Changelog — Features

New Features (TAB-562, Phase 7)

• AskUserQuestion routed through Telegram during plan mode — Completes the /plan CLI native-plan-mode integration. The CLI's AskUserQuestion tool calls now surface as Telegram inline keyboards while a plan is being generated. The CLI waits up to 120 seconds for the user's selection, then resumes plan generation with the chosen option. If no response arrives within the window, the plan continues with the CLI's default fallback.

New Features (TAB-562, Phases 1–6)

• Native plan mode integration — The /plan command now delegates to Claude Code's built-in plan mode rather than a custom planner. This means plans benefit from the CLI's full context window, tool access, and iterative refinement. The migration covered six phases: prompt rewriting, plan card adaptation, approval flow, execution handoff, error recovery, and cleanup of the legacy planner code path.

New Features (TAB-557)

• /plan command for explicit plan-first tasks — Send /plan <description> to get a structured plan card before any code runs. The bot generates an AI-powered plan with phases, then waits for your approval before delegating execution. Useful when you want to review the approach before committing to a task.

Bug Fixes (TAB-557, TAB-558, TAB-559)

• /plan AI pre-check and error handling — Added validation and graceful error recovery so malformed or ambiguous plan requests surface a clear error card instead of silently failing.
• Classifier mock pollution fixed — Resolved test isolation issues where classifier mocks bled across test suites. Closed the vi.mocked gap that allowed stale mocks to persist between runs.

New Features (TAB-549)

• Follow-up button on the Task Complete card — After a delegate task finished, there was previously no way to ask a clarifying question or request a quick follow-up change without starting a full new task cycle with a plan card and approval step. Tap Follow-up to open a prompt; the response runs via --resume <sessionId> on the exact same Claude session — no new plan generated. Claude remembers the files it changed and the reasoning it used, so corrections land in context. The answer card shows View Changes, Edit, and Dismiss actions. Edit chains another follow-up on the same session. A pendingFollowups map stores session state with a 5-minute TTL.
• Task Complete summary raised from 300 to 800 characters — The 300-character summary was too short to be useful for complex tasks. Raised to 800 characters, still well within the 4096-character Telegram message cap.
• Approval mode saved and restored safely — Approval mode state is now saved and restored inside a finally block during follow-up execution to prevent state leaks. The daily message limit is checked before follow-up execution. Ownership is verified on every callback to prevent cross-user state leaks.

New Features (TAB-537)

• Voice transcription no longer requires a user OpenAI key — Premium users can now send voice messages without supplying their own OPENAI_API_KEY. Audio flows Telegram OGG → CodePulse service → Cloudflare Worker proxy → OpenAI Whisper API → transcription, with the OpenAI key stored server-side in Worker secrets. License key is validated before forwarding. Audio files are ephemeral and never stored by the proxy. Voice usage counts against the monthly token budget. Users with their own OPENAI_API_KEY can still use it as a direct local fallback.

Features (TAB-529)

• Extractive plan generation preserving full user intent — Replaced the lossy creative plan summarization with a lossless extraction approach. Haiku now extracts constraints, cross-references, and negative constraints directly from user messages. The original user message travels alongside the structured plan to Claude CLI, ensuring no detail is lost.
• Richer plan cards — The plan card now displays Constraints, Cross-check, and Do NOT sections when extracted from the user's message, giving full visibility into how the plan was interpreted before execution begins.
• Increased maxTokens for plan generation — Bumped from 512 to 1024 tokens to prevent truncated JSON responses from the more detailed extractive prompt.

Features (TAB-518 Epic: Delegate Mode Task Flow Redesign)

• New: Task Complete card (TAB-519) — Replaces the old result card with a richer summary. Shows mode badge, 300-character CLI summary, duration in Xm Ys format, files changed count, and line delta. Includes View Changes and Dismiss buttons. The card is persistent with no timeout and stores execution metadata for the next stage.
• New: Status-aware actions card (TAB-520) — View Changes detects git state (uncommitted, committed, pushed, or clean) combined with bridge session state (alive or dead) to show one of six card variants. Supports Ask Claude to Commit or Push via --resume, direct Commit or Push when session is dead, View Full Diff as a message or .diff file, and Discard All with a two-step confirmation flow.
• Removed: Old Push and PR flow (TAB-521) — Removed resultKeyboard, sendResultCard, handleResultCallback, and all associated dead code. Cleaned up storedPlans on plan cancel, commit proposal approve, and push confirmation. Removed nine stale test cases and the orphaned 'result' entry from the CardType union.

New Features

• Premium features auto-enable on license validation (TAB-512) — Pulse, Guardian, ErrorWatch, Bridge, CommitReview, MCP Heartbeat, and TCVF now flip on automatically when the service confirms premium status. Users no longer need to edit .env by hand to activate paid features.
• /send_logs Telegram command (TAB-514) — Collects diagnostic logs from the running service and submits them to GitHub from Telegram, mirroring the launcher button (TAB-470) for remote debugging.
• /manage surfaces specific Worker errors (TAB-513) — "No Stripe customer", "key not found", and similar conditions are reported with their actual message instead of a generic failure.

New Features (TAB-470)

• Send Diagnostic Logs button in the Help tab — One click collects sanitized logs (from the running service endpoint, or log files as fallback) and POSTs them to a Cloudflare Worker which creates a structured GitHub issue with system info, health report, stop stats, hook events, and log sections in expandable <details> blocks. Rate-limited to 3 submissions per hour per license key.

New Feature

• [Wait Quietly] Button on Status Cards: When Claude sends a status update ("Claude:" card with "Reply within 240s"), you previously had only two choices: [Reply] (which caused an infinite loop of cards) or [Stop] (which killed the session). The new [Wait Quietly] button redirects Claude to the wait_for_instructions MCP tool — the session stays alive, no more cards appear, and you can send a message whenever you're ready to continue.

How It Works

When you tap [Wait Quietly]:

1. Claude enters a quiet waiting state (MCP wait_for_instructions)
2. The card updates to "Waiting quietly. Send a message when ready."
3. No more status cards appear
4. The session stays alive — send a Telegram message anytime to resume

This only affects status cards (pass classification). Question cards ("Claude Needs Your Input") are unchanged — they don't have a loop problem.