- Event-based architecture: agent forwards AgentSessionEvent to adapters
- Multi-adapter support via single config.json
- Access control: admins, DM permissions per adapter
- Channel namespacing by adapter name
- Channel isolation via bubblewrap in Linux/Docker environments
- Simplified directory structure and interfaces
- Fix //tmp issue: distinguish slash commands from file paths by checking if anything precedes the /
- Fix symlinks to directories (like /tmp) not getting trailing slash
- Add setOAuthStorage() and resetOAuthStorage() to pi-ai for custom storage backends
- Configure coding-agent to use its own configurable OAuth path via getOAuthPath()
- Model selector (/model command) now only shows models from --models scope when set
- Rewrite OAuth documentation in pi-ai README with examples
Fixes#255
When a user interrupts a tool call flow (sends a message without providing
tool results), APIs like OpenAI Responses and Anthropic fail because:
- OpenAI requires tool outputs for function calls
- OpenAI requires reasoning items to have their following items
- Anthropic requires non-empty content for error tool results
Instead of filtering out orphaned tool calls (which breaks thinking signatures),
we now insert synthetic empty tool results with isError: true and content
'No result provided'. This preserves the conversation structure and satisfies
all API requirements.
Use correct API response format (cloudaicompanionProject) and proper
onboarding flow with tierId and metadata. Add retry logic for project
provisioning which may take time.
- Add OAuth handler with PKCE flow and local callback server
- Automatic project discovery via loadCodeAssist/onboardUser endpoints
- Store credentials with projectId for API calls
- Encode token+projectId as JSON for provider to decode
- Register as 'google-cloud-code-assist' OAuth provider
- Add new API type 'google-cloud-code-assist' for Gemini CLI / Antigravity auth
- Extract shared Google utilities to google-shared.ts
- Implement streaming provider for Cloud Code Assist endpoint
- Add 7 models: gemini-3-pro-high/low, gemini-3-flash, claude-sonnet/opus, gpt-oss
Models use OAuth authentication and have sh cost (uses Google account quota).
OAuth flow will be implemented in coding-agent in a follow-up.
- setThinkingLevel() now clamps xhigh to high when model doesn't support it
- Model changes automatically re-clamp the current thinking level
- Fixed /model command to use session.setModel() instead of agent.setModel()
- Footer and editor border color update after model/thinking changes
Closes#253
The cell size query response parser was incorrectly holding back Kitty
keyboard protocol sequences, thinking they might be incomplete cell size
responses. This caused Ctrl+C/Ctrl+D to require multiple presses in iTerm2.
Fixed by only waiting for more data if the buffer doesn't end with a
terminal escape sequence terminator character.
Fixed arrow key and Enter detection in selector components to work
with Kitty protocol when Caps Lock or Num Lock is enabled.
Updated: oauth-selector, user-message-selector, hook-selector,
hook-input, model-selector, session-selector
Extended the Kitty protocol parser to handle functional keys with
~ terminator (Delete, Insert, PageUp, PageDown) and Home/End keys.
Updated editor.ts and input.ts to use the new helpers.
