Agent-Fusion

Agent-Fusion is a powerful local RAG (Retrieve, Augment, Generate) semantic search engine designed to provide AI agents with direct access to your code and documentation, including formats such as Markdown, Word, and PDF. Developers can use this tool to query their codebases without the risk of hallucinations, enhancing productivity and efficiency in code exploration and documentation retrieval. With its lightweight embedding model and optional multi-agent task orchestration, Agent-Fusion streamlines the interaction between AI agents and your development resources.

Once connected to the Agent-Fusion server, you can interact with it by sending natural language queries that reference your code and documentation. For optimal results, focus on specific questions about your codebase or documentation content. You can also leverage the multi-agent task orchestration feature to execute complex queries that require the collaboration of multiple agents. Keep your queries clear and concise to maximize the relevance of the results returned.

Prerequisites

• Java 11 or higher (for running the JAR file)
• Optional: Docker (for containerized deployment)

Option A: Quick Start with JAR

You can quickly start using Agent-Fusion by downloading the JAR file from the repository. Use the following command in your terminal:

java -jar path/to/Agent-Fusion.jar

Option B: Global Install Alternative

If you prefer to run it in a global context, you can download the JAR file and place it in a directory included in your system’s PATH. Then run the JAR file as mentioned above.

To maximize the performance of Agent-Fusion, consider configuring environment variables such as JAVA_OPTS to allocate more memory if you're working with large codebases. Make sure to check the documentation for any specific configuration settings related to your embedding model. A common gotcha is ensuring that the required file formats are properly indexed before querying, so verify that your documentation is accessible to the server.