Stack 2.9 - AI Agent Framework 🔧

Stack 2.9 is a high-performance AI Agent Framework built around a fine-tuned Qwen2.5-Coder-1.5B model. It is designed for autonomous software engineering, multi-agent orchestration, and complex tool-integrated workflows.

🌟 Key Highlights

• 57 Production-Ready Tools: From deep code intelligence (Grep, Glob, FileEdit) to agent orchestration (Spawn, TeamCreate, PlanMode).
• Cognitive Enhancements: Integrated Emotional Intelligence, Knowledge Graph RAG, and Advanced NLP pipelines.
• MCP Support: Native integration with the Model Context Protocol for standardized tool and resource access.
• Massive Context: 128K token window for processing entire repositories.
• Fine-tuned for Accuracy: Optimized on Stack Overflow Q&A for high-precision code generation and debugging.

🛠️ Architecture Overview

The framework is divided into three core layers:

1. The Brain: A LoRA-finetuned Qwen2.5-Coder-1.5B model.
2. The Toolbelt: A centralized ToolRegistry managing 57+ tools across 13 categories.
3. The Enhancements: Modular plugins for sentiment analysis, relationship mapping, and static code auditing.

🚀 Getting Started

Installation

git clone https://github.com/my-ai-stack/stack-2.9
cd stack-2.9
pip install -r requirements/requirements.txt

Quick Usage

from src.tools import get_registry
from transformers import AutoModelForCausalLM, AutoTokenizer

# 1. Load the brain
model = AutoModelForCausalLM.from_pretrained("my-ai-stack/Stack-2-9-finetuned")
tokenizer = AutoTokenizer.from_pretrained("my-ai-stack/Stack-2-9-finetuned")

# 2. Access the tools
registry = get_registry()
print(f"Available tools: {len(registry.list())}")

📂 Project Structure

• src/tools/: Implementation of the 57 agent tools.
• src/enhancements/: Cognitive modules (EI, Knowledge Graph, NLP).
• src/mcp/: Model Context Protocol client and server.
• src/voice/: TypeScript client for voice synthesis and cloning.
• stack/voice/: Python voice server (FastAPI) and integration tools.
• stack/eval/: Benchmark suites and evaluation results.
• stack/training/: Fine-tuning pipelines and dataset scripts.

🎙️ Voice Integration

Stack 2.9 includes a voice synthesis and cloning system that allows the agent to communicate via audio.

Setting up the Voice Server

1. Navigate to the voice directory:

   cd stack/voice

2. Install dependencies:

   pip install fastapi uvicorn requests

3. Start the voice server:

   python voice_server.py

   The server will start on http://localhost:8000.

Running a Voice Demo

You can run the provided demo script to see the voice integration in action:

python samples/demo_voice.py

This script simulates a voice command, processes it through the StackAgent, and generates an audio response.

📄 Documentation

For detailed information, see the Model Card and API Reference.

---

Built by Walid Sobhi