Licensing

Check out the new architecture blog post for a full overview of Fred’s modular design. This page focuses on licensing boundaries and the distinction between open source and inner source components.

Fred is an open source initiative that fosters multi-agent experimentation and development.

While not originally intended for production use, Fred can be deployed and extended to serve real-world needs — especially in secure, regulated environments. Its modularity allows you to plug in your own agents and document processors, many of which are use-case specific and remain internal — what we call inner source.

This model powers contributions both within and outside of Thales Services Numériques.

Architecture Overview#

The diagrams below summarize the architecture and clarify what’s open vs inner source.

flowchart TD
    subgraph OpenSource["Open Source"]
        FE[Frontend - React]
        BE[Agentic Backend - fred]
        KFlow[Knowledge Flow Backend]
    end

    subgraph InnerSource["Inner Source"]
        KAST[Kast Agent]
        AGENT1[Business Agent 1]
        AGENT2[Business Agent 2]
        PROCESSOR1[Processor A]
        PROCESSOR2[Processor B]
        PROCESSORN[Processor N]
    end

    FE -->|REST / WS| BE
    BE -->|Fetch Data| KFlow
    FE -->|Direct Upload| KFlow
    KFlow -->|Triggers| PROCESSOR1
    KFlow -->|Triggers| PROCESSOR2
    KFlow -->|Triggers| PROCESSORN
    BE --> KAST

    classDef opensource fill:#d0e1ff,stroke:#333,stroke-width:1.5px;
    classDef innersource fill:#fff5cc,stroke:#333,stroke-width:1.5px;
    class FE,BE,KFlow opensource;
    class KAST,AGENT1,AGENT2,PROCESSOR1,PROCESSOR2,PROCESSORN innersource;

Open Source vs Inner Source#

Open Source (Apache 2.0)#

The following components are fully open and available under the fred-agent GitHub organization:

• Frontend: A React-based chat UI
• Fred Backend:
  • FastAPI service for session, streaming, and agent orchestration
  • LangGraph-based agents (e.g. Leader, Dominic)
• Knowledge Flow:
  • Document upload, ingestion, processing, and vector search APIs
  • Includes basic document processors (e.g. text extract, vectorize)
  • Easily extensible to custom pipelines

These form a complete, usable system out of the box — ideal for research, prototyping, and educational use.

Inner Source#

Certain agents and processors are specific to Thales or domain-specific deployments. These remain internal:

• Custom Agents

  • e.g. Kast Agent: specialized for the Kast Kubernetes distribution, with built-in assumptions and integrations for secure workloads
  • Other domain-specific agents for infrastructure, threat analysis, etc.

• Domain-Specific Document Processors

  • Pipelines that extract and transform sensitive or highly structured documents
  • Typically save data to custom SQL schemas or internal systems
  • Built with Python and integrated into knowledge-flow as plugins

What Are Document Processors?#

Document processors are asynchronous modules triggered by the Knowledge Flow backend. They handle:

• 🧾 Content extraction (PDFs, Excel, Word, etc.)
• 🧠 Semantic vectorization (for retrieval)
• 🗄 Structured storage (SQL/NoSQL)

They are:

• Plugin-based
• Triggered automatically by upload
• Defined per use case (e.g. legal docs, architecture specs, contracts)

The open source processors offer generic extraction and embedding. Inner source processors include richer, domain-tuned pipelines.

What Can You Do With Fred?#

With Open Source Only#

You can:

• Run a working local system with UI + agents + document ingestion
• Upload documents and chat with agents like Leader or Dominic
• Extend the system with your own:
  • 🧠 Agents or Leaders
  • 📄 Document processors
• Use it as a base for research, demos, or internal tooling

With Inner Source or Extensions#

You can:

• Deploy in secure environments (e.g. Kast/Kubernetes)
• Use preconfigured pipelines, processors, and agents
• Scale across teams or production clusters
• Integrate with internal services (S3, PostgreSQL, monitoring, etc.)

Fred gives you the foundation to build what you need — whether you’re a research team, product group, or infrastructure developer.

💡 Curious how to get started?
Check out the documentation or explore the source on GitHub.