refactor(a2a): use OpenAI by default (#75)

fiam · web-flow · commit ef34693e39bd · 2025-07-08T16:45:03.000+01:00
- Use OpenAI by default
- Allow using DMR with a dedicated compose file
- Support Offload with an additional compose file
diff --git a/README.md b/README.md
@@ -32,13 +32,13 @@ docker compose -f compose.yaml -f compose.openai.yaml up
 # Compose for Agents Demos - Classification
 
 | Demo | Agent System | Models | MCPs | project | compose |
-| ---- | ---- | ---- | ---- | ---- | ---- |
+| ---- | ---- | ---- | ---- | ---- | ---- |]<<<<<<< HEAD
+| [A2A](https://github.com/a2a-agents/agent2agent) Multi-Agent Fact Checker | Multi-Agent | OpenAI | duckduckgo | [./a2a](./a2a) | [compose.yaml](./a2a/compose.yaml) | 
 | [Agno](https://github.com/agno-agi/agno) agent that summarizes GitHub issues | Multi-Agent | qwen3(local) | github-official | [./agno](./agno) | [compose.yaml](./agno/compose.yaml) |
 | [Vercel AI-SDK](https://github.com/vercel/ai) Chat-UI for mixing MCPs and Model | Single Agent | llama3.2(local), qwen3(local) | wikipedia-mcp, brave, resend(email) | [./vercel](./vercel) | [compose.yaml](https://github.com/slimslenderslacks/scira-mcp-chat/blob/main/compose.yaml) |
 | [CrewAI](https://github.com/crewAIInc/crewAI) Marketing Strategy Agent | Multi-Agent | qwen3(local) | duckduckgo | [./crew-ai](./crew-ai) | [compose.yaml](https://github.com/docker/compose-agents-demo/blob/main/crew-ai/compose.yaml) |
 | [ADK](https://github.com/google/adk-python) Multi-Agent Fact Checker | Multi-Agent | gemma3-qat(local) | duckduckgo | [./adk](./adk) | [compose.yaml](./adk/compose.yaml) |
 | [ADK](https://github.com/google/adk-python) & [Cerebras](https://www.cerebras.ai/) Golang Experts | Multi-Agent | unsloth/qwen3-gguf:4B-UD-Q4_K_XL & ai/qwen2.5:latest (DMR local), llama-4-scout-17b-16e-instruct (Cerebras remote) |  | [./adk-cerebras](./adk-cerebras) | [compose.yml](./adk-cerebras/compose.yml) | 
-| [A2A](https://github.com/a2a-agents/agent2agent) Multi-Agent Fact Checker | Multi-Agent | gemma3(local) | duckduckgo | [./a2a](./a2a) | [compose.yaml](./a2a/compose.yaml) | 
 | [LangGraph](https://github.com/langchain-ai/langgraph) SQL Agent | Single Agent | qwen3(local) | postgres | [./langgraph](./langgraph) | [compose.yaml](./langgraph/compose.yaml) |
 | [Embabel](https://github.com/embabel/embabel-agent) Travel Agent | Multi-Agent | qwen3, Claude3.7, llama3.2, jimclark106/all-minilm:23M-F16 | brave, github-official, wikipedia-mcp, weather, google-maps, airbnb | [./embabel](./embabel) | [compose.yaml](https://github.com/embabel/travel-planner-agent/blob/main/compose.yaml) and [compose.dmr.yaml](https://github.com/embabel/travel-planner-agent/blob/main/compose.dmr.yaml) |
 | [Spring AI](https://spring.io/projects/spring-ai) Brave Search | Single Agent | none | brave | [./spring-ai](./spring-ai) | [compose.yaml](./spring-ai/compose.yaml) |
diff --git a/a2a/Dockerfile b/a2a/Dockerfile
@@ -24,7 +24,7 @@ if test -f /run/secrets/openai-api-key; then
 fi
 
 if test -n "\${OPENAI_API_KEY}"; then
-    echo "Using OpenAI with \${MODEL_NAME}"
+    echo "Using OpenAI with \${OPENAI_MODEL_NAME}"
     export LLM_AGENT_MODEL_PROVIDER=openai
     export LLM_AGENT_MODEL_NAME=\${OPENAI_MODEL_NAME}
 else
diff --git a/a2a/README.md b/a2a/README.md
@@ -1,10 +1,11 @@
 # 🧠 A2A Multi-Agent Fact Checker
 
-This project demonstrates a **collaborative multi-agent system** built with the **Agent2Agent SDK** ([A2A]),
-where a top-level Auditor agent coordinates the workflow to verify facts. The Critic agent gathers evidence
-via live internet searches using **DuckDuckGo** through the Model Context Protocol (**MCP**), while the Reviser
-agent analyzes and refines the conclusion using internal reasoning alone. The system showcases how agents
-with distinct roles and tools can **collaborate under orchestration**.
+This project demonstrates a **collaborative multi-agent system** built with the **Agent2Agent SDK** ([A2A])
+and [OpenAI](https://platform.openai.com/api-keys), where a top-level Auditor agent coordinates the workflow
+to verify facts. The Critic agent gathers evidence via live internet searches using **DuckDuckGo** through
+the Model Context Protocol (**MCP**), while the Reviser agent analyzes and refines the conclusion using
+internal reasoning alone. The system showcases how agents with distinct roles and tools can
+**collaborate under orchestration**.
 
 > [!Tip]
 > ✨ No configuration needed — run it with a single command.
@@ -25,41 +26,41 @@ with distinct roles and tools can **collaborate under orchestration**.
 + **A laptop or workstation with a GPU** (e.g., a MacBook) for running open models locally. If you don't have a GPU, you can alternatively use [**Docker Offload**](https://www.docker.com/products/docker-offload).
 + If you're using Docker Engine on Linux or Docker Desktop on Windows, ensure that the [Docker Model Runner requirements](https://docs.docker.com/ai/model-runner/) are met (specifically that GPU support is enabled) and the necessary drivers are installed
 + If you're using Docker Engine on Linux, ensure you have Compose 2.38.1 or later installed
++ An [OpenAI API Key](https://platform.openai.com/api-keys) 🔑
 
 ### Run the project
 
+Create a `secret.openai-api-key` file with your OpenAI API key:
 
-```sh
-docker compose up --build
+```
+sk-...
 ```
 
-No configuration needed — everything runs from the container. Open `http://localhost:8080` in your browser to and select `AgentKit` in the selector at the top left, then chat with
-the agents.
-
-Using Docker Offload with GPU support, you can run the same demo with a larger model that takes advantage of a more powerful GPU on the remote instance:
+Then run:
 
 ```sh
-docker compose -f compose.yaml -f compose.offload.yaml up --build
+docker compose up --build
 ```
 
-# 🧠 Inference Options
-
-By default, this project uses [Docker Model Runner] to handle LLM inference locally — no internet connection or external API key is required.
+Everything runs from the container. Open `http://localhost:8080` in your browser and then chat with
+the agents.
 
-If you’d prefer to use OpenAI instead:
+# 🧠 Inference Options
 
-1. Create a `secret.openai-api-key` file with your OpenAI API key:
+By default, this project uses [OpenAI](https://platform.openai.com) to handle LLM inference. If you'd prefer
+to use a local LLM instead, run:
 
-```
-sk-...
+```sh
+docker compose -f compose.dmr.yaml up
 ```
 
-2. Restart the project with the OpenAI configuration:
+Using [**Docker Offload**](https://www.docker.com/products/docker-offload) with GPU support, you can run the
+same demo with a larger model that takes advantage of a more powerful GPU on the remote instance:
 
+```sh
+docker compose -f compose.dmr.yaml -f compose.offload.yaml up --build
 ```
-docker compose down -v
-docker compose -f compose.yaml -f compose.openai.yaml up
-```
+
 
 # ❓ What Can It Do?
 
diff --git a/a2a/compose.dmr.yaml b/a2a/compose.dmr.yaml
@@ -0,0 +1,60 @@
+services:
+  # Auditor Agent coordinates the entire fact-checking workflow
+  auditor-agent-a2a:
+    build:
+      target: auditor-agent
+    ports:
+      - "8080:8080"
+    environment:
+      - CRITIC_AGENT_URL=http://critic-agent-a2a:8001
+      - REVISER_AGENT_URL=http://reviser-agent-a2a:8001
+    depends_on:
+      - critic-agent-a2a
+      - reviser-agent-a2a
+    models:
+      gemma3:
+         endpoint_var: MODEL_RUNNER_URL
+         model_var: MODEL_RUNNER_MODEL
+
+  critic-agent-a2a:
+    build:
+      target: critic-agent
+    environment:
+      - MCPGATEWAY_ENDPOINT=http://mcp-gateway:8811/sse
+    depends_on:
+      - mcp-gateway
+    models:
+       gemma3:
+         # specify which environment variables to inject into the container
+         endpoint_var: MODEL_RUNNER_URL
+         model_var: MODEL_RUNNER_MODEL
+
+  reviser-agent-a2a:
+    build:
+      target: reviser-agent
+    environment:
+      - MCPGATEWAY_ENDPOINT=http://mcp-gateway:8811/sse
+    depends_on:
+      - mcp-gateway
+    models:
+       gemma3:
+         endpoint_var: MODEL_RUNNER_URL
+         model_var: MODEL_RUNNER_MODEL
+
+  mcp-gateway:
+    # mcp-gateway secures your MCP servers
+    image: docker/mcp-gateway:latest
+    use_api_socket: true
+    command:
+      - --transport=sse
+      - --servers=duckduckgo
+      # add an MCP interceptor to log the responses
+      - --interceptor
+      - after:exec:echo RESPONSE=$(cat) >&2
+
+models:
+  # declare LLM models to pull and use
+  gemma3:
+    model: ai/gemma3:4B-Q4_0
+    context_size: 10000 # 3.5 GB VRAM
+    #context_size: 131000 # 7.6 GB VRAM
diff --git a/a2a/compose.openai.yaml b/a2a/compose.openai.yaml
diff --git a/a2a/compose.yaml b/a2a/compose.yaml
@@ -8,38 +8,34 @@ services:
     environment:
       - CRITIC_AGENT_URL=http://critic-agent-a2a:8001
       - REVISER_AGENT_URL=http://reviser-agent-a2a:8001
+      - OPENAI_MODEL_NAME=o3
+    secrets:
+      - openai-api-key
     depends_on:
       - critic-agent-a2a
       - reviser-agent-a2a
-    models:
-      gemma3:
-         endpoint_var: MODEL_RUNNER_URL
-         model_var: MODEL_RUNNER_MODEL
 
   critic-agent-a2a:
     build:
       target: critic-agent
     environment:
       - MCPGATEWAY_ENDPOINT=http://mcp-gateway:8811/sse
+      - OPENAI_MODEL_NAME=o3
+    secrets:
+      - openai-api-key
     depends_on:
       - mcp-gateway
-    models:
-       gemma3:
-         # specify which environment variables to inject into the container
-         endpoint_var: MODEL_RUNNER_URL
-         model_var: MODEL_RUNNER_MODEL
 
   reviser-agent-a2a:
     build:
       target: reviser-agent
     environment:
       - MCPGATEWAY_ENDPOINT=http://mcp-gateway:8811/sse
+      - OPENAI_MODEL_NAME=o3
+    secrets:
+      - openai-api-key
     depends_on:
       - mcp-gateway
-    models:
-       gemma3:
-         endpoint_var: MODEL_RUNNER_URL
-         model_var: MODEL_RUNNER_MODEL
 
   mcp-gateway:
     # mcp-gateway secures your MCP servers
@@ -52,9 +48,6 @@ services:
       - --interceptor
       - after:exec:echo RESPONSE=$(cat) >&2
 
-models:
-  # declare LLM models to pull and use
-  gemma3:
-    model: ai/gemma3:4B-Q4_0
-    context_size: 10000 # 3.5 GB VRAM
-    #context_size: 131000 # 7.6 GB VRAM
+secrets:
+  openai-api-key:
+    file: secret.openai-api-key
