@op1/fast-mode

Fast-mode plugin for OpenCode that toggles request priority on demand and applies it only when your runtime conditions match (provider, model, agent, and enabled model target).

What the plugin does

• Exposes a TUI fast-mode controller for direct ON/OFF toggling.
• Exposes fast_mode tool actions: status, on, and off as a low-level fallback.
• Writes/reads runtime toggles from .opencode/fast-mode-state.json.
• Reads guard configuration from:
  • ~/.config/opencode/fast-mode.json (global)
  • .opencode/fast-mode.json (project, overrides global)
• Applies options.serviceTier = "priority" only when a request passes provider, model, and agent allowlists and the matching provider/model toggle is ON.

Install

bun add @op1/fast-mode

If you install from source workspace, run the local build first:

bun run build --cwd packages/fast-mode

Configure

Add plugin access in ~/.config/opencode/opencode.json:

{
	"plugin": ["@op1/fast-mode"]
}

If your opencode.json already has a plugin list, merge @op1/fast-mode into it.

To enable the TUI controller, also add @op1/fast-mode to ~/.config/opencode/tui.json:

{
	"$schema": "https://opencode.ai/tui.json",
	"plugin": ["@op1/fast-mode"]
}

OpenCode resolves the package ./tui export automatically from the package root.

Fast-mode config file shape

Create or edit ~/.config/opencode/fast-mode.json:

{
	"enabled": true,
	"providers": {
		"openai": {
			"enabled": true,
			"agents": ["build", "coder"],
			"models": ["gpt-5.3-codex", "gpt-5.3-codex-mini"]
		}
	}
}

Supported fields:

• enabled (boolean): global feature gate for fast-mode.
• providers (record): provider-level config, keyed by provider ID.
• providers.<provider>.enabled (optional boolean): per-provider gate.
• providers.<provider>.agents (optional string[]): agent allowlist.
• providers.<provider>.models (optional string[]): model allowlist.

Project-level /.opencode/fast-mode.json config is merged over the global file.

You can also add a .opencode/fast-mode.json in a repo to tune behavior per project.

TUI usage

Open the command palette in the OpenCode TUI and run Fast Mode.

The dialog shows the current ON/OFF state for:

• openai / gpt-5.4
• openai / gpt-5.3-codex

Selecting a configured provider/model pair flips its state immediately and persists it to .opencode/fast-mode-state.json.

Notes on behavior

• This plugin is provider-first: it matches provider ID first, then model and agent allowlists.
• It is currently OpenAI-focused because it writes serviceTier in the request options contract used by OpenAI-style providers.
• It is model-guarded: if model/agent/provider checks fail, no request mutation occurs.
• TUI toggles only control runtime provider/model state; config gates in fast-mode.json still decide whether fast mode can actually apply.

Publishing readiness

Build now emits declarations for npm publishing:

bun run build --cwd packages/fast-mode

That script writes:

• dist/fast-mode.js
• declaration files under dist/ (for example dist/index.d.ts)

Keep dist/ committed for package publishing.

License

MIT