Repolode
Submit
labsai

labsai/EDDI

Config-driven middleware that lets you build multi-agent conversational AI from JSON, with 19+ LLM providers, MCP and A2A support, RAG and persistent memory.

View on GitHubWebsite
Stars
389
Forks
131
Open issues
84
Last commit
1 hour ago

What is it?

EDDI (Enhanced Dialog Driven Interface) is a multi-agent orchestration middleware for conversational AI. Agent behavior is defined in versioned JSON configurations rather than application code, and the engine coordinates users, AI agents and business systems through routing rules, persistent memory and API calls. It is written in Java 25 on Quarkus and ships as a Docker image that includes a Manager dashboard and a standalone chat UI.

The project supports group conversations in which several agents discuss a topic in defined phases, along with routing, human approval gates and scheduled or heartbeat-triggered runs. On the model side it connects to OpenAI, Anthropic, Gemini, Azure OpenAI, Bedrock, Ollama and other OpenAI-compatible endpoints, and includes RAG with several embedding providers and vector stores such as pgvector, Chroma, MongoDB Atlas and Qdrant. Data can be stored in MongoDB or PostgreSQL.

EDDI exposes MCP and A2A interfaces, an OpenAI-compatible Chat Completions endpoint, Slack integration and OpenAPI-based tool generation. It also includes an encrypted secrets vault, an audit ledger, and documented features for GDPR and HIPAA-related deployments.

Key features

  • 01Agents are defined as versioned JSON configs, with no application code required.
  • 02Group conversations with multiple agents in defined discussion phases, including voting and negotiation.
  • 03Connects to 19+ LLM providers, including cloud APIs and self-hosted models through Ollama.
  • 04Supports MCP server and client, A2A, and an OpenAI-compatible API.
  • 05RAG with multiple embedding providers and vector stores, plus persistent user memory.
  • 06Human-in-the-loop approvals for conversation turns and individual tool calls.

Use cases

  • Building customer-facing or internal chatbots that call business APIs without writing custom orchestration code.
  • Running multi-agent review or debate workflows, such as peer review of a draft by several specialized agents.
  • Deploying self-hosted assistants on Docker or Kubernetes with MongoDB or PostgreSQL storage.
  • Exposing existing agents to Open WebUI or other OpenAI-compatible clients.

Quick start

  1. 1

    Run the one-command installer for Linux, macOS or WSL2 (runs a remote script, so no command is given)

  2. 2

    Or pull the Docker image from Docker Hub

    docker pull labsai/eddi
  3. 3

    Start EDDI with Docker Compose (default EDDI and MongoDB)

    docker compose up
  4. 4

    Open http://localhost:7070 to reach the Manager dashboard and create your first agent

Things to know

  • Requires Docker for the quick start path.
  • The Ollama local-LLM overlay needs Docker Compose 2.30.0 or newer, and the GPU overlay also needs the NVIDIA Container Toolkit.

In collections

Multi-agent orchestration frameworksOpen-source AI agent frameworks & appsMCP serversAI chatbots & chat clients

Topics

#ai-agents#multi-agent#llm#rag#mcp#chatbot#java#quarkus#a2a#ai-orchestration

FAQ

What is EDDI?
Config-driven middleware that lets you build multi-agent conversational AI from JSON, with 19+ LLM providers, MCP and A2A support, RAG and persistent memory. EDDI (Enhanced Dialog Driven Interface) is a multi-agent orchestration middleware for conversational AI. Agent behavior is defined in versioned JSON configurations rather than application code, and the engine coordinates users, AI agents and business systems through routing rules, persistent memory and API calls. It is written in Java 25 on Quarkus and ships as a Docker image that includes a Manager dashboard and a standalone chat UI.
Is EDDI free to use? What license does it have?
EDDI is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is EDDI still maintained?
Actively maintained. The last commit to EDDI was on Nov 7, 2026.
What can I use EDDI for?
Building customer-facing or internal chatbots that call business APIs without writing custom orchestration code. Running multi-agent review or debate workflows, such as peer review of a draft by several specialized agents. Deploying self-hosted assistants on Docker or Kubernetes with MongoDB or PostgreSQL storage. Exposing existing agents to Open WebUI or other OpenAI-compatible clients.
How do I install EDDI?
Run the one-command installer for Linux, macOS or WSL2 (runs a remote script, so no command is given) Or pull the Docker image from Docker Hub: docker pull labsai/eddi Start EDDI with Docker Compose (default EDDI and MongoDB): docker compose up Open http://localhost:7070 to reach the Manager dashboard and create your first agent
Where can I get EDDI?
The source code and installation instructions are on GitHub: https://github.com/labsai/EDDI

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects