[Bug]: fail to read audit content (v1.0.1)

[Qwenjin]

What happened?

api 500 error

500 response be like:
{
"error": "Failed to read file at any known location for .spec-workflow/..../requirements.md"
}

Component

Web Dashboard

Steps to Reproduce

1. (cursor ) run as mcp server, using config.toml
2. open web dashboard and click 【audit/审批】

Expected Behavior

Display approved content normally

Error Output (if any)

Version

1.0.1

Operating System

macOS

Additional Context

No response

[Pimzino]

Can you please restart your device to ensure there are no running processes or old web socket service that could be clashing as new API endpoints have been added for diffs

[Qwenjin]

Can you please restart your device to ensure there are no running processes or old web socket service that could be clashing as new API endpoints have been added for diffs

I've tried restarting cursor and reinstalling @pimzino/spec-workflow-mcp

[Pimzino]

Can you please restart your device to ensure there are no running processes or old web socket service that could be clashing as new API endpoints have been added for diffs

I've tried restarting cursor and reinstalling @pimzino/spec-workflow-mcp

Hey, I mean restarting your device please.

[Qwenjin]

still error

[Qwenjin]

In addition, the message pop-up has become transparent, and the button color does not distinguish between colors as before：

[Pimzino]

The popup has always been transparent post version 0.0.5.

I am unsure but this looks like a local issue because I am unable to recreate on my end. Have tried with source files by running node and npx commands.

The buttons were changed intentionally for a more unified look. I wonder if this is a macOS issue but I dont have a mac device to test on.

[Pimzino]

@Qwenjin Were you able to troubleshoot this further? I wonder if someone else with a macOS device can help us out here.

[Qwenjin]

I found the reason. May be ai used the previous version of the specification and generated the file: .spec-workflow/new-function/requirements.md.

After I deleted all the memories of ai and stat a new conversation, I let ai re-understand the specification of spec-workflow mcp. Now the generated file is in the right location: .spec-workflow/specs/new-function/requirements.md

[Qwenjin]

I also found that sometimes the "tasks" generated by ai do not fully base on .spec-workflow/templates/tasks-template.md, and I cannot check whether the format error of the task on the dashboard.

[Pimzino]

What Agent / AI Models are you using because in my testing I have used GPT5, Claude Sonnet / Opus, Qwen, Kimi k2 basically most of the current big models and they all follow the instructions pretty well surprisingly.

[Qwenjin]

The GPT5 model is used and developed under cursor IDE. It is necessary to clearly set the mdc rules to let Ai understand the usage method of mcp, otherwise Ai will directly modify the documents to be approved by itself.