Files
openchamber/packages/web/server/lib/openchamber-control/DOCUMENTATION.md
T
Bohdan Triapitsyn a5aa32446d feat(browser): replace the preview proxy with a real browser panel and an agent web tool (#2883)
The preview panel worked by proxying a dev server through OpenChamber's own
origin and rewriting the HTML that came back. Anything the rewriter did not
anticipate broke, and pages that refuse to be embedded never loaded at all.
This deletes the proxy (-1604 lines and its tests) and merges the preview and
browser panels into one surface backed by a real Chromium view.

What the panel is now

- A `<webview>` in its own session partition: logins and cookies persist, hot
  reload works because nothing is rewritten, DevTools are one click away.
- Annotation: pick one element, drag a region, or draw freehand, write a note,
  and it reaches chat with a screenshot of the visible page with the marks on it.
- Toolbar: hard reload, page zoom, device sizes, a light/dark switch that
  applies to the page rather than the app, and cookie/cache clearing scoped to
  the panel alone.
- Several pages at once, each tab showing the page's own favicon, and an address
  bar that suggests pages already visited in this project.
- Dev servers are listed from what is actually listening on the machine, checked
  against what a project announced, so a server is offered no matter how it was
  started. One that is still starting is waited for instead of failing.

Remote dev servers

The desktop app binds a local port and pipes raw bytes to the OpenChamber host
over the existing authenticated connection, so the page keeps its own origin at
the root of its own host. The reachable set is exactly what discovery reports
and is re-checked per connection, so an authenticated client cannot dial
arbitrary local services on the host. Links and redirects to another loopback
port stay on the machine that served the page. A tunnel that cannot be opened is
reported; it is never replaced by the plain loopback URL, which would answer
from the user's own machine under a remote address.

Agent control

Browser actions are a separate `openchamber_web` tool: open, snapshot, click,
type, scroll, inspect computed styles, resize between mobile/tablet/desktop, and
capture a screenshot into `.openchamber/screenshots/` in the project. The
existing `openchamber` tool keeps sessions, worktrees and scheduled tasks. Each
has its own setting in the new Settings -> General -> OpenChamber Tools section,
and the plugin is not injected at all when both are off.

Capability belongs to the connected client, not to configuration: a client
declares on its event stream that it can drive a page, which only a Chromium
host does. Exactly one client performs each request — it claims the request
before acting, and the first claim wins — because deciding by whose result
arrives first would be too late for a click that already happened. No client
listening is answered immediately with an explanation rather than a timeout.

Runtime boundaries

Web tabs get a plain iframe that can display a page but not inspect one. The
VS Code extension no longer offers the surface at all, since nothing that makes
the panel worth having works there. Mobile is unaffected.

Native boundary

Camera, microphone, location and device-picker requests from panel pages are
denied — Electron grants them by default when no handler is set, and the panel
loads whatever address the user types. Page capture, appearance emulation and
storage clearing verify that their target belongs to the panel's own session
instead of trusting a web-contents id from the renderer.

Persisted state

Stored `preview` tabs migrate to `browser` (v13 -> v14). Context panel tab
limits are now per surface, so filling one surface no longer evicts another's
tabs. Address history is stored per project and per runtime.

Documentation

`preview.mdx` and `desktop-browser.mdx` rewritten across all locales, the agent
tool settings path corrected, new `DOCUMENTATION.md` for the browser-control
broker and the dev tunnel, and the `ui-api-decoupling` skill updated where it
still described the deleted proxy.
2026-08-13 22:44:13 +03:00

3.7 KiB

OpenChamber Control Service

Purpose

This module owns the typed control contract shared by the OpenChamber CLI and the managed OpenCode openchamber tool. Both adapters delegate to createOpenChamberControlService(); neither adapter may call or spawn the other.

Boundaries

  • service.js validates and executes the fixed project, model, session, and scheduled-task action allowlist. actions.js marks CLI-only actions with agentExposed: false (currently schedule.status); the agent tool consumes the filtered OPENCHAMBER_AGENT_TOOL_* exports. schedule.toggle requires the disabled boolean and replaces separate enable/disable actions; schedule.list also returns scheduler status as scheduler.
  • routes.js is the authenticated CLI HTTP adapter. It forwards one action, preserves service status and partial-result details, and propagates request cancellation.
  • ../agent-tool/runtime.js is the managed-tool adapter. It wraps service results in the versioned native-tool envelope and uses a separate ephemeral loopback credential.
  • ../openchamber-sessions/routes.js and ../scheduled-tasks/service.js own their domain operations and are composed into this service.

Invariants

  • Session status and messages come from official directory-scoped OpenCode APIs. Message output includes only ordered text parts.
  • Wait never treats an initial idle response as completion after dispatch. It requires observed activity or a newly completed assistant message.
  • Timeout and cancellation are failures, never authoritative idle results.
  • Validation that protects side effects runs before session creation or dispatch. An explicitly requested model, agent, or variant is checked against the directory's own OpenCode agent and provider lists before any session, worktree, or goal is created, because prompt_async accepts an unusable selection and then fails only on the event stream. A failed or empty lookup never turns a valid selection into a rejection.
  • promptDispatched reports an observed dispatch, never an accepted request. After prompt_async the service confirms a new user message reached the session; when it does not, the result reports promptDispatched: false with promptError instead of claiming success.
  • Send and fork dispatches without an explicit model/agent/variant reuse the target session's last user-message selection before falling back to the configured defaults; only session creation resolves defaults directly.
  • Usage errors name the missing or conflicting input so CLI and agent-tool callers can correct an invalid request without an upfront usage manual.
  • Explicit projectId or directory scope takes precedence over the managed tool's current-session directory fallback; the fallback never creates a conflicting second scope.
  • One failed directory status lookup produces unknown for only that directory and does not erase other session results.
  • Destructive session/worktree deletion and project-path registration are not part of the action contract.
  • browser.capture writes its image on the server, into .openchamber/screenshots/ under the scoped project directory, and returns the project-relative path rather than the image bytes. The client that took the picture may be on a different machine than the repository, and a path is what an answer, a commit, or a review can use; base64 in a tool result cannot be any of those. The agent's label is reduced to a filename fragment, never used as a path. The result also states how to present the image, because chat renders the image paths written in a finished answer below that message — saving the file is not what shows it to anyone.