[FEATURE]: desktop: show cross-session token totals and cache hit rate in the Context panel #58024

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

OpenCode Desktop 2.1.2, Windows 11 (10.0.26300), custom OpenAI-compatible provider.

What I want

The desktop Review → Context panel reports one session in isolation: messages, total tokens, input/output/reasoning tokens, cache read/write, cost, context limit and usage percentage. Two numbers that decide whether prompt caching is actually working are missing:

1. Total tokens used across sessions — a running total for the project, not just the current session.
2. Cache hit rate — cache.read / (input + cache.read), for the current session and as a rolling average.

Why

With a custom OpenAI-compatible provider, cache hits are the difference between a cheap session and an expensive one. The panel shows raw cache read/write counts but no ratio, and nothing aggregates across sessions, so there is no way to tell from the app whether caching is working.

The data already exists

GET /api/experimental/session/stats already returns the aggregate: SessionStats.Info → tokens: TokenUsage.Info { input, output, reasoning, cache: { read, write } }, cost, and a per-model models[] breakdown. This is a presentation gap in the Context panel, not a new computation.

Proposal

• Add session-total and cache-hit-rate rows to the Context panel, derived from the existing usage fields.
• Add an aggregate section (this project / all sessions, optional time range) fed by experimental.session.stats.
• Render "not reported by provider" when cache fields are absent instead of 0.

No plugin path

Per the V2 docs only CLI/TUI plugins (@opencode/plugin/tui, OpenTUI renderables) can draw UI, and only in the terminal. The desktop app exposes no UI extension point, so this cannot be filled in by a plugin.

Related

• [FEATURE]: Per-model cache efficiency and configuration diagnostics in Settings #49335 (open) — per-model cache efficiency in Settings > Models. Different surface; this is the session Context panel plus cross-session totals.
• [FEATURE]: tokens-per-second in the desktop app (context usage indicator) #49380 (open) — tokens per second in the desktop context usage indicator.
• [FEATURE]: Desktop 2.0 must ship built-in session metrics (tokens/sec + cache hit rate) — no UI plugin path exists #52029 — adjacent; auto-closed by the compliance bot for template non-compliance, not on the merits.
• [FEATURE]:Show context usage in OpenCode Desktop, including for local OpenAI-compatible models #34900 — desktop context usage incl. cached tokens, closed as not planned.