Commit Graph

13 Commits

Author SHA1 Message Date
hikari 4134e11c88 chore: CLI v2.1.75–v2.1.80 audit and support (#223–#232) (#233)
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 1m8s
CI / Build Linux (push) Has been cancelled
CI / Build Windows (cross-compile) (push) Has been cancelled
CI / Lint & Test (push) Has been cancelled
## Summary

This PR implements all tickets filed from the CLI v2.1.74 → v2.1.80 changelog audit (issues #223–#232).

### Changes by Issue

- **#223** — `feat: handle Elicitation and ElicitationResult hook events`
  New `ElicitationModal.svelte` component, Rust parsing for `[Elicitation Hook]` and `[ElicitationResult Hook]`, new store methods, and TypeScript event types.

- **#224** — `feat: handle StopFailure hook event for API error turns`
  Rust parsing for `[StopFailure Hook]`; frontend shows error toast + error character state.

- **#225** — `feat: handle PostCompact hook event`
  Rust parsing for `[PostCompact Hook]`; frontend shows info toast + success character state.

- **#226** — `feat: expose --name CLI flag as session name at startup`
  Added `session_name` field to `ClaudeStartOptions`; `StatusBar.doConnect()` passes the conversation name.

- **#227** — `fix: tighten startup watchdog and correct misleading comment`
  Startup watchdog tightened from 60 s → 30 s; corrected a comment that said "5 minutes" whilst the code used 60 seconds.

- **#228** — `fix: document cost estimation review and update default model fallback`
  Default model fallback updated from `claude-sonnet-4-5-20250929` → `claude-sonnet-4-6`; added doc comment explaining why char-based estimation is unaffected by v2.1.75 token overcounting fix.

- **#229** — `chore: update supported CLI version constant to 2.1.80`
  `SUPPORTED_CLI_VERSION` bumped in `CliVersion.svelte`.

- **#230** — `feat: surface memory file last-modified timestamps in MemoryBrowserPanel`
  Backend populates `last_modified` Unix timestamp; frontend formats and displays it per file.

- **#231** — `feat: update max_output_tokens upper bound and helper text for 128k`
  Input max raised to 128 000; placeholder and helper text updated to reflect model-dependent defaults and 128 k ceiling for Opus/Sonnet 4.6.

- **#232** — `fix: document non-streaming fallback compatibility with mid-session watchdog`
  Added doc comment above `STUCK_TIMEOUT` explaining the 5-minute watchdog is intentionally larger than the CLI's 2-minute non-streaming API fallback.

---

 This PR was created with help from Hikari~ 🌸

Reviewed-on: #233
Co-authored-by: Hikari <hikari@nhcarrigan.com>
Co-committed-by: Hikari <hikari@nhcarrigan.com>
2026-03-23 14:28:08 -07:00
hikari 452fe185df feat: CLI v2.1.68–v2.1.74 compatibility updates (#221)
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 1m21s
CI / Lint & Test (push) Has started running
CI / Build Linux (push) Has been cancelled
CI / Build Windows (cross-compile) (push) Has been cancelled
## Summary

This PR brings Hikari Desktop up to full compatibility with Claude Code CLI versions v2.1.68 through v2.1.74, implementing all changelog items audited in issues #200–#218.

## Changes

### Bug Fixes
- Remove deprecated Claude Opus 4.0 and 4.1 models from the model selector
- Auto-migrate users pinned to deprecated models to Opus 4.6

### New Features
- Add cron tool support (`CronCreate`, `CronDelete`, `CronList`) with character state mapping and `CLAUDE_CODE_DISABLE_CRON` settings toggle
- Handle `EnterWorktree` and `ExitWorktree` tools in character state mapping and tool display
- Add CLI update check with npm registry indicator in the version bar
- Add `agent_type` field and support the Agent tool rename from CLI v2.1.69
- Consume `worktree` field from status line hook events
- Display per-agent model override in the agent monitor tree
- Expose Claude Code CLI built-in slash commands (`/simplify`, `/loop`, `/batch`, `/memory`, `/context`) in the command menu with CLI badges
- Add `includeGitInstructions` toggle in settings
- Add `ENABLE_CLAUDEAI_MCP_SERVERS` opt-out setting
- Linkify MCP binary file paths (PDFs, audio, Office docs) in markdown output
- Add auto-memory panel, `/memory` slash command shortcut, and unified toast notification system
- Toast notifications for `WorktreeCreate` and `WorktreeRemove` hook events
- Sort session resume list by most recent activity, with most recent user message as preview
- Convert WSL Linux paths to Windows UNC paths when opening binary files via `open_binary_file` command
- Expose `autoMemoryDirectory` setting in ConfigSidebar (Agent Settings section)
- Add `/context` as a CLI built-in in the slash command menu
- Expose `modelOverrides` setting as a JSON textarea in ConfigSidebar (for AWS Bedrock, Google Vertex, etc.)

> **Note:** The CLI update check commit does not have a corresponding issue — it was a bonus addition during the audit sprint.

## Closes

Closes #200
Closes #201
Closes #202
Closes #205
Closes #206
Closes #207
Closes #208
Closes #209
Closes #210
Closes #211
Closes #212
Closes #213
Closes #214
Closes #215
Closes #216
Closes #217
Closes #218

Reviewed-on: #221
Co-authored-by: Hikari <hikari@nhcarrigan.com>
Co-committed-by: Hikari <hikari@nhcarrigan.com>
2026-03-13 01:34:44 -07:00
hikari 2816e33257 fix: resolve message submission and stuck processing bugs (#199)
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 1m2s
CI / Build Linux (push) Has been cancelled
CI / Build Windows (cross-compile) (push) Has been cancelled
CI / Lint & Test (push) Has been cancelled
## Summary

- **Fix `isProcessing` tracking**: The `isProcessing` store field was initialised as `false` and never set to `true` in production, making all submission guards no-ops. Now `setProcessing(true)` is called after `send_prompt` succeeds in both `handleSubmit` and `handleQuickAction`, and `setProcessingForConversation(id, false)` is called when the backend emits an idle/success/error state.
- **Fix auto-granted tools dropped on permission reconnect** (closes #198): `PermissionModal.svelte` was passing only session-granted tools when reconnecting after a permission approval, silently dropping `config.auto_granted_tools`. Fixed to merge both sets, matching the behaviour of every other `start_claude` call site.
- **Add mid-session watchdog**: A watchdog thread now kills the Claude Code process if a user message is sent but no `Result` arrives within 5 minutes. This triggers the existing disconnect/error flow so the user is notified and can reconnect. A generation counter ensures watchdogs from previous sessions exit cleanly when a new session starts.

## Test plan

- [ ] Send a message and verify the textarea is disabled and the stop button is visible while Claude is processing
- [ ] Verify the textarea re-enables after Claude finishes responding
- [ ] Enable a tool in default permissions (e.g. Read), start a session, trigger a permission approval for another tool, approve it — verify the previously auto-granted tool is no longer re-prompted
- [ ] Verify all existing tests pass (`./check-all.sh`)

 This PR was created with help from Hikari~ 🌸

Reviewed-on: #199
Co-authored-by: Hikari <hikari@nhcarrigan.com>
Co-committed-by: Hikari <hikari@nhcarrigan.com>
2026-03-09 16:53:09 -07:00
hikari 89a0bdd8f1 fix: assorted bug fixes for lists, sounds, interrupts, and permissions (#173)
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 59s
CI / Build Linux (push) Has been cancelled
CI / Build Windows (cross-compile) (push) Has been cancelled
CI / Lint & Test (push) Has been cancelled
## Summary

- **Markdown lists**: Explicitly set `list-style-type: disc` / `decimal` in the Markdown renderer — Tauri's WebView strips browser defaults, leaving bullets and numbers invisible.
- **Notification sounds**: Moved all per-task sounds (success, error, permission, task-start) from a global `characterState` subscription into the per-conversation `claude:state` event handler, so background tabs receive their sounds correctly and tab-switching never replays a sound that already fired. Closes #172
- **Draft text**: Persists `inputValue` per conversation tab so a half-typed prompt survives switching to another tab and back.
- **Interrupt messages**: Replaced vague "Process interrupted" / "Disconnected" strings with source-specific descriptions (keyboard shortcut, stop button, unexpected crash) so it's clear what actually happened.
- **Silent prompt loss**: When Claude Code exits whilst a prompt is in-flight, emits a visible error line telling the user their last prompt was not processed and to reconnect and retry.
- **Double disconnect**: Added an `intentional_stop` flag to `WslBridge` so that `stop()` / `interrupt()` — which kill the process themselves — suppress the duplicate "Disconnected unexpectedly" message that `handle_stdout`'s EOF path was also emitting.
- **Permission modal**: Fixed two cooperating reactivity bugs — `pendingPermissions` was mutated in-place (`.push()`), causing Svelte's derived-store chain to receive the same array reference and skip re-rendering; `PermissionModal.svelte` also used `$state()` (runes mode) where plain `let` is required for correct store-subscription reactivity.

## Test plan

- [ ] Unordered and ordered lists render with visible bullets and numbers in the chat terminal
- [ ] Completion sound plays once when a background tab finishes; switching back to that tab does not replay it
- [ ] Sounds for error, permission request, and task-start also play for background tabs and do not replay on tab switch
- [ ] Typing a prompt, switching tabs, and switching back restores the draft text
- [ ] Pressing Ctrl+C shows "keyboard shortcut (Ctrl+C)"; clicking the stop button shows "via stop button"
- [ ] If Claude exits mid-request, an error message appears prompting the user to resend
- [ ] Clicking stop or pressing Ctrl+C produces exactly one disconnect message (not two)
- [ ] When a tool requires permission, the permission modal appears and the user can approve or dismiss it

 This PR was created with help from Hikari~ 🌸

Reviewed-on: #173
Co-authored-by: Hikari <hikari@nhcarrigan.com>
Co-committed-by: Hikari <hikari@nhcarrigan.com>
2026-02-26 23:34:51 -08:00
hikari bf411adeb7 fix: critical permission modal and config issues (#127)
CI / Lint & Test (push) Has started running
CI / Build Linux (push) Has been cancelled
CI / Build Windows (cross-compile) (push) Has been cancelled
Security Scan and Upload / Security & DefectDojo Upload (push) Has been cancelled
## Summary

This PR resolves several critical bugs that were blocking the permission modal and causing config loss:

- **Permission modal not appearing** - Fixed z-index issues and runtime errors
- **Config store race condition** - Resolved critical race condition causing settings to be lost
- **Excessive logging** - Removed redundant fmt layer that was writing to hidden stdout
- **System tool prompts** - Prevented unnecessary permission prompts for built-in tools
- **Permission batching** - Added support for parallel permission requests
- **ExitPlanMode tool** - Fixed ExitPlanMode tool not functioning correctly

## Changes Made

### Permission Modal Fixes
- Updated z-index to proper value (9999) to ensure modal appears above all other UI elements
- Fixed runtime errors that were preventing modal from rendering
- Resolved issues with permission grants not being properly applied

### Config Store Race Condition
- Fixed critical race condition where multiple rapid config updates would result in lost settings
- Ensured config writes are properly sequenced to prevent data loss
- Added proper synchronisation for config store operations

### Logging Cleanup
- Removed redundant fmt formatting layer that was outputting to hidden stdout
- Cleaned up excessive debug logging added during troubleshooting
- Removed temporary debugging documentation files

### UX Improvements
- Added close confirmation modal with minimise to tray option
- Implemented batching for parallel permission requests
- Added debug console for viewing frontend and backend logs

### ExitPlanMode Fix
- Fixed ExitPlanMode tool not functioning correctly, ensuring proper transitions out of plan mode

## Issues Resolved

Closes #112 - Permission flow now properly handles multiple tool requests
Closes #113 - ExitPlanMode tool now functions correctly
Closes #126 - Debug console feature added (partial - basic implementation complete)

## Test Plan

- [x] Permission modal appears and functions correctly
- [x] Config settings persist across app restarts
- [x] No excessive logging in production builds
- [x] System tools don't trigger permission prompts
- [x] Parallel permission requests are properly batched
- [x] Debug console displays frontend and backend logs
- [x] ExitPlanMode properly exits plan mode

---
 This PR was created with help from Hikari~ 🌸

Co-authored-by: Naomi Carrigan <commits@nhcarrigan.com>
Reviewed-on: #127
Co-authored-by: Hikari <hikari@nhcarrigan.com>
Co-committed-by: Hikari <hikari@nhcarrigan.com>
2026-02-07 01:55:49 -08:00
naomi 852a4d6661 feat: add native clipboard support for screenshot paste (#67)
CI / Lint & Test (push) Successful in 14m34s
CI / Build Linux (push) Successful in 18m19s
CI / Build Windows (cross-compile) (push) Successful in 27m57s
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 1m3s
## Summary
- Adds Tauri clipboard-manager plugin to read images from native clipboard
- Falls back to native clipboard when WebView clipboard API returns empty (fixes screenshot paste)
- Allows sending messages with just attachments (no text required)
- Logs attached files to output with 📎 emoji

## Test plan
- [ ] Build and run the app natively on Windows
- [ ] Copy a screenshot (Win+Shift+S) and paste in the chat input
- [ ] Verify the screenshot appears as an attachment preview
- [ ] Send the attachment and verify Claude receives the file path
- [ ] Test sending a message with only an attachment (no text)
- [ ] Verify the 📎 log line shows the attached filename

**Note:** Paste will not work in WSLg dev environment due to clipboard isolation - needs native Windows build to test.

 This PR was created with help from Hikari~ 🌸

Co-authored-by: Hikari <hikari@nhcarrigan.com>
Reviewed-on: #67
Co-authored-by: Naomi Carrigan <commits@nhcarrigan.com>
Co-committed-by: Naomi Carrigan <commits@nhcarrigan.com>
2026-01-25 13:08:38 -08:00
naomi 3f30997f0e feat: another wave of features (#61)
CI / Build Linux (push) Has been cancelled
CI / Build Windows (cross-compile) (push) Has been cancelled
Security Scan and Upload / Security & DefectDojo Upload (push) Has been cancelled
CI / Lint & Test (push) Has been cancelled
## Explanation

This PR bundles several user-facing improvements and feature additions for the v0.3.0 release, including quality-of-life improvements to the UI, new slash commands, better state persistence, and auto-update checking.

## Included Changes

- **Resizable chat input** with drag handle (#58 partial)
- **Arrow key navigation fix** - cursor keys now navigate text when user has typed input (#58)
- **Scroll position persistence** per conversation tab
- **/skill command** for invoking Claude Code skills (#57)
- **Stats persistence fix** - stats now persist across session changes, only reset on disconnect (#59)
- **Auto-update checker** on startup (#17)
- **Resizable character panel** with full-height sprites (#10)
- **Font size and zoom settings** with keyboard shortcuts (Ctrl++/Ctrl+-/Ctrl+0) (#19)

## Closes

Closes #10, #17, #19, #57, #58, #59

## Attestations

- [x] I have read and agree to the Code of Conduct
- [x] I have read and agree to the Community Guidelines
- [x] My contribution complies with the Contributor Covenant
- [x] I have run the linter and resolved any errors
- [x] My pull request uses an appropriate title, matching the conventional commit standards
- [x] My scope of feat/fix/chore/etc. correctly matches the nature of changes in my pull request
- [x] All new and existing tests pass locally with my changes
- [x] Code coverage remains at or above the configured threshold

## Documentation

N/A - Internal app features

## Versioning

Minor - My pull request introduces new non-breaking features.

---
 This PR was created with help from Hikari~ 🌸

Co-authored-by: Hikari <hikari@nhcarrigan.com>
Reviewed-on: #61
Co-authored-by: Naomi Carrigan <commits@nhcarrigan.com>
Co-committed-by: Naomi Carrigan <commits@nhcarrigan.com>
2026-01-23 19:07:22 -08:00
naomi 06810537a9 feat: add AskUserQuestion tool support (#60)
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 53s
CI / Lint & Test (push) Successful in 14m12s
CI / Build Linux (push) Successful in 16m41s
CI / Build Windows (cross-compile) (push) Successful in 27m0s
## Summary

Implements support for Claude's `AskUserQuestion` tool, allowing Claude to ask the user questions with multiple choice options during a conversation.

## Changes

- Add `UserQuestionEvent` and `QuestionOption` types (Rust and TypeScript)
- Detect `AskUserQuestion` in permission denials and emit `claude:question` event
- Create `UserQuestionModal` component with option selection and custom answer input
- Use stop/reconnect approach (same as `PermissionModal`) since Claude API doesn't accept tool_result for permission-denied tools
- Add `pendingQuestion` to conversation store and `hasQuestionPending` derived store

## Technical Notes

We discovered that Claude Code's permission denial system doesn't allow sending tool results back directly - the API rejects them with "unexpected tool_use_id found in tool_result blocks". The solution was to use the same stop/reconnect pattern that permissions use: stop the session, reconnect with context, and include the user's answer in the context restoration message.

## Test Plan

- [x] Build compiles without errors (Rust + TypeScript)
- [x] Question modal appears when Claude uses `AskUserQuestion`
- [x] Can select options and submit answer
- [x] Answer is properly restored to Claude after reconnect

Closes #51

---

 This PR was created with help from Hikari~ 🌸

Co-authored-by: Hikari <hikari@nhcarrigan.com>
Reviewed-on: #60
2026-01-23 14:11:18 -08:00
naomi 947e56ef41 feat: naomi did too much at once (#53)
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 53s
CI / Lint & Test (push) Successful in 14m10s
CI / Build Linux (push) Successful in 16m47s
CI / Build Windows (cross-compile) (push) Successful in 26m36s
- feat: add slash commands
- feat: toggle window always on top
- fix: save settings button closes settings panel
- feat: input history (both text and commands)
- feat: add keyboard shortcuts
- feat: add confirmation modal when closing connected tabs
- fix: better text colours in light mode
- fix: handle multiple tabs requesting permission

Closes #6
Closes #13
Closes #21
Closes #28

Reviewed-on: #53
Co-authored-by: Naomi Carrigan <commits@nhcarrigan.com>
Co-committed-by: Naomi Carrigan <commits@nhcarrigan.com>
2026-01-21 17:38:36 -08:00
naomi d83697e5cf feat: add ability to run multiple agents via tabbed views (#47)
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 54s
CI / Lint & Test (push) Successful in 14m18s
CI / Build Linux (push) Successful in 16m46s
CI / Build Windows (cross-compile) (push) Successful in 26m39s
### Explanation

_No response_

### Issue

Closes #30 Closes #41

### Attestations

- [ ] I have read and agree to the [Code of Conduct](https://docs.nhcarrigan.com/community/coc/)
- [ ] I have read and agree to the [Community Guidelines](https://docs.nhcarrigan.com/community/guide/).
- [ ] My contribution complies with the [Contributor Covenant](https://docs.nhcarrigan.com/dev/covenant/).

### Dependencies

- [ ] I have pinned the dependencies to a specific patch version.

### Style

- [ ] I have run the linter and resolved any errors.
- [ ] My pull request uses an appropriate title, matching the conventional commit standards.
- [ ] My scope of feat/fix/chore/etc. correctly matches the nature of changes in my pull request.

### Tests

- [ ] My contribution adds new code, and I have added tests to cover it.
- [ ] My contribution modifies existing code, and I have updated the tests to reflect these changes.
- [ ] All new and existing tests pass locally with my changes.
- [ ] Code coverage remains at or above the configured threshold.

### Documentation

_No response_

### Versioning

_No response_

Reviewed-on: #47
Co-authored-by: Naomi Carrigan <commits@nhcarrigan.com>
Co-committed-by: Naomi Carrigan <commits@nhcarrigan.com>
2026-01-20 13:57:48 -08:00
naomi 2d3adcab1c feat: add chat modes and interrupt feature (#46)
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 52s
CI / Lint & Test (push) Successful in 14m15s
CI / Build Linux (push) Successful in 16m37s
CI / Build Windows (cross-compile) (push) Successful in 26m35s
### Explanation

_No response_

### Issue

Closes #40

### Attestations

- [ ] I have read and agree to the [Code of Conduct](https://docs.nhcarrigan.com/community/coc/)
- [ ] I have read and agree to the [Community Guidelines](https://docs.nhcarrigan.com/community/guide/).
- [ ] My contribution complies with the [Contributor Covenant](https://docs.nhcarrigan.com/dev/covenant/).

### Dependencies

- [ ] I have pinned the dependencies to a specific patch version.

### Style

- [ ] I have run the linter and resolved any errors.
- [ ] My pull request uses an appropriate title, matching the conventional commit standards.
- [ ] My scope of feat/fix/chore/etc. correctly matches the nature of changes in my pull request.

### Tests

- [ ] My contribution adds new code, and I have added tests to cover it.
- [ ] My contribution modifies existing code, and I have updated the tests to reflect these changes.
- [ ] All new and existing tests pass locally with my changes.
- [ ] Code coverage remains at or above the configured threshold.

### Documentation

_No response_

### Versioning

_No response_

Reviewed-on: #46
Co-authored-by: Naomi Carrigan <commits@nhcarrigan.com>
Co-committed-by: Naomi Carrigan <commits@nhcarrigan.com>
2026-01-20 08:33:39 -08:00
naomi c241544743 feat(tools): set up proper CI (#2)
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 57s
CI / Lint & Test (push) Successful in 14m1s
CI / Build Linux (push) Successful in 16m8s
CI / Build Windows (cross-compile) (push) Successful in 26m18s
### Explanation

_No response_

### Issue

_No response_

### Attestations

- [ ] I have read and agree to the [Code of Conduct](https://docs.nhcarrigan.com/community/coc/)
- [ ] I have read and agree to the [Community Guidelines](https://docs.nhcarrigan.com/community/guide/).
- [ ] My contribution complies with the [Contributor Covenant](https://docs.nhcarrigan.com/dev/covenant/).

### Dependencies

- [ ] I have pinned the dependencies to a specific patch version.

### Style

- [ ] I have run the linter and resolved any errors.
- [ ] My pull request uses an appropriate title, matching the conventional commit standards.
- [ ] My scope of feat/fix/chore/etc. correctly matches the nature of changes in my pull request.

### Tests

- [ ] My contribution adds new code, and I have added tests to cover it.
- [ ] My contribution modifies existing code, and I have updated the tests to reflect these changes.
- [ ] All new and existing tests pass locally with my changes.
- [ ] Code coverage remains at or above the configured threshold.

### Documentation

_No response_

### Versioning

_No response_

Reviewed-on: #2
Co-authored-by: Naomi Carrigan <commits@nhcarrigan.com>
Co-committed-by: Naomi Carrigan <commits@nhcarrigan.com>
2026-01-15 20:06:47 -08:00
naomi f393dfb359 feat: initial prototype
Security Scan and Upload / Security & DefectDojo Upload (push) Successful in 47s
2026-01-14 20:56:28 -08:00