[VibeCoding] Python uv project with starlette - thangchung/mcp-labs GitHub Wiki

Use `uv init` to start the simple boilerplate code using the `Starlette framework`, and update the readme file to guide the user to get started as well. You build and make sure it works well.

Export .env:

GOOGLE_API_KEY=<get from Google AI Studio>
GOOGLE_GENAI_USE_VERTEXAI=FALSE
MODEL_GEMINI_PRO=gemini-2.0-flash

uv sync
uv venv # activate env then
uv pip install -r requirements.txt
uv run chainlit run chainlit_app.py
adk web

TODO 1:

Add mcp-servers (https://pypi.org/project/mcp/)
Make it work with starlette (https://github.com/dapr/dapr-agents/blob/main/quickstarts/07-agent-mcp-client-sse/server.py)
Call of it from Agent (https://github.com/Tsadoq/a2a-mcp-tutorial/blob/main/adk_agents_testing/mcp_tools/mcp_tool_search.py#L30)

uv run .\mcp_server\calculator_server.py --server_type sse --port 8080
uv run chainlit run .\main.py
uv run adk web

High Level Architecture - A2A_MCP - archive 1

High Level Architecture - A2A_MCP

TODO 2:

What about running a devcontainer in Linux and running Docker model in that VM, like https://k33g.hashnode.dev/docker-model-runner-is-available-on-linux
litellm with Ollama: https://github.com/google/adk-python/issues/49

TODO 3:

From MCP Server call to A2A for resolving AgentCard: https://github.com/kaaquist/a2a-mcp-fun/blob/master/weather-agent/src/weather_agent/a2a_agent_mcp_tool.py#L26

TODO 4: Discord <-> kagent (A2A)

Here it is: https://kagent.dev/docs/examples/discord-a2a
The article at https://kagent.dev/docs/examples/slack-a2a, shows us how to use a Slack bot and kagent
Learn this template https://github.com/kagent-dev/a2a-slack-template, and convert it to a Discord bot (https://discordpy.readthedocs.io/en/stable/discord.html and https://discord.com/developers/docs/resources/application)