Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions AGENTS.md
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@
- In Solid code, follow existing signal/memo/effect patterns and signal accessors like `value()`, not React-style state APIs.
- Before changing message-list virtualization, pagination, sticky navigation, row measurement, attachments, or inline editing, read `docs/message-list-virtualization.md` and preserve its identity, height, visible-anchor, and scroll-ownership invariants.
- When the user asks to "Run AI tests" or "Run fuzzy tests" (case-insensitive), follow `docs/ai-fuzzy-verification.md`. Unless the user narrows the scope, run its automated preflight and standard real-editor scenarios, use GPT Luna or Terra as specified, run realistic edit/tool streams only in the clean `tmp/opencode` fixture, save the run ledger under `artifacts/ai-fuzzy/`, and report blocked visual checks rather than treating browser or host-only tests as substitutes.
- AI parity and regression runs must include `docs/ai-action-matrix.md`: test sends, queueing, pause/edit/resume/remove, steering, inline edits, tools, actual subagents, root/child permissions, questions, cancellation, and view ownership recovery. Report action-by-backend results; streaming-only coverage is incomplete.
- When the user asks to "Run streaming tests" (case-insensitive), follow `docs/ai-streaming-verification.md`. This separate AI-test suite selects existing OpenCode history read-only and replays it in an isolated VS Code host with short gaps preserved and long waits capped. Review selection coverage, attach AI watchers before starting playback, save evidence under `artifacts/ai-streaming/`, and never treat scheduler/canonical checks alone as a visual or performance pass. Never execute recorded tools or mutate source sessions.
- Tests must never change existing production sessions, including timestamps, permissions, titles, or messages, unless the user explicitly requests that specific change. A test request is not authorization to migrate or repair production data. Live AI tests require a verified separate database under `artifacts/ai-test-data/`; never fall back to the normal server, reuse a production editor, or bypass isolation checks. Historical session reads remain read-only; test imported copies instead.
- Before changing permission modes or rules, auto-approval, permission events or snapshots, prompt grouping, attention state, or child-session approvals, read `docs/permission-lifecycle.md` and preserve its pending-request visibility and server-acknowledgement invariants.
Expand Down
20 changes: 19 additions & 1 deletion CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,25 +2,43 @@

This changelog summarizes the main user-facing improvements in each Varro minor release line. Patch releases are consolidated into their parent minor version. Dates reflect the development history.

## 0.29.x - September 2026

- Added OpenCode v2.0.5+ support alongside v1.16.0+, with automatic API detection and authenticated connections.
- Recommend v2 via `npm install -g @opencode/cli` or the standalone Windows CLI. V1 remains available via `npm install -g opencode-ai`; updates preserve the installed package family.
- Prefer `opencode2`, falling back to `opencode`, unless `varro.server.command` is set. Both current packages use `opencode`.
- After installing v2, let active work finish, run `Varro: Restart Server`, and confirm the version in the status bar. See [restart recovery steps](docs/usage.md#if-restart-does-not-switch-to-v2) if Varro stays on v1.
- Added `Varro: Import OpenCode v1 Session into v2` to copy conversations and child sessions with history intact, leaving originals unchanged.
- Included both versions in local usage reports without double-counting migrated or imported sessions.
- Kept v2 provider failures visible in live chat and history, even before a response starts.
- Updated tested OpenCode v2 support to 2.0.7, including request-start timing and hidden authentication defaults.
- Preserved session-scoped Always approvals across Default-mode sends and kept declined v2 tools out of completed-work summaries.

## 0.28.x - September 2026

- Added VS Code Problems context with workspace and individual diagnostic selection through the composer, slash command, and editor code actions. Captured diagnostics persist across queued messages, history, and message edits, with settings to disable inclusion.
- Added inline `$` skill references with completion and attachments, preserving skill references in message history and edits.
- Added compact auxiliary windows for new and existing chats, with a per-window light/dark theme toggle and a remembered preference.
- Added database context and table-attachment support in the shared chat interface, including persistent snapshots and message rendering for sessions with database attachments.
- Synchronized session read state across folder and multi-root views so read conversations stay read when switching workspaces.
- Improved Windows server startup and recovery, event handling, and OpenCode update support while limiting background work under heavy load.
- Added layered permission controls for session and project rules, inherited configuration, and server-memory allowances. Simplified session choices to Default, Auto, and Full access.
- Let agents continue permitted work after a rejected permission request, required scope selection when saving an always-allow rule, and prevented startup permission recovery from changing session history or recency.
- Show the requesting subagent in permission prompts.
- Added provider visibility and ordering controls, shared model preferences across workspaces, and usage-limit reporting for Claude Code. OpenAI and Z.ai limits now include reset details.
- Added model pricing details and provider quota percentages to the model picker, with clearer usage and polling details in provider-limit popups.
- Coordinated provider quota polling across instances, added available Grok quota reset times, kept authentication failures visible until credentials are verified, and fixed providers disappearing when all their models are deselected.
- Added drag-and-drop and keyboard reordering for pinned sessions, with custom order preserved across session lists and navigation.
- Expanded redacted diagnostics and Ralph verification evidence, with clearer provider failures and reconnect feedback.
- Expanded redacted diagnostics and Ralph verification evidence, with clearer provider failures and reconnect feedback. Simplified diagnostics export to copy or save reports directly.
- Opened plans in VS Code's Markdown preview while keeping them available as editor context.
- Sped up local usage reports for large histories with worker-based database aggregation and bounded fallbacks.
- Improved pasted images, draft and queued-message persistence, and attachment cleanup.
- Kept sent drag-and-drop files available for the lifetime of their sessions, and preserved the original attachments, agent, model, and reasoning variant when retrying failed turns.
- Strengthened sidebar and editor-chat recovery across delayed state, stalled delivery, stale documents, cross-folder prompts, concurrent session-scope changes, and asynchronous session transitions.
- Fixed editor-session initialization when VS Code recreates a webview while reusing its document identity.
- Stabilized long conversations, scrolling during activity transitions, prompt navigation, streaming Markdown, and deeply nested session lists.
- Smoothed streamed text and activity updates, preserved scroll anchors during resizing and collapsed activity, and kept file-change previews stable across repeated edits.
- Reduced streaming Markdown rendering work and stabilized follow-to-bottom during row resizing and resending edited messages.
- Fixed session-picker sends losing newly created sessions, highlighted reselected editor sessions, and removed stale Stop controls after completed responses.
- Fixed custom-answer submission and kept active auto-approve reviews visible.

Expand Down
18 changes: 13 additions & 5 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -8,19 +8,25 @@

Varro runs [OpenCode](https://opencode.ai) inside VS Code. It adds project-aware chat, parallel sessions, plan and change review, model and permission controls, commit-message generation, and usage reports.

**OpenCode v1 and v2 are both supported.** Varro automatically detects the connected API and uses the same workbench for either version. OpenCode v2 is recommended for new installations. Supported versions start at v2 **2.0.5** and v1 **1.16.0**.

OpenCode remains responsible for agents, providers, models, commands, skills, MCP servers, and their configuration. Varro reads that configuration and provides a VS Code interface for it.

![The Varro OpenCode workbench in VS Code](https://raw.githubusercontent.com/koltyakov/varro/main/assets/demo.png)

## Quick start

1. [Install Varro](https://marketplace.visualstudio.com/items?itemName=koltyakov.varro) from the VS Code Marketplace.
2. Install the OpenCode CLI with `npm install -g opencode-ai`.
2. Install [OpenCode v2](https://opencode.ai/v2/docs/) with `npm install -g @opencode/cli` on macOS, Linux, or WSL. For native Windows, download the standalone CLI from that page.
3. Run `opencode auth login`, or use `/connect` in Varro, if no provider is configured.
4. Open a folder in VS Code and select `Varro` from the Activity Bar.
5. Start a session. Varro starts or connects to OpenCode when needed.

On native Windows, install OpenCode from a Windows terminal because Varro runs it on the local Windows extension host. In a VS Code WSL window, install OpenCode inside that WSL distribution instead. Varro and OpenCode then run on the Linux extension host. OpenCode [recommends WSL for the best Windows experience](https://opencode.ai/docs/windows-wsl), although native Windows is supported. Native OpenCode data, including `auth.json`, logs, and sessions, is under `%USERPROFILE%\.local\share\opencode`.
Already using v1? You can keep it, or install it with `npm install -g opencode-ai`. Set `varro.server.command` to your v1 executable to select it explicitly. With this setting empty, Varro searches for `opencode2` first, then `opencode`. Current v1 and v2 packages both install `opencode`, so the command name alone does not identify the version. Run `opencode --version` to check. See [choosing and updating OpenCode](https://github.com/koltyakov/varro/blob/main/docs/usage.md#choose-and-update-opencode) before switching versions.

> After switching from v1 to v2, you may need to run `Varro: Restart Server` from the Command Palette. Installing v2 does not replace an already running v1 server. If restart fails or Varro still shows v1, stop the old OpenCode process explicitly, especially on Windows, then restart with the v2 executable. Let active work finish first. See [restart recovery steps](https://github.com/koltyakov/varro/blob/main/docs/usage.md#if-restart-does-not-switch-to-v2) and confirm the connected version in Varro's status bar.

Install OpenCode on the host where Varro runs. A native Windows VS Code window needs the Windows CLI; add it to `PATH` or set `varro.server.command` to its executable path. A VS Code WSL window needs OpenCode installed inside that WSL distribution. Native OpenCode data is under `%USERPROFILE%\.local\share\opencode`; WSL uses the Linux data directory.

Varro supports VS Code and VSCodium. Support for other VS Code forks is limited; see [VS Code fork compatibility](https://github.com/koltyakov/varro/blob/main/docs/vscode-forks.md).

Expand Down Expand Up @@ -59,6 +65,8 @@ Varro supports VS Code and VSCodium. Support for other VS Code forks is limited;
| LSP, formatters, and tools | Integrated | Show language-server status and tool activity. OpenCode owns definitions and execution. |
| Configuration and TUI | Handoff | Edit instruction files or continue a session in the OpenCode TUI. |

Coverage depends on the connected version. The supported v2 API has no session sharing or OpenCode LSP service. Varro disables sharing on v2; VS Code Problems can still be attached as context. V2 session annotations are stored locally by Varro. See [version differences and history import](https://github.com/koltyakov/varro/blob/main/docs/usage.md#version-differences-and-history).

## Sessions

Varro filters sessions to the current workspace and groups them into `Recent`, `Archive`, and `Recycle Bin`. Filters identify `Running`, `Needs attention`, `Failed`, `Plan ready`, and `Completed` sessions. Rows can show queued prompts, changed files, added and removed lines, token use, duration, and current state.
Expand Down Expand Up @@ -95,7 +103,7 @@ Permission requests provide `Reject`, `Once`, and `Always` actions. Each session

Child sessions inherit the nearest selected mode in their session tree. Manual child requests appear in the parent conversation while remaining owned by the child session.

Varro initially selects `Auto` unless `varro.chat.defaultPermissionMode` or a saved project/global selection says otherwise. `Default` does not mean that every action requires approval. Without custom rules, OpenCode allows most permissions, asks for `doom_loop` and `external_directory`, and blocks reading `.env` files and variants while allowing `.env.example`. See [OpenCode permissions](https://opencode.ai/docs/permissions/) to configure global or per-agent `allow`, `ask`, and `deny` rules, including rules matched by tool, command, path, subagent, URL, and external directory.
Varro initially selects `Auto` unless `varro.chat.defaultPermissionMode` or a saved project/global selection says otherwise. `Default` follows the connected OpenCode version's global and agent rules. V2 uses ordered `permissions` rules with `action`, `resource`, and `effect`; supported v1-format rules remain accepted. See [v2 permissions](https://opencode.ai/v2/docs/permissions/) or [v1 permissions](https://opencode.ai/docs/permissions/) for syntax and defaults.

See the [Varro permissions guide](https://github.com/koltyakov/varro/blob/main/docs/permissions.md) for mode behavior, manual approvals, and automatic review details.

Expand Down Expand Up @@ -137,15 +145,15 @@ Varro never mixes staged and unstaged changes, stages files, or commits automati

Varro connects to `http://127.0.0.1:4096` by default. Set `varro.server.port` to another port from 1 through 65535. For manual server management, disable the deprecated debug setting `varro.server.autoStart` and run `opencode serve --port 4096`.

The status bar shows the active OpenCode version and available updates. On macOS and Linux, `varro.server.autoUpdate` installs the latest OpenCode CLI update. Native Windows does not replace the CLI in the background. It shows an update prompt, waits for active work to finish, and stops a Varro-managed server before opening the update command so Windows releases its lock on `opencode.exe`. Stop a separately managed server yourself before updating.
The status bar shows the active OpenCode version and available updates. On macOS and Linux, `varro.server.autoUpdate` installs updates within the installed CLI family: `@opencode/cli` for v2 or `opencode-ai` for v1. Updating v1 does not migrate it to v2. Native Windows does not replace the CLI in the background. It shows an update prompt, waits for active work to finish, and stops a Varro-managed server before opening the update command so Windows releases its lock on `opencode.exe`. Stop a separately managed server yourself before updating.

Varro reloads global OpenCode configuration when OpenCode is idle. Changes to project configuration may require `Varro: Restart Server`. This command waits for active work and only restarts a server managed by Varro. Restart a manually launched server in its terminal.

## Requirements

- [VS Code](https://code.visualstudio.com/) or [VSCodium](https://vscodium.com/) 1.120 or newer
- [Node.js](https://nodejs.org/) 22.22.2+ on Node 22, or Node 24.15.0+
- [OpenCode CLI](https://opencode.ai/docs) 1.16.0 or newer on `PATH`, or configured through `varro.server.command`
- [OpenCode v2 CLI](https://opencode.ai/v2/docs/) 2.0.5+ recommended, or [v1 CLI](https://opencode.ai/docs/) 1.16.0+, on `PATH` or configured through `varro.server.command`. Tested with v2 2.0.7 and v1 1.18.31
- A trusted, non-virtual workspace. Remote workspaces run Varro and OpenCode on the remote extension host

## Documentation
Expand Down
Loading
Loading