Task-plan tracking and progress-message management

[dimonb]

📝 Description

This PR introduces task-plan tracking and progress-message management from scratch.

It adds a new tasktool with persisted per-session task state, wires the feature into the agent loop, and adds the channel plumbing needed to post and later update a live execution plan. It also rounds out the feature with config support, Telegram chunk-aware message editing, regression tests, and English documentation.

High-level changes:

• add session.TaskManager for persisted session-scoped task plans and task updates
• add tasktool actions for create_plan, update_task, list_plan, and resend_plan
• add synchronous channel send/edit support so tools can immediately capture message IDs and update progress later
• handle multi-part Telegram plan messages by tracking and editing all chunk IDs
• introduce SequentialTool so stateful tools like tasktool can preserve model order within one turn
• document tasktool runtime behavior and add a short developer note for tool authors

🗣️ Type of Change

• 🐞 Bug fix (non-breaking change which fixes an issue)
• ✨ New feature (non-breaking change which adds functionality)
• 📖 Documentation update
• ⚡ Code refactoring (no functional changes, no api changes)

🤖 AI Code Generation

• 🤖 Fully AI-generated (100% AI, 0% Human)
• 🛠️ Mostly AI-generated (AI draft, Human verified/modified)
• 👨‍💻 Mostly Human-written (Human lead, AI assisted or none)

🔗 Related Issue

N/A

📚 Technical Context (Skip for Docs)

• Reference URL: N/A
• Reasoning: The planning feature needs persistent task state tied to agent sessions, plus synchronous access to platform message IDs so plan messages can be updated later. Because task updates mutate shared session state, the tool execution layer also needs an ordered path for stateful tools within a single LLM turn.

🧪 Test Environment

• Hardware: Apple Silicon Mac
• OS: macOS 26.3
• Model/Provider: Mock providers in Go unit tests
• Channels: Direct/CLI, Telegram (unit-test coverage)

📸 Evidence (Optional)

Click to view Logs/Screenshots

go test ./pkg/agent ./pkg/tools ./pkg/channels/telegram ./pkg/session

Passed locally.

☑️ Checklist

• My code/docs follow the style of this project.
• I have performed a self-review of my own changes.
• I have updated the documentation accordingly.

[CLAassistant]

All committers have signed the CLA.

[yinwm]

Hi @dimonb, thanks for this valuable PR! Task progress tracking is definitely a feature users need.

However, I noticed there's a major Agent Loop refactor being planned: #1316

#1316 will introduce:

• EventBus - for emitting tool execution events
• Hook system - for intercepting and observing execution
• Smart tool grouping - sequential execution for stateful tools

These mechanisms could implement tasktool's functionality more elegantly:

• Task state changes emitted via EventBus, instead of the tool managing message ID tracking itself
• SequentialTool could reuse the "smart grouping" mechanism from [Agent refactor] Event-driven agent loop with hooks, interrupts, and steering #1316

Suggestion: Would you consider waiting for #1316 to be completed before implementing tasktool? This would:

1. Avoid duplicate work of merging first, then refactoring later
2. Make tasktool's implementation cleaner by leveraging EventBus + Hook directly
3. Maintain architectural consistency

Alternatively, if this feature is urgently needed, it could be merged as a temporary solution and refactored together with #1316 later.

Would love to hear your thoughts on this.

[dimonb]

@yinwm sounds reasonable

[sipeed-bot]

@dimonb Hi! This PR has had no activity for over 2 weeks, so I'm closing it for now to keep things tidy. If it's still relevant, feel free to reopen it anytime and we'll pick it back up.