Cerno - Local Agentic Deep Research

Cerno: Local-first, open-source agentic AI for deep research. Transparent workflows, model flexibility, and full control over your data.

Comment

Inspiration

I wanted a research tool that doesn't treat AI as a black box. Instead, it should be transparent, local-first, and adaptable, letting researchers observe, verify, and control every step of the reasoning process while harnessing the full potential of open-source language models.

What it does

Cerno is an open-source agentic AI workspace that breaks down complex queries into transparent multi-step workflows. It supports both local and cloud models, provides complete visibility into each reasoning step, and generates verifiable artifacts like comprehensive reports, production-ready code, and structured datasets.

The platform leverages the power of open-source GPT OSS models through local inference, ensuring data privacy while delivering enterprise-grade capabilities. Beyond reasoning, Cerno can execute SQL queries, run live code, browse and scrape the web, and process structured data - making it a comprehensive research intelligence platform.

How we built it

Cerno uses a modular architecture that connects sophisticated agentic orchestration with a model-agnostic backend. Built on Django for robust backend services and Vite.js for a responsive frontend, it creates an elegant bridge between researchers and AI capabilities.

The platform integrates seamlessly with Ollama to serve cutting-edge open-source GPT OSS models locally, enabling researchers to harness models like Llama, Mistral, and CodeLlama without compromising data sovereignty or incurring cloud costs.

Challenges we ran into

  • Balancing complete transparency with intuitive usability across complex multi-agent workflows
  • Ensuring robust privacy and local-first execution while maintaining flexibility for diverse research needs
  • Optimizing performance and managing computational costs across different local and cloud LLM providers
  • Designing a clean, responsive interface for real-time workflow visualization and task lifecycle tracking
  • Managing memory and processing efficiency when running large open-source models locally

Accomplishments that we're proud of

  • Created a truly local-first research platform with zero vendor lock-in and complete data sovereignty
  • Achieved transparent, step-by-step workflow execution with full audit trails and reproducible outputs
  • Harnessed OpenAI's open-source GPT models to deliver sophisticated reasoning and intelligent task orchestration locally
  • Delivered verifiable, citation-ready research outputs that meet academic and professional standards by leveraging GPT OSS capabilities
  • Extended the intelligence of OpenAI's open-source models with practical tools: database querying, code execution, web interaction, and data processing
  • Democratized access to GPT-level intelligence by making OpenAI's open-source models accessible for complex research workflows

What we learned

Building trustworthy AI systems requires more than powerful models—it demands architectural commitment to transparency, reproducibility, and user control. We discovered that OpenAI's open-source GPT OSS models provide exceptional capabilities while offering superior privacy, cost advantages, and freedom from vendor dependency.

The importance of adaptive orchestration became clear when handling everything from simple queries to complex, multi-faceted research challenges that traditionally require specialized tools and expertise.

What's next for Cerno - Local Agentic Deep Research

  • Expanding support for emerging open-source foundation models and specialized research-focused models
  • Enhanced UI/UX with advanced workflow visualization and collaborative research features
  • Improved team collaboration tools for shared research projects and knowledge synthesis
  • Advanced cost optimization and performance monitoring across local and cloud model deployments
  • Extended integrations: spreadsheet automation, API orchestration, vector databases, and enterprise knowledge management systems
  • Enhanced security features and compliance frameworks for regulated research environments

Built With

Share this project:

Updates