lazycook 1.3.0

LazyCook is an autonomous multi-agent conversational assistant designed to intelligently process user queries, manage documents, store conversations, and maintain iterative AI reasoning loops. It uses Gemini 2.5 Flash model with a four-agent architecture for high-quality responses and continuous learning.

🧠 LazyCook — Multi-Agent AI Assistant

Version: 1.3.0

Author: Hitarth Trivedi and Harsh Bhatt

Language: Python 3.10+

Powered by: Google Gemini API (Select any supported model)

---

📘 Overview

LazyCook is an autonomous multi-agent conversational assistant that intelligently processes user queries, manages documents, tracks tasks, and maintains context across sessions.
It leverages Google Gemini API and a four-agent architecture — Generator, Analyzer, Optimizer, and Validator — to deliver accurate, coherent, and high-quality responses through iterative reasoning.

LazyCook is ideal for developers, researchers, and productivity users who want an intelligent assistant capable of local storage, contextual memory, and automated document analysis — all inside a single Python app.

---

⚙️ Core Features

Feature	Description
🎛️ Model Selection	Choose any Gemini model of your choice as input (e.g., gemma-3-27b-it, gemini-2.5-flash) to power the assistant.
🤖 Multi-Agent System	Four specialized AI agents collaborate to generate, analyze, optimize, and validate every response.
🧠 Smart Context Management	Maintains conversation context from current and past sessions (default 70, configurable).
📄 Document Processing	Supports .pdf, .docx, .txt, .md, and .csv files (default 50 docs, configurable).
📊 Smart Visualization	Automatically generates insightful graphs and plots for data-heavy queries.
🎯 Intelligent Query Routing	Adjusts API usage automatically based on query complexity (Simple / Medium / Complex).
📊 Quality Metrics	Evaluates completeness, accuracy, and polish with weighted scoring.
💾 Persistent Storage	Stores all data — chats, tasks, documents — as JSON files for easy access.
📦 Export Options	Export past conversations in .txt, .md, or .json formats.
🔧 Maintenance Tools	Clear old chats, documents, and caches for optimal performance.
🧮 Real-Time Logging	Colorful terminal output and progress visualization using rich.

---

🧩 System Architecture

User Query
    ↓
AutonomousMultiAgentAssistant
├── TextFileManager         # File & context storage
├── MultiAgentSystem        # Core orchestrator
│     ├── Generator Agent   # Draft creation
│     ├── Analyzer Agent    # Error detection
│     ├── Optimizer Agent   # Refinement
│     └── Validator Agent   # Final verification
├── QueryComplexityAnalyzer # Routing logic
└── QualityMetrics          # Evaluation engine

---

🚀 Installation & Setup

1. Prerequisites

• Python 3.10+
• Google Gemini API key
• Internet connection

2. Install Dependencies

pip install google-genai rich PyPDF2 python-docx matplotlib pandas seaborn numpy

3. Run the Application

export GEMINI_API_KEY="your-api-key"
python -m lazycook

---

💬 Example Usage

import lazycook
import asyncio

# You can select any model of your choice as input along with the API key
config = lazycook.create_assistant("your-api-key", conversation_limit=9, document_limit=1, model_name="gemma-3-27b-it")

# Run CLI
asyncio.run(config.run_cli())

---

📂 Directory Structure

project/
├── lazycook.py                    # Main application
├── multi_agent_data/              # Stored data
│   ├── conversations.json
│   ├── tasks.json
│   ├── documents.json
│   └── new_convo.json
├── exported_chats/                # Exported chat files
└── multi_agent_assistant.log      # Application logs

---

🧠 Multi-Agent Roles

Agent	Role	Purpose
Generator	Creative	Drafts the initial solution using context and user query.
Analyzer	Critical	Detects logical or factual errors and missing details.
Optimizer	Refinement	Enhances clarity, formatting, and completeness.
Validator	Assurance	Final accuracy and factual verification.

---

⚡ Quality Scoring System

Metric	Weight	Description
Completeness	40%	Ensures all query points are addressed.
Accuracy	40%	Checks factual and logical correctness.
Length	20%	Evaluates concise vs. detailed balance.
Structure & Polish	—	Considers clarity, readability, and formatting.

Tiers:

• 🔥 Excellent: ≥ 0.95
• ✅ Very Good: 0.90–0.94
• 📈 Good: 0.85–0.89
• ⚠️ Acceptable: 0.75–0.84

---

🛠 Maintenance Commands

Command	Function
maintenance	Access cleanup, reset tools, and upload documents (via Manage Documents)
docs	View current uploaded documents
download	Export chat history
quality	View session quality metrics
stats	View performance statistics
context	Preview current conversation context
quit	Exit application safely

---

🧾 Logging Example

2025-10-31 13:22:51 - INFO - Query classified as: complex
2025-10-31 13:22:52 - INFO - Iteration 1: Objective=0.913, Subjective=0.867, Combined=0.890
2025-10-31 13:22:52 - INFO - ✓ Quality threshold met: 0.890 >= 0.880

---

🧱 Future Enhancements

• VERSION THAT CAN BE DOWNLOADED AND USED WITHOUT API-KEY(using gemma2.0)
• Multi-model fusion (Gemini + LLaMA)
• Long-term vector memory
• Web-based dashboard and analytics
• Speech-to-text and voice integration

---

🧰 Troubleshooting

Issue	Possible Fix
API Connection Failed	Verify GEMINI_API_KEY and internet access.
Context Not Loading	Check user ID consistency and clear cache.
Document Upload Error	Ensure file < 5MB and supported format.
Low Quality Scores	Add context or documents for deeper responses.
Slow Responses	Reduce context size or clean old conversations.

---

📜 License

Copyright (c) 2025 Harsh Bhatt, Hitarth Trivedi. All Rights Reserved.

This software and associated documentation files (the "Software") may not be copied, modified, merged, published, distributed, sublicensed, and/or sold without explicit written permission from the copyright holder.

---

💡 Credits

• AI Framework: Google Gemini API
• Terminal UI: rich
• PDF Handling: PyPDF2
• DOCX Handling: python-docx
• Visualization: matplotlib, seaborn, pandas
• Developer: Hitarth Trivedi, Harsh Bhatt

Let it cook 🔥