake Forensic Models	PRNU Analysis, Audio Artifact Libraries
08	Timeline Reconstruction	Google Gemini Flash / Neo4j	Neo4j Graph Database, Temporal Chronology
09	Intelligent Retrieval	OpenAI Embeddings / Qdrant	Qdrant Vector Search, Neo4j Graph RAG
10	Automated Reporting	Google Gemini Flash / Structured Outputs	FRE 901 Certificate, Section 65B PDF Engine
11	Risk Assessment	Google Gemini Flash / Custom Engine	Victim Threat & Danger Indexing
12	Intelligence Fusion	Google Gemini Flash / Fusion Engine	Neo4j, Qdrant Unified Intelligence Graph
13	Hypothesis Generation	Google Gemini Flash / Graph RAG	Knowledge Graph Reasoning Engine
14	Verification	Google Gemini Flash / Rule Engine	Rule Engine, Anti-Hallucination Guardrails
15	Investigation Copilot	Google Gemini Flash / Streaming	Interactive Live AI Chat Assistant
16	Explainability & Legal	Google Gemini Flash / Legal AI	Judicial Audit Trail, Magistrate Review

---

🗄️ Multi-Database Architecture

Database	Technology	Purpose
Relational DB	PostgreSQL 16	Case management, users, evidence records, audit logs
Knowledge Graph	Neo4j 5	SUSPECT, VICTIM, DEVICE, LOCATION, EVIDENCE entity relationships
Vector DB	Qdrant	1536-dimensional semantic vector embeddings & RAG search
Evidence Storage	MinIO S3	Immutable binary evidence storage & pre-signed URLs
Cache & Realtime	Redis 7	Task queues, real-time agent execution state
Event Bus	RabbitMQ 3	Topic exchange event bus (srishti.agents.*, srishti.pipeline)

---

🚀 Key Features & Capabilities

1. Live Interactive 16-Agent 1-by-1 Tester:

   • Access /agents to click on any agent card and run a live execution test with your real Google Gemini API Key (gemini-flash-latest), observing latency timer (ms), token usage, and AI model output.

2. Live SRISHTI Copilot Chat:

   • Interactive right-side chat widget powered by Google Gemini API for real-time investigation Q&A with Inspector Jayesh.

3. Drag & Drop Evidence Intake:

   • Drag and drop evidence files at /evidence with real-time WebCrypto SHA-256 hashing, EXIF metadata parsing, and interactive Evidence Viewer Modal.

4. Real-time API Credentials Manager:

   • Live key validation page at /settings supporting Google Gemini API Key, HuggingFace Fine-Grained Token, and OpenAI API Key.

---

🛠️ Quick Start & Local Setup

Prerequisites

• Node.js: v20.0.0+
• pnpm: v9.0.0+

Installation

# Clone the repository
git clone https://github.com/mandalejayesh10-stack/ACPIA.git
cd ACPIA

# Install all workspace dependencies
pnpm install

Environment Configuration (.env)

Create or edit .env in the root directory:

GEMINI_API_KEY=your_google_gemini_api_key_here
HUGGINGFACE_API_KEY=your_huggingface_api_key_here
OPENAI_API_KEY=your_openai_api_key_here
DATABASE_URL=postgresql://acpia_admin:acpia_secure_password_2026@localhost:5432/acpia_db
NEO4J_URI=bolt://localhost:7687
QDRANT_URL=http://localhost:6333
MINIO_ENDPOINT=localhost

Run Web Application

# Build packages
pnpm --filter @acpia/ai-provider build

# Start Next.js Web App
pnpm --filter web dev

Open http://localhost:3000 in your browser!

---

🏛️ Legal & Court Admissibility

All SRISHTI evidence outputs comply strictly with:

• FRE 901 (Federal Rules of Evidence - Authenticating & Identifying Evidence)
• Section 65B of the Indian Evidence Act / Bharatiya Sakshya Adhiniyam (Certificate of Digital Evidence Authenticity)

---

📜 License

MIT License — Built for Hac'KP 2026 & Kerala Police Cyberdome.