Replies: 2 comments
-
my fast investigation with CProfile, it took most time in compress_schema(), occupying almost 87% of total time. time occupations in compresss_schema() are:
when I replaced copy.deepcopy to json.dumps+loads, it gets almost 1.24x faster than before.
compresss_schema() occupies 72% of total time, and its NEW breakdowns are:
more investigations, more faster... I hope. |
Beta Was this translation helpful? Give feedback.
-
I tested with new OpenAPI parser in main branch, and found:
thank you, jlowin |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
I'm trying fastmcp.from_openapi() for OpenAPI spec which got from REST production server.
openAPI spec for production server has lots of paths and methods.
i.e: github REST API has over 1000 APIs in paths (num of path x methods).
I found it takes too much time to create instance as below:
seconds measured with time.time(), i.e. user experience time, took diff just before/after FastMCP.from_openapi().
as the result, production mcp clients, such as VC copilot chat UI, often give up waiting FastMCP ready to serve when transport is stdio
Beta Was this translation helpful? Give feedback.
All reactions