FAME Knowledge Agent Gateway

FAME Knowledge Agent Gateway serves as a universal external plugin tailored for agents, integrating a graph-based knowledge network, project memories, governance for tool calls, FAME route feedback, and a visual workbench into a single external runtime. This solution is designed not just as a basic retrieval augment (RAG) or standard chat memory; it aims to enhance agents' operational efficiency in real-world projects by aligning goals, navigating structured knowledge, and ensuring controlled tool usage, while also retaining important evidence and lessons outside of the immediate context.

Key Features

• External Memory & Execution Governance: Functions as a plugin that maintains operational memory and execution control for agents.
• 3D/2D Knowledge Workbench: Provides a visual workspace that enhances navigation and editing of knowledge graphs.
• Project Memory Overlay: Ensures that project memories do not interfere with the core knowledge network, maintaining clarity and organization.
• Tool Gateway: Implements safeguards against potential misuse of tools by requiring an approval process before calls can be executed.
• Language Tree Center: Facilitates reasoning through abstractions and associations, going beyond simple information retrieval.
• Multimodal Asset Indexing: Keeps raw assets stored externally, only retaining previews and references within the graph.

Advantages

• Structured Knowledge Organization: Knowledge is structured around routes, abstraction layers, associations, and FAME parameters, offering more than just database retrieval.
• Scoped Contexts: Contexts are healthily segmented using identifiers like project_id, subject, route_id, and task_id, ensuring relevant information is encapsulated.
• Safe Tool Usage: All mutation-related tool interactions go through a comprehensive approval process, bolstering safety in operations.
• Verifiable Memory: The integration of testing, tool results, and user feedback contribute towards updating the status of knowledge routes, facilitating continuous improvement.
• Comprehensive Retention: Original traces and lessons learned are preserved rather than discarded, supporting learning and adaptation.
• Open-source Extensibility: The project includes both a complete Chinese edition and a seed English version, enabling further customization and expansion.

Usage Example

To initiate the project, execute the following commands:

npm run install:all
npm run generate:all
npm run dev

After setup, access the workbench via:

http://127.0.0.1:5178/

The initial interface of the Agent External Brain console presents various operational controls related to agent interaction, including safety features and memory management.

For detailed usage instructions, refer to the documentation: User Guide in English and 使用手册 for Chinese.