- Add resetApiProviders() to clear and re-register built-in providers
- Add createAssistantMessageEventStream() factory for extensions
- Add streamSimple support in ProviderConfig for custom API implementations
- Call resetApiProviders() on /reload to clean up extension providers
- Add custom-provider.md documentation
- Add custom-provider.ts example with full Anthropic implementation
- Update extensions.md with streamSimple config option
- Package deduplication: same package in global+project, project wins
- Collision detection for skills, prompts, and themes with ResourceCollision type
- PathMetadata tracking with parent directory lookup for file paths
- Display improvements: section headers, sorted groups, accent colors for packages
- pi list shows full paths below package names
- Extension loader discovers files in directories without index.ts
- In-memory SettingsManager properly tracks project settings
fixes#645
- Add ParsedSkillBlock interface and parseSkillBlock() function
- Change skill expansion to use XML-style <skill> tags
- Add SkillInvocationMessageComponent for collapsible display
- Collapsed: single line with skill name and expand hint
- User message rendered separately after skill block
Fixes#894
There's only ever one bindings instance per session, so the Set/Array
approach was unnecessary. Changed from Set<ExtensionErrorListener> to
optional single listener.
- Add ResourceLoader interface and DefaultResourceLoader implementation
- Add PackageManager for npm/git extension sources with install/remove/update
- Add session.reload() and session.bindExtensions() APIs
- Add /reload command in interactive mode
- Add CLI flags: --skill, --theme, --prompt-template, --no-themes, --no-prompt-templates
- Add pi install/remove/update commands for extension management
- Refactor settings.json to use arrays for skills, prompts, themes
- Remove legacy SkillsSettings source flags and filters
- Update SDK examples and documentation for ResourceLoader pattern
- Add theme registration and loadThemeFromPath for dynamic themes
- Add getShellEnv to include bin dir in PATH for bash commands
Detect OAuth authentication failures (expired credentials, offline) and provide helpful error message instead of crashing with generic 'No API key found' error.
Co-authored-by: Mario Zechner <badlogicgames@gmail.com>
When auto-compaction fails (e.g., quota exceeded), emit the error via
the auto_compaction_end event instead of throwing. The UI now displays
the error message, allowing users to take action (switch models, wait
for quota reset, etc.) instead of crashing.
fixes#792
When 429/500 errors occur during tool execution, empty assistant messages
with stopReason='error' get persisted. These break the tool_use -> tool_result
chain for Claude/Gemini APIs.
Added centralized filtering in transformMessages to skip assistant messages
with empty content and no tool calls. Provider-level filters remain for
defense-in-depth.
- Skill commands (/skill:name) now expanded in AgentSession instead of
interactive mode, enabling them in RPC and print modes
- Input event can now intercept /skill:name before expansion
- Updated extensions.md with clearer input event docs and processing order
- Updated rpc.md: hook -> extension terminology, added skill expansion mentions
- Added PR attribution to changelog entries for #761
* feat(coding-agent): add input event for extension input interception
Extensions can now intercept, transform, or handle user input before the
agent processes it. Three result types: continue (pass through), transform
(modify text/images), handled (respond without LLM). Handlers chain
transforms and short-circuit on handled. Source field identifies origin.
* fix: make source public, use if/else over ternary
* fix: remove response field, extension handles own UI
* Improve Cloud Code Assist error messages
- Extract just the message from verbose JSON error responses
- Extract cause from generic 'fetch failed' errors for better diagnostics
* Make 'other side closed' network error retryable
* Make 'other side closed' network error retryable
Breaking change: pi.getAllTools() now returns Array<{ name, description }>
instead of string[]. Extensions needing just names can use .map(t => t.name).
Removes redundant getToolInfo() method added in original PR.
Fixes#647
- New /models command with toggle UI for each available model
- Changes persist to enabledModels in settings.json
- Updates take effect immediately for Ctrl+P cycling
The --no-skills flag set options.skills = [] in main.ts, but the
interactive mode UI would rediscover skills anyway because it called
loadSkills() directly instead of using the already-loaded skills.
Changes:
- Add AgentSession.skills and AgentSession.skillWarnings properties
- discoverSkills() now returns { skills, warnings } instead of Skill[]
- Interactive mode uses session.skills instead of calling loadSkills()
- Update SDK docs and examples for new return type
Fixes#577
- user_bash event for intercepting ! and !! commands (#528)
- Extensions can return { operations } or { result } to redirect/replace
- executeBashWithOperations() for custom BashOperations execution
- session.recordBashResult() for extensions handling bash themselves
- Theme API: getAllThemes(), getTheme(), setTheme() on ctx.ui
- mac-system-theme.ts example: sync with macOS dark/light mode
- Updated ssh.ts to use user_bash event
- Add systemPrompt to BeforeAgentStartEvent so extensions can see current prompt
- Change systemPromptAppend to systemPrompt in BeforeAgentStartEventResult for full replacement
- Extensions can now chain modifications (each sees the result of previous)
- Update ssh.ts to replace local cwd with remote cwd in system prompt
- Update pirate.ts, claude-rules.ts, preset.ts to use new API
fixes#575
When aborting a retry attempt, surface the retry-aware abort message
in the assistant output and tool results instead of a generic "Aborted".
- Set errorMessage for aborted streaming messages
- Render abort message without forcing a leading newline when no content
Three related fixes:
1. google-gemini-cli: Handle abort signal in stream reading loop
- Add abort event listener to cancel reader immediately when signal fires
- Fix AbortError detection in retry catch block (fetch throws AbortError,
not our custom message)
- Swallow reader.cancel() rejection to avoid unhandled promise
2. agent-session: Fix retry attempt counter showing 0 on cancel
- abortRetry() was resetting _retryAttempt before the catch block could
read it for the error message
3. interactive-mode: Restore main escape handler on agent_start
- When auto-retry starts, onEscape is replaced with retry-specific handler
- auto_retry_end (which restores it) fires on turn_end, after streaming begins
- Now restore immediately on agent_start if retry handler is still active
Amended: suppress reader.cancel() rejection on abort.
Adds sendUserMessage() to the extension API, allowing extensions to send
actual user messages (role: user) rather than custom messages. Unlike
sendMessage(), this always triggers a turn and behaves as if the user
typed the message.
- Add SendUserMessageHandler type and sendUserMessage() to ExtensionAPI
- Wire handler through loader, runner, and all modes
- Implement via prompt() with expandPromptTemplates: false
- Add send-user-message.ts example with /ask, /steer, /followup commands
- Document in extensions.md
fixes#483
- Add sessionId to StreamOptions for providers that support session-based caching
- OpenAI Codex provider uses sessionId for prompt_cache_key and routing headers
- Agent class now accepts and forwards sessionId to stream functions
- coding-agent passes session ID from SessionManager and updates on session changes
- Update ai package README with table of contents, OpenAI Codex OAuth docs, and env vars table
- Increase Codex instructions cache TTL from 15 minutes to 24 hours
- Add tests for sessionId forwarding in ai and agent packages
- Remove per-thinking-level model variants (gpt-5.2-codex-high, etc.)
- Remove thinkingLevels from Model type
- Provider clamps reasoning effort internally
- Omit reasoning field when thinking is off
fixes#472
Breaking changes:
- Settings: 'hooks' and 'customTools' arrays replaced with 'extensions'
- CLI: '--hook' and '--tool' flags replaced with '--extension' / '-e'
- API: HookMessage renamed to CustomMessage, role 'hookMessage' to 'custom'
- API: FileSlashCommand renamed to PromptTemplate
- API: discoverSlashCommands() renamed to discoverPromptTemplates()
- Directories: commands/ renamed to prompts/ for prompt templates
Migration:
- Session version bumped to 3 (auto-migrates v2 sessions)
- Old 'hookMessage' role entries converted to 'custom'
Structural changes:
- src/core/hooks/ and src/core/custom-tools/ merged into src/core/extensions/
- src/core/slash-commands.ts renamed to src/core/prompt-templates.ts
- examples/hooks/ and examples/custom-tools/ merged into examples/extensions/
- docs/hooks.md and docs/custom-tools.md merged into docs/extensions.md
New test coverage:
- test/extensions-runner.test.ts (10 tests)
- test/extensions-discovery.test.ts (26 tests)
- test/prompt-templates.test.ts
- event-bus.ts: await async handlers to catch errors properly
- agent-session.ts: clear _pendingNextTurnMessages on newSession/switchSession/branch
- sdk.ts: make eventBus first (required) param for discoverHooks/discoverCustomTools
- docs/sdk.md: document eventBus sharing pattern for hook/tool communication
* feat(coding-agent): add event bus for tool/hook communication
Adds pi.events API enabling custom tools and hooks to communicate via
pub/sub. Tools can emit events, hooks can listen. Shared EventBus instance
created per session in createAgentSession().
- EventBus interface with emit() and on() methods
- on() returns unsubscribe function
- Threaded through hook and tool loaders
- Documented in hooks.md and custom-tools.md
* fix(coding-agent): wrap event handlers to catch errors
* docs: note async handler error handling for event bus
* feat(coding-agent): add sendMessage to tools, nextTurn delivery mode
- Custom tools now have pi.sendMessage() for direct agent notifications
- New deliverAs: 'nextTurn' queues messages for next user prompt
- Fix: hooks and tools now share the same eventBus (was isolated before)
* fix(coding-agent): nextTurn delivery should always queue, even when streaming
- before_agent_start handlers can return systemPromptAppend to dynamically
append text to the system prompt for that turn
- Multiple hooks' systemPromptAppend strings are concatenated
- Multiple hooks' messages are now all injected (not just first)
- Tool registry now contains ALL built-in tools (read, bash, edit, write,
grep, find, ls) regardless of --tools flag
- --tools only sets initially active tools, hooks can enable any via
setActiveTools()
- System prompt automatically rebuilds when tools change, updating tool
descriptions and guidelines
- Add pirate.ts example hook demonstrating systemPromptAppend
- Update hooks.md with systemPromptAppend documentation
- New text_delta hook event fires for each chunk of streaming text
- Enables real-time monitoring of agent output
- Plan-mode hook now updates todo progress as [DONE:id] tags stream in
- Each todo item has unique ID for reliable tracking
- Add pi.getTools() and pi.setTools(toolNames) to HookAPI
- Hooks can now enable/disable tools dynamically
- Changes take effect on next agent turn
New example hook: plan-mode.ts
- Claude Code-style read-only exploration mode
- /plan command toggles plan mode on/off
- Plan mode tools: read, bash, grep, find, ls
- Edit/write tools disabled in plan mode
- Injects context telling agent about restrictions
- After each response, prompts to execute/stay/refine
- State persists across sessions
- Hook commands now execute immediately during streaming (they manage their own LLM interaction via pi.sendMessage())
- File-based slash commands are expanded and queued via steer/followUp during streaming
- prompt() accepts new streamingBehavior option ('steer' or 'followUp') for explicit queueing during streaming
- steer() and followUp() now expand file-based slash commands and error on hook commands
- RPC prompt command accepts optional streamingBehavior field
- Updated docs: rpc.md, sdk.md, CHANGELOG.md
fixes#420
Use !!command to execute bash commands that are shown in the TUI and
saved to session history but excluded from LLM context, compaction
summaries, and branch summaries.
- Add excludeFromContext field to BashExecutionMessage
- Filter excluded messages in convertToLlm()
- Parse !! prefix in interactive mode
- Use dim border color for excluded commands
fixes#414
- Rename queueMessage to steer(), add followUp()
- Split _pendingMessages into _steeringMessages and _followUpMessages
- Update sendHookMessage to accept deliverAs option
- Rename hasQueuedMessages to hasPendingMessages
- Rename queuedMessageCount to pendingMessageCount
- Update clearQueue() return type to { steering, followUp }
- Update UI to show steering vs follow-up messages differently
WIP: settings-manager, sdk, interactive-mode, rpc-mode still need updates
Agent.prompt() and Agent.continue() now throw if called while already
streaming, preventing race conditions and corrupted state. Use
queueMessage() to queue messages during streaming, or await the
previous call.
AgentSession.prompt() has the same guard with a message directing
users to queueMessage().
Ref #403
CustomToolContext now has:
- isIdle() - check if agent is streaming
- hasQueuedMessages() - check if user has queued messages
- abort() - abort current operation (fire-and-forget)
Changed abort() signature from Promise<void> to void in both
HookContext and CustomToolContext. The abort is fire-and-forget:
it calls session.abort() without awaiting, so the abort signal
is set immediately while waitForIdle() runs in the background.
Fixes#388
HookContext (all events):
- isIdle() - read-only state check
- hasQueuedMessages() - read-only state check
- abort() - fire-and-forget, does not wait
HookCommandContext (slash commands only):
- waitForIdle() - waits for agent to finish
- newSession(options?) - create new session
- branch(entryId) - branch from entry
- navigateTree(targetId, options?) - navigate session tree
Session control methods moved from HookAPI (pi.*) to HookCommandContext (ctx.*)
because they can deadlock when called from event handlers that run inside
the agent loop (tool_call, tool_result, context events).
HookAPI additions:
- pi.newSession(options?) - create new session with optional setup callback
- pi.branch(entryId) - branch from a specific entry
- pi.navigateTree(targetId, options?) - navigate the session tree
HookContext additions:
- ctx.isIdle() - check if agent is streaming
- ctx.waitForIdle() - wait for agent to finish
- ctx.abort() - abort current operation
- ctx.hasQueuedMessages() - check for queued user messages
These enable hooks to programmatically manage sessions (handoff, templates)
and check agent state before showing interactive UI.
Fixes#388
- Remove session_before_new and session_new hook events
- Add reason: 'new' | 'resume' to session_before_switch and session_switch events
- Remove 'new' reason from custom tool onSession (use 'switch' for both /new and /resume)
- Rename reset() to newSession(options?) in AgentSession
- Add NewSessionOptions with optional parentSession for lineage tracking
- Rename branchedFrom to parentSession in SessionHeader
- Rename RPC reset command to new_session with optional parentSession
- Update example hooks to use new event structure
- Update documentation and changelog
Based on discussion in #293
- Add tree sidebar with search and filter (Default/All/Labels)
- Client-side markdown/syntax highlighting via vendored marked.js + highlight.js
- Base64 encode session data to avoid escaping issues
- Reuse theme.ts color tokens via getResolvedThemeColors()
- Sticky sidebar, responsive mobile layout with overlay
- Click tree node to scroll to message
- Keyboard shortcuts: Esc to reset, Ctrl/Cmd+F to search
