ETHOSIntel

An advanced multi-agent system for cumulative intelligence analysis, built on the ASI Alliance stack and integrated with Flow and Filecoin for decentralized operation and verifiable evidence storage.

https://youtu.be/kPgYt1Xs4CM   •   [github.com/Ash-394/ethIndia2025]

---

Project Summary

ETHOSIntel is a decentralized, multi-agent system designed to assist human analysts in complex investigations. It addresses the critical challenge of synthesizing a coherent narrative from a continuous stream of unstructured, text-based intelligence. The platform leverages a stateful, cumulative architecture, allowing its understanding of a case to grow more sophisticated with each new piece of evidence.

Core Innovation: Hybrid Reasoning

The central innovation is a hybrid reasoning architecture that combines a formal logic engine with a powerful LLM. This creates a final output that is both creatively insightful and logically sound.

1. AI for Translation: An initial agent uses ASI:One to handle the "fuzzy" task of translating unstructured English reports into a structured, formal MeTTa script of facts.
2. MeTTa for Logic: A specialist agent then uses this script to build a cumulative knowledge base, executing a ruleset and complex queries to produce verifiable, logical insights.
3. AI for Synthesis: Finally, the system feeds the raw evidence and the MeTTa findings into a second ASI:One prompt to generate a high-level, intuitive theory of the case.

System Architecture

The system is a decentralized, two-agent collaboration:

• Collector Agent: Serves as the system's public interface. It ingests tips via a live webhook, uses ASI:One to create a MeTTa script, archives the evidence to Filecoin via Lighthouse, and forwards the complete MeTTa script to the Detective.

• Detective Agent: The analytical core. It receives the MeTTa script, loads it into its MeTTa reasoning space to generate logical facts, and then uses ASI:One to synthesize all available information. The final report is then pushed to clients via a WebSocket and a summary is logged on the Flow blockchain.

Web3 Integration

This project is built as a Web3-native application, utilizing decentralized technologies for security, persistence, and transparency.

Decentralized Storage: Filecoin & Lighthouse

To ensure the integrity and persistence of evidence, our Collector Agent uses Lighthouse to store all incoming tips and evidence files on the Filecoin Calibration Testnet. This provides a meaningful, decentralized storage solution for critical case data, protecting it from tampering and ensuring long-term availability.

On-chain Verification: Flow EVM Testnet

The system is built on Flow and interacts with the blockchain to provide a transparent, on-chain record of investigative summaries.

• Network: Flow EVM Testnet
• Deployed Contract Address: [0x4BdE24c005DE13a5d8B47CEb18D6485E17A444Fb] On Flowscan:
  • (Note: The contract was deployed in transaction 0x8c601859c86547d38bce138ac49bb47accd96bfe4c32d6b4d289795d0a1fda89)

Technology Stack

Category	Technology
Agent Framework	Fetch.ai uAgents
AI & LLM	ASI:One
Logical Reasoning	MeTTa (hyperon-experimental)
Smart Contracts	Flow EVM Testnet
Decentral Storage	Filecoin (via Lighthouse)
API & Webhooks	FastAPI, Uvicorn, WebSockets

Getting Started

1. Prerequisites

• Python 3.10+
• A virtual environment tool (venv)

2. Installation & Setup

# Clone the repository
git clone [github.com/Ash-394/ethIndia2025]
cd [agents/fetch-agents]

# Create and activate a virtual environment
python3 -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

3. Configuration

Create a file named .env in the root directory and add your ASI Alliance API key:

ASI_API_KEY="your_key_here"

4. Execution

The system requires three separate terminals to run.

• Terminal 1: Start the Collector Agent

  python3 evidence_collector.py

• Terminal 2: Start the Detective Agent

  python3 detective_agent.py

• Terminal 3: Interact with the System Open the index.html file in a browser, or use curl to send a tip to the live webhook.

  curl -X POST -H "Content-Type: application/json" \
  -d '{"case_id": "CASE-101", "text": "A new witness saw a blue car leaving the scene."}' \
  [http://127.0.0.1:8003/submit_tip](http://127.0.0.1:8003/submit_tip)