Add a shared OpenChamber control service with two thin adapters — a native
`openchamber` tool injected into managed OpenCode, and new CLI commands — so
users can manage parallel sessions, worktrees, and scheduled tasks
conversationally through agents or from the terminal.
Control plane:
- New openchamber-control service owning a fixed action contract:
projects.list, models.list, session list/create/send/fork/status/messages,
and schedule list/create/run/delete/toggle. Session and worktree deletion
and project registration are deliberately not exposed.
- New openchamber-sessions module owning create/worktree/prompt orchestration,
Goal Mode dispatch, wait semantics (initial idle never counts as completion;
timeout and cancellation are failures), and explicit partial-failure results.
- Scheduled-task logic extracted into a service shared by routes, CLI, and the
agent tool.
Agent tool:
- Managed OpenCode gets a materialized plugin registering one typed tool with
a loopback-only callback, per-child ephemeral bearer (timing-safe, never
persisted or logged), and abort propagation into the service.
- The ~1.5k-token schema applies progressive disclosure: short descriptions,
server-side validation returning actionable usage errors, and intent
guardrails — created sessions/tasks are user-facing work (not age
self-delegation); worktree/goal/agent/variant/wait are omit-by-default;
dispatches produce no completion notification, and later result r
to session.messages, which now returns the authoritative sessionStatus.
- session.create without a user-named model picks from favorites/re
send/fork omit the selection and the service reuses the target session's
last user-message model, agent, and variant before falling back t
- An "Agent control tool" setting (default on, Save + Reload to apply)
disables plugin injection entirely.
CLI:
- New `openchamber session`, `schedule`, `projects`, and `models` commands
with automatic instance targeting, --wait/--timeout/--last-assist
worktree flags, and Goal Mode, preserving interactive, non-TTY, --quiet,
and --json contracts. The control HTTP timeout derives from the w
instead of the 4-second default.
UI:
- New built-in "Schedule a Task" starter (/schedule-task) running a
dialogue that defines a task and offers to create it via the tool after
explicit confirmation; Craft a Goal and Feature Planning gain the
handoff offer, and guided starters reserve the question tool for concrete
option choices. Localized in all 10 locales, migrated into custom
starter lists, hidden on VS Code.
- Sidebar shows CLI/agent-created sessions live via the control eve
- openchamber tool calls render with per-action titles and metadata.
Adds /craft-goal autocomplete and chat handling for starting a Goal crafting session.
Introduces new Magic Prompts content and localized labels/descriptions for Goal crafting.
Migrates desktop draft starters to include Craft a Goal once and persists the migration marker.
Adds a server-side "small model" capability: direct, cheap LLM calls that
reuse the user's existing OpenCode provider logins — the mechanism OpenCode
uses internally for titles and summaries but does not expose through the
SDK or plugins. Zero new dependencies; plain fetch with per-provider wire
formats, credentials never leave the server.
Core (packages/web/server/lib/small-model):
- Resolution mirrors OpenCode's session scoping: explicit settings override
→ small_model from the OpenCode config → family scan within the session's
provider → the session's own model. The global provider scan only serves
callers without a session context, and background callers forbid it
entirely (restrictToPreferredProvider), so conversation content never
reaches a provider the user didn't pick — explicit choices excepted.
- Per-provider auth replicating OpenCode's plugin loaders: GitHub Copilot
(device token as bearer, no exchange), ChatGPT plan via the codex
Responses API (single-flight OAuth refresh written back to auth.json),
Anthropic messages, Google generateContent, generic OpenAI-compatible.
- OpenCode's free models (opencode/big-pickle, *-free) are never called
directly; unauthenticated providers are skipped by design.
- Prompt clamping to the model's catalog context limit; thinking disabled
where a wire switch exists (Z.AI/GLM, MiniMax-M3, Gemini Flash); robust
content parsing with a clear error when a thinking model spends its whol
budget on reasoning.
- Settings → Sessions gains a Small Model group: use-default checkbox plus
an override picker limited to authenticated providers, persisted with
web/desktop/VS Code sanitization parity.
Consumers:
- Session assist: a server-side watcher on the global SSE hub generates a
short recap and one suggested follow-up after a session idles quietly fo
a minute, stored on session metadata (openchamber.assist). Freshness is
keyed to the last assistant message id, so new activity invalidates the
payload everywhere with no extra writes. The chat shows the recap under
the last message after five quiet minutes and the suggestion as a
dismissible chip above the composer (tap fills the input, never sends).
Gated by a new Chat setting (default on) that is a hard generation
switch. Language is anchored to the conversation itself, with a
script-mismatch guard against model/backend language hallucination.
- TTS: a third input mode, summarized — long replies are condensed to
spoken prose before playback on any TTS engine.
- Git: commit-message and PR generation moved off the active chat session
onto the small model fed with real diffs and the commit list (bodies
included), with a session-transport fallback for free-model-only setups.
- Notes: Add to notes distills long selections into 1-3 dense sentences
preserving exact identifiers, with verbatim fallback on failure.
Fixes along the way:
- The global event watcher now starts unconditionally; it was gated behind
the desktop-notify env, leaving the server-side event hub dead in
packaged apps.
- OpenCode re-emits message.updated for old user messages after idle; the
watcher no longer mistakes those for new activity.
- Session metadata merges from a fresh read right before the PATCH, so
writes made during the generation window (suggestion dismissals, review
links) are preserved; the assist runtime stops during graceful shutdown.
Introduce a desktop/web-only /handoff-review flow that generates a handoff from the active implementation session, creates or reuses a separate review session in the same directory, and links the pair through hidden OpenChamber session metadata.
Add review flow orchestration, metadata helpers, magic prompts, localized command/action labels, session metadata create/update support, and assistant message transfer actions for sending reviewer feedback back to the implementer or implementation responses back to the reviewer.
Review sessions are ordinary sessions, not child sessions. The flow avoids exposing session IDs or routing metadata to agents, hides review controls on mobile and VS Code, hides unrelated assistant actions inside review sessions, cleans up stale metadata where possible, and uses the optimistic send path so cross-session sends scroll like normal composer messages.
Add a packaged-client runtime boundary so the shared UI can talk to local,
desktop, remote, and VS Code runtimes through the right transport instead of
assuming one same-origin web server.
Centralize OpenChamber-owned API access behind RuntimeAPIs, runtimeFetch, and
runtime URL helpers, while keeping official OpenCode traffic on the SDK path.
Support runtime switching, remote host selection, desktop client credentials,
and headless connection links for pairing packaged clients with remote
OpenChamber servers.
Harden the new auth model by moving long-lived client tokens out of browser
URLs, introducing short-lived scoped URL tokens for browser-owned transports,
restricting URL-token access to explicit readable/realtime routes, and making
client-token management session-scoped or self-scoped as appropriate.
Update browser-owned assets and preview proxy flows to work with the split
runtime model, including authenticated project icons, preview token propagation,
CSP-safe preview bridge injection, and preview proxy auth that survives
short-lived URL-token expiry.
Tighten Electron security boundaries for packaged clients by gating privileged
preload state to trusted origins and requiring explicit confirmation before
connect deep-links import or switch remote runtimes.
Also refresh agent guidance and project skills so future runtime/API, auth,
preview, UI, CLI, settings, locale, and drag-to-reorder work follows the new
architecture.
The old workspace-review prompt only hunted for catastrophic bugs and
omitted everything it wasn't highly certain about, so it usually returned
'No high-signal issues found' and felt useless.
Rework it to first understand the diff's intent and judge whether the
implementation actually achieves it and is adequate (smallest correct
solution, completeness), then review correctness across concrete failure
modes (concurrency, lifecycle, state fanout, comparators, a11y,
regressions, targeted-test gaps) plus a security/supply-chain lens, and
classify findings by severity (blocker / non-blocker / nit). Keeps the
existing false-positive discipline (diff is source of truth, validation
pass, no nitpicks, cite rules) and stays review-only by default.
Adapted from a battle-tested PR-review prompt, minus the GitHub-specific
parts that don't apply to a local workspace diff. Updates command and
settings descriptions across locales.
Reword the /catch-up prompt and its Magic Prompts settings descriptions to
emphasize an easy-to-digest summary rather than a 'warm'/'friendly' tone —
the goal is ease of comprehension, not friendliness. Updated across all
locales.
Rework the /catch-up decision tree from mutually-exclusive cases into
complementary layers. It now determines whether the current branch is the
repo's default branch and builds context accordingly: on a feature branch
it reads enough of the branch's own commits to grasp the intent and checks
the branch's PR even when there are uncommitted changes (so in-progress
work is understood as part of the larger effort); on the default branch it
lightly skims recent commits. Uncommitted changes remain the focal point
but are interpreted through that context.
The suggested next step now favors continuing the actual work over
housekeeping — since running catch-up implies the user stepped away and
the work is likely unfinished, push/PR/checks are only a fallback. Updates
the Magic Prompts settings descriptions across locales to match.
Promote the last plain-prompt draft chip into a proper /explore command so
all draft welcome presets are now commands. Its hidden prompt investigates
the repository and gives a structured orientation — big picture, main
modules, how they connect, conventions, and where to start — instead of a
file-by-file dump.
Follows the established pattern: visible + hidden magic prompts wired into
command autocomplete, the submit handler, the draft chip, and the Magic
Prompts settings page, with i18n across all locales. Drops the now-unused
explore prompt string.
Rework the /weigh hidden prompt so the recommendation is driven by what
best serves the user's actual goal — not by what is fastest or least
effort. Effort and complexity are surfaced as honest consequences rather
than reasons to steer toward a weaker option, strong-but-harder
approaches that fully cover the idea must be included and discussed, and
a watered-down solution is never recommended just because the proper one
is more work.
Add /weigh as a draft chip and command for the moment before planning —
when you know what to build but not how. Its hidden prompt investigates
the code, then lays out 2-3 genuinely distinct approaches with trade-offs
(complexity, risk, blast radius, effort) and a clear recommendation,
without writing a plan or code.
Follows the established command pattern: visible + hidden magic prompts
wired into command autocomplete, the submit handler, draft preset chips,
and the Magic Prompts settings page, with i18n across all locales. Adds a
scales-3 icon to the sprite.
Rework the /catch-up hidden instructions: write like a teammate catching
you up rather than a CI summary, scope strictly to the current branch,
and never surface unrelated branches, other people's PRs, or review
requests. When the current branch has no PR of its own, skip pull-request
talk entirely and just recap recent work plus a next step.
Replace the 'What changed recently' draft chip with /catch-up, a command
whose hidden prompt branches on git state: reconstruct intent from an
in-progress diff, check an open PR's review state, or summarize recent
commits.
Add /debug as a new draft chip and command: a guided root-cause
investigation that captures the symptom, forms hypotheses, checks them
against the code, and confirms the cause before proposing a fix.
Both follow the /workspace-review pattern — visible + hidden magic
prompts wired into command autocomplete, the submit handler, draft preset
chips, and the Magic Prompts settings page, with i18n across all locales.
Add starter preset chips under the composer on the desktop draft welcome
screen (explore, what changed, plan, review). Clicking a chip submits
immediately; prompt chips send a natural-language prompt, command chips
reuse built-in slash commands.
Add a new /plan-feature built-in command, modeled on /workspace-review:
visible + hidden magic prompts that run a guided, batched-question
planning dialogue (investigate the code, ask up to 3 clarifying
questions at a time, surface pitfalls, then produce an implementation
plan). Wired into command autocomplete, the submit handler, the draft
plan chip, and the Magic Prompts settings page, with i18n across all
locales.
Adds Run fusion for multi-run-like sessions
Combines sibling outputs into a new fusion session
Adds configurable Fusion prompts in Magic Prompts settings
Replace server-enforced structured output with local JSON parsing for Git generation
Render generated commit and PR JSON responses as compact chat cards
Tighten generation prompts while preserving active session context
Conflict resolution prompts now ask for an analysis-first strategy
Response style instructions are wrapped as system reminders
Prompt templates stay separate from reminder formatting
Add /review command with high-signal diff review instructions
Expose workspace review prompts in Magic Prompts settings
Show /review in command autocomplete with OpenChamber command metadata
Typing /summary (optionally with a topic hint) produces a structured
summary as a normal assistant message — without compacting or mutating
session history. Reuses the existing "Start new session from this answer"
button for hand-off.
Prompts are customizable in Settings → Magic Prompts under a new Session
group. Autocomplete shows an 'openchamber' badge to distinguish our
commands from OpenCode-built-in ones.
Add Save-as-plan action on assistant messages that writes a markdown plan
into the project plans directory and suggests a title from the text. Plan
view opens a specific saved plan via targetPath, supports inline edits, and
can kick off improve or implement flows in a new or worktree session.
Ship the matching hidden magic prompts (plan.todo/improve/implement) so the
instructions applied by these flows are visible and editable in the Magic
Prompts page.
Remove the plan editor font override, CodeMirror now uses the app font stack.
- Add configurable visible/instructions prompt families for commit/PR generation, PR checks/comments flows, and git conflict resolution helpers.
- Refactor prompt sending to explicit visible + synthetic parts instead of newline-based splitting, with legacy override migration for old keys.
- Polish Magic Prompts settings UX with grouped sidebar entries, tooltip-based descriptions, AI icon, and validation that visible prompts cannot be empty across web and VS Code runtimes.
Bohdan Triapitsyn