OpenCode Dashboard

A live usage dashboard for OpenCode, served by a single Python file — no framework, no build step, no dependencies.

• Single file: opencode_dashboard.py is the whole server and page.
• Auto-refreshing every 15s; shows session history, per-day token/message charts, cost by model, tool usage, and a work-hour heatmap.
• Click to filter: days, tools, agents, models. Hover for tooltips.
• Standalone app: the included OpenCode Dashboard.command launcher opens a dedicated Chrome window and shuts the server down when the window closes.

Requirements

• Python 3.9+
• An OpenCode installation (it reads the SQLite db that OpenCode maintains)
• Chrome (optional, for the .command launcher's dedicated window)

Usage

python3 opencode_dashboard.py --port 8765
# open http://localhost:8765

Or double-click OpenCode Dashboard.command for a dedicated app window.

Options

Flag	Default	Description
--port	8765	HTTP port to serve on
--path	auto	Path to the directory containing OpenCode's SQLite db
--open	—	Launch the app window immediately
--time	0	Also re-render a plain-text stats file every N seconds

API

• GET /api/stats — everything the page renders, as JSON
• GET / — the dashboard page
• POST /api/close — stop the server (used by the launcher on quit)

Why a single file?

Copy opencode_dashboard.py (and, optionally, the .command launcher) to any machine with a Python 3 and an OpenCode db. Point it at that db and you're done. Everything — HTML, CSS, JS, SVG charts, KPI math, HTTP server, graceful shutdown — lives in one importable file you can read end to end.

CSV export

From the dashboard you can export the full stats JSON with the Export button.

Notes

• The sessions.cost column is not always populated by every session; the dashboard estimates the per-session cost as the model's share of that session's tokens instead.
• Reasoning tokens are counted as output ("Output · reasoning"); cached tokens are shown separately from the input/output bars.