- Explain AgentMessage vs LLM Message separation and why it's needed
- Document conversion flow: AgentMessage -> transformContext -> convertToLlm -> LLM
- Clarify that prompt/queueMessage/continue must convert to user or toolResult
- Document message_start/end events for prompt() and queued messages
- Explain message_update is assistant-only with partial content
- Add pattern for handling partial messages in reactive UIs
- Document agent.state.streamMessage for accessing current partial
- Removed Agent API section from pi-ai README (moved to agent package)
- Rewrote agent package README for new architecture:
- No more transports (ProviderTransport, AppTransport removed)
- Uses streamFn directly with streamProxy for proxy usage
- Documents convertToLlm and transformContext
- Documents low-level agentLoop/agentLoopContinue API
- Updated custom message types documentation
- Renamed AppMessage to AgentMessage throughout
- New agent-loop.ts with AgentLoopContext, AgentLoopConfig
- Removed transport abstraction, Agent now takes streamFn directly
- Extracted streamProxy to proxy.ts utility
- Removed agent-loop from pi-ai (now in agent package)
- Updated consumers (coding-agent, mom) for AgentMessage rename
- Tests updated but some consumers still need migration
Known issues:
- AgentTool, AgentToolResult not exported from pi-ai
- Attachment not exported from pi-agent-core
- ProviderTransport removed but still referenced
- messageTransformer -> convertToLlm migration incomplete
- CustomMessages declaration merging not working properly
HookMessage (role: hookMessage) was being passed directly to transport
without transformation. Now it's transformed via messageTransformer
which converts it to a proper user message for the LLM.
HookMessage.content can be string or array, but LLM Message.content
must be an array. This was causing 'messages: at least one message
is required' errors when hooks sent string content.
- Rename HookAppMessage to HookMessage, isHookAppMessage to isHookMessage
- Remove entries array from SessionContext (use isHookMessage type guard instead)
- HookMessage.content now accepts string directly (not just array)
- Fix streamMessage type in AgentState (AppMessage, not Message)
- Rename CustomMessageComponent to HookMessageComponent
- Fix test hook to use pi.sendMessage
- Change from contextTransform (runs once at agent start) to preprocessor
- preprocessor runs before EACH LLM call inside the agent loop
- ContextEvent now uses Message[] (pi-ai format) instead of AppMessage[]
- Deep copy handled by pi-ai preprocessor, not Agent
This enables:
- Pruning rules applied on every turn (not just agent start)
- /prune during long agent loop takes effect immediately
- Compaction can use same transforms (future work)
- Add contextTransform option to Agent (runs before messageTransformer)
- Deep copy messages before passing to contextTransform (modifications are ephemeral)
- Add ContextEvent and ContextEventResult types
- Add emitContext() to HookRunner (chains multiple handlers)
- Wire up in sdk.ts when creating Agent with hooks
Enables dynamic context pruning: hooks can modify messages sent to LLM
without changing session data. See discussion #330.
- Full box border around title, game area, and instructions
- ESC pauses and saves state to session via pi.appendEntry()
- Resume shows 'PAUSED - press any key to continue'
- Q quits and clears saved state
- High score persists across games
- Rounded box corners (╭╮╰╯)
- Better characters: ● head, ○ body, ◆ food, · empty
- Colored title with emoji
- Dimmed borders and help text
- Bold highlights for score and controls
- Extract execCommand to src/core/exec.ts, shared by hooks and custom-tools
- Rename CustomMessageRenderer -> HookMessageRenderer
- Rename registerCustomMessageRenderer -> registerMessageRenderer
- Renderer now receives HookMessage instead of CustomMessageEntry
- CustomMessageComponent now has setExpanded() and responds to Ctrl+E toggle
- Re-export ExecOptions/ExecResult from exec.ts for backward compatibility
- compaction-hooks-example.test.ts: get sessionManager/modelRegistry from ctx
- compaction-hooks.test.ts:
- Pass sessionManager/modelRegistry to HookRunner constructor
- Remove setSessionFile call
- Update tests to use session.sessionManager instead of event.sessionManager
- exec() moves from HookEventContext/HookCommandContext to HookAPI
- sessionManager/modelRegistry move from SessionEventBase to HookEventContext
- HookCommandContext keeps sessionManager/modelRegistry (command handlers need them)
- Both sendMessage and exec accessed via pi closure in command handlers
Cleans up the temporary emitLastMessage plumbing since we now use
Agent.prompt(AppMessage) for hook messages instead of appendMessage+continue.
- Remove emitLastMessage parameter from Agent.continue()
- Remove from transport interface and implementations
- Remove from agentLoopContinue()
Instead of using continue() which validates roles, prompt() now accepts
an AppMessage directly. This allows hook messages with role: 'hookMessage'
to trigger proper agent loop with message events.
- Add overloads: prompt(AppMessage) and prompt(string, attachments?)
- sendHookMessage uses prompt(appMessage) instead of appendMessage+continue
Following the same pattern as BashExecutionMessage:
- HookAppMessage has role: 'hookMessage' with customType, content, display, details
- isHookAppMessage() type guard for checking message type
- messageTransformer converts to user message for LLM context
- TUI checks isHookAppMessage() for rendering as CustomMessageComponent
This makes the API clean for anyone building on AgentSession - they can
use the type guard instead of knowing about internal marker fields.
When calling continue() with emitLastMessage=true, the agent loop
emits message_start/message_end events for the last message in context.
This allows messages added outside the loop (e.g., hook messages via
sendHookMessage) to trigger proper TUI rendering.
Changes across packages:
- packages/ai: agentLoopContinue() accepts emitLastMessage parameter
- packages/agent: Agent.continue(), transports updated to pass flag
- packages/coding-agent: sendHookMessage passes true when triggerTurn
- Command handler now returns Promise<void> instead of Promise<string | undefined>
- To trigger LLM response, use sendMessage() with triggerTurn: true
- Simplify _tryExecuteHookCommand to return boolean
Added example hook and slash command in .pi/:
- .pi/hooks/test-command.ts - /greet command using sendMessage
- .pi/commands/review.md - file-based /review command
Hook commands registered via pi.registerCommand() are now handled in
AgentSession.prompt() alongside file-based slash commands. This:
- Removes duplicate tryHandleHookCommand from interactive-mode and rpc-mode
- All modes (interactive, RPC, print) share the same command handling logic
- AgentSession._tryExecuteHookCommand() builds CommandContext using:
- UI context from hookRunner (set by mode)
- sessionManager, modelRegistry from AgentSession
- sendMessage via sendHookMessage
- exec via exported execCommand
- Handler returning string uses it as prompt, undefined returns early
Also:
- Export execCommand from hooks/runner.ts
- Add getUIContext() and getHasUI() to HookRunner
- Make HookRunner.emitError() public for error reporting
Breaking changes to Hook API:
- pi.send(text, attachments?) replaced with pi.sendMessage(message, triggerTurn?)
- Creates CustomMessageEntry instead of user messages
- Properly handles queuing during streaming via agent loop
- Supports optional turn triggering when idle
- New pi.appendEntry(customType, data?) for hook state persistence
- New pi.registerCommand(name, options) for custom slash commands
- Handler types renamed: SendHandler -> SendMessageHandler, new AppendEntryHandler
Implementation:
- AgentSession.sendHookMessage() handles all three cases:
- Streaming: queues message with _hookData marker, agent loop processes it
- Not streaming + triggerTurn: appends to state/session, calls agent.continue()
- Not streaming + no trigger: appends to state/session only
- message_end handler routes based on _hookData presence to correct persistence
- HookRunner gains getRegisteredCommands() and getCommand() methods
New types: HookMessage<T>, RegisteredCommand, CommandContext
- CustomMessageComponent accepts optional CustomMessageRenderer
- If hook provides a renderer, call it and use returned Component inside Box
- Falls back to default rendering (label + Markdown) if no renderer or null returned
- renderSessionContext gets renderer from hookRunner and passes to component
- Add getMessageFromEntry helper to extract AppMessage from any context-producing entry
- Update findValidCutPoints to treat branch_summary/custom_message as valid cut points
- Update findTurnStartIndex to recognize branch_summary/custom_message as turn starters
- Update all message extraction loops to use getMessageFromEntry
- Add renderCustomMessage to HookAPI for registering custom renderers
- Add CustomMessageRenderer type and CustomMessageRenderOptions
- Store customMessageRenderers in LoadedHook
- Add getCustomMessageRenderer(customType) to HookRunner
- SessionContext.entries now aligned with messages (same length, corresponding indices)
TUI can now correlate messages with their source entries to identify
custom_message entries and use hook-provided renderers.
content: string | (TextContent | ImageContent)[]
This matches the UserMessage type from pi-ai, so content can be
passed directly to AppMessage without conversion.
- CustomMessageEntry<T> type with customType, content, display, details
- appendCustomMessageEntry() in SessionManager
- buildSessionContext() includes custom_message entries as user messages
- Exported CustomEntry and CustomMessageEntry from main index
CustomEntry is for hook state (not in context).
CustomMessageEntry is for hook-injected content (in context).
- CompactionEntry<T> and CompactionResult<T> now have optional details?: T
- appendCompaction() accepts optional details parameter
- Hooks can return compaction.details to store custom data
- Enables structured compaction with ArtifactIndex (see #314)
- Fix CompactionResult export location (now from compaction.ts)
- Update plan with remaining compaction refactor items
- Import CompactionResult in hooks/types.ts
- Replace inline type with CompactionResult for SessionEventResult.compaction
- Add labels feature to changelog
- Add LabelEntry type with targetId and label (string | undefined)
- Add labelsById map built on load via linear scan
- Add getLabel(id) and appendLabelChange(targetId, label) methods
- Add label field to SessionTreeNode, populated by getTree()
- Update createBranchedSession to preserve labels for entries on path
- Labels are ignored by buildSessionContext (not sent to LLM)
- Add comprehensive tests for label functionality
