fix: extract text from TG Rich Message for bot-to-bot communication#92
Open
changhansung wants to merge 332 commits into
Open
fix: extract text from TG Rich Message for bot-to-bot communication#92changhansung wants to merge 332 commits into
changhansung wants to merge 332 commits into
Conversation
changhansung
commented
Jun 17, 2026
changhansung
commented
Contributor
- Add extractRichText() helper that recursively walks RichBlock/RichText nodes
- When msg.text is empty but msg.rich_message exists, extract plain text from blocks
- Remove TG-DEBUG console.log statements
- Enables bot-to-bot @mention to work when sender uses sendRichMessage
- daemon.ts: emit message_queued (when pasteQueueDepth > 0) and message_delivered (after deliverMessage) with chatId + messageId - instance-lifecycle.ts: listen for both events, call ctx.reactMessageStatus - fleet-manager.ts: reactMessageStatus reacts on the user's message, gated to Discord adapter only
- /chat respond now returns the bot reply message ID - Pass channelId as chatId + reply messageId to daemon meta - Optimize react() to try chatId as channelId first (fast path) - Fixes: ⏳/✅ reactions not showing because messageId was empty
…sy TUI When multiple messages are queued and the TUI is busy outputting, the first send-keys Enter after paste-buffer can be swallowed. Add a retry Enter after 1s delay. All supported CLIs (kiro-cli, claude-code, gemini-cli, codex, opencode) ignore Enter on empty input, making the retry safe.
…allow Root cause: with 10s/30s timeout, waitForIdle expires while the agent is still outputting. The forced paste succeeds but Enter is swallowed by the busy TUI. With 120s, paste only happens when the CLI has been silent for 2s (truly idle), so Enter is reliably accepted. The retry Enter in pasteText remains as a safety net for the rare case where the 120s timeout expires.
Part 1: restart_instance MCP tool - outbound-schemas.ts: RestartInstanceArgs (name only) - outbound-handlers.ts: restartInstance handler calling ctx.restartSingleInstance - OutboundContext: add restartSingleInstance method - mcp-tools.ts: register restart_instance tool - daemon.ts: add to CROSS_INSTANCE_TOOLS set Part 2: lightweight waitForIdle - deliverMessage: lightweight 30s, normal 120s (was 120s for both)
After /start registers a new classic channel, setOpenChannels on the adapter wasn't updated. Messages from non-primary guilds to the new channel were silently dropped because the channel wasn't in openChannels. This also affected attachment URL storage — images couldn't be downloaded.
1. handleClassicStop: add reregisterClassicChannels() after unregister to remove stopped channel from adapter openChannels. 2. reregisterClassicChannels: always call setOpenChannels even when channels list is empty, so the last /stop properly clears all stale entries from the adapter.
No admin required. Reads context from statusline.json first, falls back to tmux capture-pane with kiro-cli regex parser. Shows context percentage, backend, and instance name.
OpenCode shows 'Permission required' when accessing external directories, followed by a 'confirm' prompt. Add getRuntimeDialogs to auto-press Enter on both layers.
Runtime dialogs (Permission required, confirm) need fast response. 30s meant each prompt waited up to 30s before auto-dismiss. 5s keeps CPU impact minimal (single capture-pane per cycle).
Select 'Allow always' instead of 'Allow once' so the same directory won't prompt again. Added Right/Left to sendSpecialKey and dialog handler SPECIAL_KEYS set.
Kiro CLI 2.0 TUI mode uses ~1.1GB per instance. Classic mode uses ~300MB. Add --classic flag to kiro-cli chat command.
1. classic-channel-manager.ts:
- ClassicChannel.collab field (persisted to classicBot.yaml)
- toggleCollab(channelId): toggle + save, returns new state
- isCollab(channelId): check collab status
2. discord-adapter.ts:
- Register /collab slash command
- Only block own messages (not all bots) — let fleet-manager decide
- Pass isBotMessage flag in message emit
- Emit bot user ID with 'started' event
3. channel/types.ts:
- InboundMessage.isBotMessage?: boolean
4. fleet-manager.ts:
- /collab handler: admin-only toggle, responds with new state
- handleInboundMessage: bot messages only pass through to collab channels
- handleClassicChannelMessage collab mode:
- Log all messages (including bots) to chat-logs
- Trigger on exact @mention (<@BOT_USER_ID>) only
- @everyone/@here/@ROLE don't trigger (not <@id> format)
- Strip @mention, forward clean text to agent
- Store botUserId from adapter 'started' event
Part 1: pre_task_command - types.ts: InstanceConfig.pre_task_command?: string - classic-channel-manager.ts: ClassicChannel.preTaskCommand, load/save, getPreTaskCommand(channelId) - daemon.ts: in pasteLock, deliverMessage(pre_task_command) before the actual message (waits for idle between both) - fleet-manager.ts: startClassicInstance passes preTaskCommand from classicChannels config into InstanceConfig Part 2: General steering - GENERAL_INSTRUCTIONS: add 'Instance Configuration Tips' section with reviewer pre_task_command, collab mode, and cost_guard suggestions
Fleet start syncs src/general-knowledge/*.md to General instance's .kiro/steering/ directory. Knowledge files are version-controlled — updating AgEnD automatically gives General new skills. Includes initial skills.md: tmux health check, reviewer session management, fork instance (session cloning).
Previously, any instructions change (including knowledge file sync) forced skipResume=true, losing conversation context on every restart. Now: resume is always preserved. When instructions change, a [system] notice is injected before the next user message telling the agent to re-read its steering files. Also: syncGeneralKnowledge only writes files when content actually changed, preventing unnecessary instructions hash diffs.
Telegram ClassicBot allows /start in private chats and regular groups (not the primary forum group) to spawn lightweight agent instances. Changes: - telegram.ts: allow non-primary-group messages through access control for classic mode routing (fleet-manager handles its own ACL) - classic-channel-manager.ts: add allowed_groups/allowed_users to ClassicBotYaml interface + isGroupAllowed/isUserAllowed methods - fleet-manager.ts: add Telegram classic mode routing in handleInboundMessage — /start, /stop, /chat commands + private chat auto-forward (no /chat prefix needed) - topic-commands.ts: register /start, /stop, /chat as default-scope bot commands via setMyCommands API Coexists with topic mode on the same bot token: - Primary forum group (group_id) → topic mode (unchanged) - Other groups/private chats → classic mode
Telegram classic mode messages (threadId=null, chatId != primaryGroupId) were blocked by fleet-level access control before reaching the classic mode routing block. Now these messages bypass fleet ACL and rely on classicBot.yaml's own isGroupAllowed/isUserAllowed checks.
Core multi-adapter architecture enabling simultaneous Telegram + Discord: types.ts: - ChannelConfig: add optional 'id' field (adapter identifier) - FleetConfig: add 'channels?: ChannelConfig[]' (backward compat with 'channel') config.ts: - loadFleetConfig: normalize 'channel' singular → 'channels' array - Auto-assign id = type when not specified fleet-manager.ts: - Add adapters Map<string, ChannelAdapter> for multi-adapter lookup - Add instanceAdapterBinding Map to track which adapter each instance uses - Add getAdapterForInstance() / bindInstanceAdapter() helpers - Split startSharedAdapter into loop + startSingleAdapter + startAdditionalAdapter - Primary adapter (first channel) stays as this.adapter for backward compat - Additional adapters get full event wiring (message, callback_query, slash_command) - handleOutboundFromInstance: route via getAdapterForInstance - handleToolStatusFromInstance: route via getAdapterForInstance - Bind instances to adapters on inbound message (topic, classic, general) - stopAll: stop all adapters Backward compatible: single 'channel' config works exactly as before. Multi-channel: add 'channels' array to run both platforms simultaneously.
feat: quickstart supports adding platform while fleet is running
… for platform detection
…assign fix: quickstart systemd restart + assignTopicIds channel_id platform detection
…ates When adding a second platform, existing general without channel_id was not recognized as belonging to any adapter. Auto-create spawned duplicates. Fix: 3-phase approach: 1. Match unbound generals by topic_id (TG general has topic_id=1, DC has general_channel_id) 2. Adopt remaining unbound generals first-come 3. Only create new instances if no unbound generals left This correctly handles: existing TG general + add DC → general gets channel_id=telegram, new general-discord created for DC.
fix: adopt existing general when adding second platform
Without it, General has no valid topic_id and routing breaks. Loop until user provides a value with explanation why it's needed.
fix: Discord general_channel_id required in quickstart
On machines without D-Bus user session (root, SSH without linger, containers), systemctl --user fails. Now falls back to fleet.pid kill like agend fleet stop does.
fix: agend stop fallback to PID kill when systemd unavailable
…lable Same D-Bus issue as agend stop. When startService() fails or no service is installed, spawns 'agend fleet start' in background instead of giving up.
fix: agend start fallback to direct fleet start when systemd unavailable
docs: complete configuration and commands reference
fix: add admin check to /restart + correct docs permissions
fix: add version to /sysinfo + improve kiro-cli context parser
fix: /status adds tmux capture fallback for context percentage
fix: remove /chat from TG ClassicBot menu
…reach classic instances
fix: TG ClassicBot chat-log + error notification delivery
Reviewer feedback: rawText starting with / would trigger /chat forwarding. Guard: any /command without @mention is logged as empty (won't forward).
fix: guard against /command forwarding without @mention
fix: TG ClassicBot bot reply logging + bot-to-bot @mention
…pter 1. TG adapter was never setting isBotMessage (from.is_bot) — now included 2. Debug logs at adapter access control + fleet-manager entry point 3. This is likely the root cause: without isBotMessage, the bot-to-bot fix in beta.17 never triggered
debug: TG bot message trace + fix missing isBotMessage
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.