KIRMADA(AZTEC

README.md

@@ -1,43 +1,213 @@
 # Agentic Ethereum Hackathon India
 
-# 🛠 Project Title - [Team Name]
+# 🛠 Project Title - KIRMADA
 
 Welcome to our submission for the *Agentic Ethereum Hackathon* by Reskilll & Geodework! This repository includes our project code, documentation, and related assets.
 
 ---
 
 ## 📌 Problem Statement
 
-We addressed the challenge: *“[Problem Statement Title]”*  
+We addressed the challenge: 
+🧩 Problem Statement for Aztec
+"Decentralized, Verifiable AI Model Training Without Compromising Data Privacy"
+🔍 Context
+In traditional machine learning workflows, training AI models often involves centralized infrastructure and massive datasets—frequently requiring sensitive user data to be shared with third-party services. This approach raises serious data privacy, trust, and verifiability concerns:
+
+How do we verify if a model was trained as claimed?
+
+Can we incentivize honest participation in model training without central enforcement?
+
+Is it possible to prove learning without revealing the training data?
+
+❗The Problem
+There exists no trustless, decentralized system that allows users to train AI models on private data and prove their contribution without exposing the data itself.
+
+Current solutions either:
+
+Require centralized authorities (e.g., hosted model competitions),
+
+Cannot prove the honesty of local training,
+
+Or compromise data sovereignty and privacy.
+
+
 Brief description of the challenge and why it matters.
 
 ---
 
 ## 💡 Our Solution
 
-*Project Name:* [Your Project Name]  
-A short pitch of your solution — what you built, who it’s for, and why it’s impactful.
+# Product Requirements Document (PRD)
+
+## Project Name: Aztec
+
+**Subtitle:** The Verifiable Learning Protocol — Autonomous AI Agents Coordinating Trustless, Incentivized Training on Ethereum
+
+---
+
+## 1. **Overview**
+
+Aztec is a decentralized coordination protocol that powers autonomous AI agents to participate in the training and evolution of open-source AI models. Agents locally fine-tune models, generate zero-knowledge proofs to verify their training, upload updates to decentralized storage, and receive on-chain micropayments based on verifiable contributions.
+
+Unlike traditional federated learning systems, Aztec allows independent agents to autonomously:
+
+- Discover training opportunities
+- Make economic decisions
+- Learn over time via agent memory
+- Submit updates to a public, auditable protocol layer
+
+Aztec is deployed on Ethereum L2 (Optimism/Scroll) and leverages IPFS, ZK circuits, and ERC-20 micropayment contracts.
+
+---
+
+## 2. **Goals & Objectives**
+
+### 🎯 Primary Goals:
+
+- Launch a decentralized agentic learning protocol
+- Enable autonomous agents to verifiably contribute model updates
+- Incentivize meaningful training through crypto-based rewards
+- Maintain an immutable, transparent model evolution graph
+
+### 🧠 Secondary Goals:
+
+- Build foundational infrastructure for multi-agent decentralized AI
+- Implement memory and strategy modules in the agents
+- Support forkable, reputation-weighted model evolution
+
+---
+
+## 3. **Target Users**
+
+- Developers training LLMs with private/local datasets
+- Agent operators who run Aztec agents for passive income
+- Coordinators or DAOs managing open-source model rounds
+- Researchers exploring decentralized agentic AI systems
+
+---
+
+## 4. **User Stories**
+
+1. *As an agent operator*, I want my AI agent to detect high-reward rounds, train, prove, and earn automatically.
+2. *As a protocol coordinator*, I want to launch rounds and get high-quality model updates verifiably.
+3. *As a DAO*, I want to fork, merge, and vote on model versions based on proof quality and contributor reputation.
+4. *As an observer*, I want to explore the full evolution of any public model in Aztec's ecosystem.
+
+---
+
+## 5. **Core Features**
+
+### 🤖 Autonomous Aztec Agent
+
+- Runs on local machine or node
+- Reads training round intent and evaluates expected reward
+- Trains a base model (LoRA, PEFT, QLoRA, etc.)
+- Generates model diff and zero-knowledge proof
+- Uploads to IPFS, submits to contract, tracks rewards
+- Stores **long-term memory** of previous proofs, rewards, model types, round quality
+
+### 🧠 Agent Memory Module
+
+- Stores history of:
+    - Proof success/failure
+    - Token payouts
+    - Training context (model, dataset, diff delta)
+- Uses memory to avoid bad rounds, retry smarter
+- Agents learn to optimize strategy over time (e.g. pick rounds with better validation scores)
+
+### 🔐 ZK Proof Generation
+
+- Circom or Noir-based circuit for training validity
+- Proves: honest training, valid diff, dataset constraints
+- Integrated with verifier contract on Ethereum L2
+
+### 🧾 On-Chain Contracts (Ethereum L2)
+
+- **Orchestrator:** Starts rounds, manages submissions, coordinates payouts
+- **Verifier:** Accepts proof & IPFS hash, validates via circuit
+- **ERC-20 Treasury:** Rewards contributors with micropayments
+
+### 🌐 IPFS + Forkable Model Graph
+
+- All model diffs stored on IPFS with metadata
+- Agents sign each submission → models form a versioned graph
+- Forkable by contributors and DAOs ("Aztec/BERT-hindi-v1.2")
+
+### 🖥 Frontend Interface
+
+- View live training rounds
+- Track agent submissions, round history, and model forks
+- Leaderboard of contributors with proof rate + reputation
+- DAO governance modules (stretch)
+
+---
+
+## 6. **System Architecture: Agent-Centric Design**
+
+**Agent Layer**:
+
+- Autonomous Python-based agent
+- Local GPU/TPU for training
+- Memory cache (JSON/SQLite/VectorDB)
+- Event loop listens for new rounds, evaluates reward curves
+
+**Proof Layer**:
+
+- Noir/Circom-based training proof circuits
+- Fast prover integrated with agent
+- ZK-SNARK verifier deployed to smart contract
+
+**Contract Layer**:
+
+- Orchestrator (Solidity, Hardhat)
+- Verifier (ZK circuit-generated)
+- Payment Vault (ERC-20)
+
+**Storage Layer**:
+
+- IPFS (for diffs, metadata)
+- Optional Filecoin backend
+
+**Aggregation + Governance**:
+
+- Off-chain aggregator (Node.js/Python)
+- IPFS hash of aggregated model
+- DAO voting + merge governance (stub or full)
+
 
 ---
 
 ## 🧱 Tech Stack
 
-- 🖥 Frontend: [React / Next.js / etc.]
-- ⚙ Backend: [Node.js / Python / etc.]
-- 🧠 AI: [Llama 3 / LangChain / OpenAI / etc.]
-- 🔗 Blockchain: [Ethereum / Solidity / Foundry / etc.]
-- 🔍 DB/Storage: [IPFS / PostgreSQL / Firebase / etc.]
-- 🚀 Hosting: [Vercel / Netlify / Render / etc.]
+Layer	Tooling
+Agent	Python, PyTorch, LoRA, SQLite
+Proof	Noir, Circom, snarkjs, zk-SNARK
+Contracts	Solidity, Hardhat, Optimism/Scroll
+Backend	Node (Hono) or Python (FastAPI)
+Frontend	Next.js, Tailwind, wagmi, ShadCN
+Storage	IPFS + Infura or web3.storage
+DevOps	Docker, GitHub Actions, Render
+DB/Cache	PostgreSQL, Redis
 
 ---
 
 ## 📽 Demo
 
-- 🎥 *Video Link*: [YouTube/Drive Link]  
+- 🎥 *Video Link*:
+- 1:Frontend: https://drive.google.com/file/d/1syox6r6V8NNKn9yrrFMviiCY1_1kkBMu/view?usp=sharing
+- 2.Backend: https://drive.google.com/file/d/1Xbe6IDRmkrMmk5f_uNQVA404Hyr75cV4/view?usp=sharing
+- 
 - 🖥 *Live App (if available)*: [URL]
 
 ---
 
+# Notion overview
+https://www.notion.so/22ea091db562802181afe18184eb6747
+https://www.notion.so/22ea091db56280c1af50d5246a9d81b3
+https://www.notion.so/22ea091db56280358b69de2705a45d62
+https://www.notion.so/22ea091db56280d696dffb0169e8f240
+
 ## 📂 Repository Structure
 
 ```bash