On-device autonomous research and content writing with open-source LLMs and CrewAI.
This repository provides a fully local, on-device framework for autonomous research, analysis, and content generation using open-source large language models (LLMs) and multi-agent AI workflows. Built with CrewAI for agent orchestration and Ollama for local LLM inference, the system enables offline, privacy-first, and cost-predictable AI pipelines without reliance on cloud APIs.
The architecture has been tested and validated using the following locally hosted open-source LLMs:
- Gemma 3 (12B) – used for efficient reasoning, research automation, and high-quality content drafting
- GPT-OSS (20B) – used for extracting insights, advanced analytical reasoning, synthesis, and structured long-form content
By running all models on-device, this project ensures data sovereignty, reproducibility, and secure AI workflows. Specialized AI agents collaborate to autonomously plan research tasks, collect and evaluate information, synthesize insights, and generate professional-grade written outputs. The modular design allows seamless model swapping, agent role customization, and workflow adaptation across domains.
Features
- Fully on-device execution via Ollama
- Multi-agent research and writing workflows with CrewAI
- Support for Gemma 3 (12B) and GPT-OSS (20B) models
- Offline, privacy-first, and reproducible pipelines
- Flexible design for research, reports, documentation, and editorial content
- Modular and extensible architecture for research automation
- Designed for reports, technical documentation, and editorial content
This repository is intended for developers, researchers, and AI practitioners seeking local-first, autonomous AI systems for research and content creation, with full control over models, infrastructure, and data.