[bug]: NanoGPT doesn't show the correct data

[kvn1351]

Pre-flight checks

• I searched existing issues and did not find a duplicate.
• I verified this on the current production released OpenCode version, or I explain why not below.

OpenCode version tested

1.2.27

opencode-quota version tested

2.10.0

Bug summary

Hey @kvn1351, thank you so much for this PR! Adding NanoGPT support is exactly the kind of contribution that moves the plugin forward, and your work here directly inspired what I ended up building.

Unfortunately, I can't merge this as-is because it's missing a few things the plugin requires, e.g. some provider metadata fields. But my template checklist didn't make those requirements explicit enough, so I'll be updating the template to reflect this clearly.

I also extended the implementation a bit, to include check balance as a fallback, incase someone stops using NanoGPT as a subscription source.

Thanks again!

Hey @slkiser, no worries. I'm very happy to help and thanks for this great plugin. Unfortunately, some of your changes have "broken" it. First of all, it's important to know that there isn't any daily quota for the LLM endpoint that is of any actual interest to the user. Daily quota only counts for images. Weekly quota is of interest however. Here's a screenshot of what we see:

So for a subscriber only the daily image usage and weekly token usage (out of the max being 60 million) count.

My branch correctly visualises this:

Whereas today's release shows me this:

It should show the usage in % relative to the 60 mil token limit for the week.

Steps to reproduce

.

Expected behavior

.

Actual behavior

.

Relevant logs/output

If not tested on current production OpenCode, explain why

No response

[slkiser]

Whoops, thanks for the clarification @kvn1351.

I realized after publishing a beta build that the current NanoGPT docs for /api/subscription/v1/usage show a daily / monthly response shape, while your PR #25 and handoff imply a different shape with weeklyInputTokens, dailyImages, and possibly dailyInputTokens.

Because I can’t test the NanoGPT API myself, I want to make sure I'm getting things right.

First, can you use this beta build to test what I implemented:

{
  "plugin": ["@slkiser/opencode-quota@beta"]
}

Then restart OpenCode and check:

• /quota
• /quota_status

---

Then, if you’re okay with it, could you help with one of these questions since I am totally confused:

1. Did the live NanoGPT API response actually contain weeklyInputTokens / dailyImages, or was that based on trial and error / observed behavior in the app or just testing you performed?

2. If possible, do you have a sanitized example of your real /api/subscription/v1/usage API response? You can remove anything sensitive.

[slkiser]

I published a beta for this, but I’m holding off on merging to stable because I still can’t verify the live NanoGPT response shape myself.

The blocker is that the public docs for /api/subscription/v1/usage show a daily / monthly schema, while the original PR/issue discussion implied a different weeklyInputTokens / dailyImages schema.

If anyone using NanoGPT can share a sanitized real response payload, that would let me finish this safely. Until then, the beta remains available for testing:

{
  "plugin": ["@slkiser/opencode-quota@beta"]
}

[kvn1351]

I published a beta for this, but I’m holding off on merging to stable because I still can’t verify the live NanoGPT response shape myself.

The blocker is that the public docs for /api/subscription/v1/usage show a daily / monthly schema, while the original PR/issue discussion implied a different weeklyInputTokens / dailyImages schema.

If anyone using NanoGPT can share a sanitized real response payload, that would let me finish this safely. Until then, the beta remains available for testing:

{
  "plugin": ["@slkiser/opencode-quota@beta"]
}

I'll give you the response today. I've unfortunately not had the time these past days. Sorry for that.