genai-function-calling: Adds MCP example to openai-agents #59
Conversation
codefromthecrypt
left a comment
codefromthecrypt
left a comment
There was a problem hiding this comment.
mind doing stdio so that the impl matches the flow?
|
|
||
| For example, to run with Docker: | ||
| ```bash | ||
| docker compose run --build --rm genai-function-calling -- --mcp |
There was a problem hiding this comment.
I think..
| docker compose run --build --rm genai-function-calling -- --mcp | |
| docker compose run --build --rm genai-function-calling --mcp |
There was a problem hiding this comment.
Oops yeah just copied this from vercel-ai without testing it. I think it may need the CMD -> ENTRYPOINT change
| @@ -1,5 +1,6 @@ | |||
| openai-agents~=0.0.8 | |||
There was a problem hiding this comment.
mind checking latest doesn't break this?
Co-authored-by: Adrian Cole <[email protected]>
codefromthecrypt
left a comment
codefromthecrypt
left a comment
There was a problem hiding this comment.
Hmm I think this demonstrates that SSE is worse than stdio wrt the propagation concern. We can still merge it, but I'd go back and try to change it to stdio later. We should take the lessons about SSE with us as we progress MCP propagation.
SG?
| mcp_server = FastMCP(log_level="WARNING") | ||
| for tool in tools: | ||
| mcp_server.add_tool(tool) | ||
| # Mysteriously, cleanup such as from opentelemetry-instrument does not run on exit |
There was a problem hiding this comment.
Sorry absolutely no clue what's going on here
|
Went ahead and switched to stdio |
codefromthecrypt
left a comment
codefromthecrypt
left a comment
There was a problem hiding this comment.
great job with pragmatic workarounds!
2 participants
Same implementation as vercel-ai