[v6 features] OpenCode 2: what should OpenCode Quota become?

[slkiser]

OpenCode 2 support is coming in v.5.0.0 as a "make it work" release. After that, I plan a bigger remake built for OpenCode 2. This issue collects your feedback.

Good to know

• OpenCode 2's built-in /stats page shows year-to-date tokens, a heatmap, and streaks. It doesn't show cost per model, plan limits, or reset times, and plugins can't change it.
• Plugins can add their own terminal (TUI) screens. On Web and Desktop, plugins can't add screens.
• Everything stays on your machine. The plugin never sends requests to your AI model to check quota.

1. What you use today
Which parts do you use (sidebar, toasts, the compact line, the prompt bar, /quota, /tokens_*, exports)? What should stay, and what annoys you?

2. What's missing

• What do you want to know about your usage, cost, or limits that you can't see today?
• Which providers and plans do you use?
• Do you mainly use the terminal (TUI), or Web/Desktop? Something else entirely?

[Cyberistic]

Alright I've been using this for a few months, I will give my feedback. Only TUI.
Current plans, based on price: KIMI, Minimax, XAI, Opencode Go, Copilot, OpenRouter (pay-as-you-go). Next month dropping most in favor of Deepseek.

1. I only use the sidebar, never anything else. I have no use for any of the commands.
2. Slightly annoying: when I'm on a small pane (in tmux) and I go full-screen, the plugin blocks the thread while it checks API usage. Ideally, it should be non-blocking and happen in the background, or be optimistic (use stale values then update if available)
3. Adding profiles is hard. I have multiple accounts for the same providers, and each time it's very hard to configure for quota to work. Though I don't mind this now since the pattern is established and my agent can figure it out.
4. Some providers give info I don't care about, such as video/image generations per day/week. It stays at 0 and occupies a few bars. Per provider config would be nice.
5. Some providers are stale and I don't use their services right now. I don't want to remove the .env just in case, but would be nice to hide them.
6. Error consistency: each provider gives a different error message, would be nice to standardize based on code or something. (limit reached, key invalid, bad request..)
7. Ideally, I would want all settings editable within the TUI, so I don't have to look through commands. Things such as disabling toasts, etc.

[TheStaticMage]

1. I use CodeNomad and look at the toasts. I never use the built-in commands.
2. OpenCode Go, OpenAI, Z.AI, and Deepseek API balance show up now.

• In CodeNomad the progress bars split across two lines so it doesn't look great, but it's still usable.
• OpenRouter doesn't ever seem to work right despite the key being valid.
• CodeNomad has recently added its own support for showing the quota so I'm less dependent on this plugin. However, since I'm using OmniRoute for the Z.AI and Deepseek, CodeNomad sees the provider as OmniRoute and doesn't know how to retrieve the data to show, so the toasts from this plugin are still useful.

[Samson026]

I this is a great plugin! However one thing that could be great is a visualiser of how much usage I should have used by that time in the month. Like a line indicator if I'm above or under my quoter if it was evenly distributed throughout the month / day or week.

[slkiser]

Feature ideas moved here from their own issues. v5 takes bug fixes and small upgrades that don't change how quota is shown; display ideas go to v6, which will pick from the ideas collected here:

• [feature]: optionally hide Anthropic Fable quota row #303 (@balaji-dutt): hide just the Fable row, keeping the other Anthropic rows.
• [feature]: show reset minutes only below one hour #302 (@balaji-dutt): reset times without minutes until under an hour, e.g. 3h then 42m.
• [FEATURE] Colorize the TUI sidebar Quota panel by usage threshold #292 (@liulianggui520-ops): color the sidebar Quota panel by usage level.

[slkiser]

Moved from #303 (@balaji-dutt): [feature]: optionally hide Anthropic Fable quota row

On AI assistance: OpenCode (GPT-6 Sol) helped me narrow down the behavior and draft the wording below, including the repro steps. The actual issue and workflow are mine. — @balaji-dutt

When the Anthropic quota response includes a Fable limit, the sidebar shows a Fable weekly row (see #257 and PR #259). I rarely use Fable and would like to hide that row without losing Anthropic’s other quota windows.

With @slkiser/opencode-quota@4.10.0, enable Anthropic and the sidebar using an account whose response includes Fable. The Fable row appears. Disabling Anthropic also hides the other Claude rows, while onlyCurrentModel filters more broadly. I could not find a Fable-only visibility option in the v5.0.0 configuration loader either.

Could there be an opt-in setting to hide just the Fable row across human-facing quota displays, defaulting to today’s behavior? It should leave the Anthropic fetch and non-Fable rows intact. Guidance on an existing way to filter only this row would also help.

[slkiser]

Moved from #302 (@balaji-dutt): [feature]: show reset minutes only below one hour

On AI assistance: OpenCode (GPT-6 Sol) helped me narrow down the behavior and draft the wording below, including the repro steps. The actual issue and workflow are mine. — @balaji-dutt

I use the quota sidebar to see roughly when windows reset. Minute-precise strings for multi-hour or multi-day windows add visual noise, but I still want accurate minutes in the final hour.

With @slkiser/opencode-quota@4.10.0, enable the sidebar and compare a reset above and below one hour. The normal countdown includes minutes in both cases. resetTimeDecimals: 0 reduces clutter, but rounds hours and can show 1h with less than an hour remaining; resetTimeSpaced: false only changes spacing. I also checked the v5.0.0 formatter and configuration docs and did not find a threshold option. This is an optional preference alongside #238’s minute-precision work, not a request to change the default.

Could the plugin offer an opt-in format that suppresses minutes at one hour or more, then shows accurate minute countdowns below one hour? Ideally it would work consistently across human-facing quota displays and retain existing behavior when a provider supplies no reset timestamp.