abdullahed/av-k.1

AVK.1 Agentic Brain

AVK.1 is a decentralized agentic system that ensembles Kimi-K2.6 and Qwen3.6-27B into a unified, high-performance intelligence. It is designed to learn and synthesize information in real-time from four major knowledge vectors:

1. Wikipedia: For factual and historical summaries.
2. YouTube: For deep-dives into video content and transcripts.
3. GitHub: For technical research and code analysis.
4. Twitter: For real-time news, sentiment, and trends.

Architecture

AVK.1 operates using a Collaborative Ensemble strategy:

• Planner (Qwen): Analyzes user requests, identifies knowledge gaps, and selects appropriate tools.
• Synthesizer (Kimi): Aggregates tool outputs and the planner's insights to generate a premium, professional response.

Usage

This model is designed to be deployed as a Hugging Face Inference Endpoint.

Local Setup

1. Clone this repository.
2. Install dependencies: pip install -r requirements.txt.
3. Set your HF_TOKEN environment variable.
4. Run python brain.py.

Capabilities

• Multi-source RAG (Retrieval Augmented Generation).
• Agentic tool orchestration.
• Real-time data ingestion.