feat(gateway): add Telegram temporary tool progress

[mrcharlesiv]

What does this PR do?

Adds an opt-in Telegram display setting, display.platforms.telegram.temporary_tool_progress, so Hermes can show temporary work/status visibility while a turn is running and then delete the bot-sent temporary bubbles after the final response is successfully delivered.

This keeps the useful live “working” visibility without leaving Telegram chats cluttered. The implementation is Telegram-gated, default-off, generation-safe, and best-effort: it tracks only current-run bot-sent temporary message IDs, never deletes the final response or user messages, preserves breadcrumbs on failure, and ignores Telegram deletion failures.

Covered temporary messages now include:

• tool-progress bubbles
• long-running ⏳ Still working... notices
• status-callback messages such as context-pressure/compaction notices

This PR is intentionally separate from #18266’s display.ephemeral_system_ttl lifecycle. #18266 deletes slash-command/system notices after a TTL; this PR deletes tracked Telegram progress/status messages only after confirmed final delivery.

Related Issue

N/A — no issue filed.

Potentially related PRs found during duplicate check:

• feat(gateway): auto-delete tool progress messages after response #4767 feat(gateway): auto-delete tool progress messages after response
• feat(gateway): silent tool progress messages on Telegram #6646 feat(gateway): silent tool progress messages on Telegram
• fix(gateway): update Telegram progress lines on tool completion #13599 fix(gateway): update Telegram progress lines on tool completion
• [codex] Add message transport for gateway tool progress #18050 [codex] Add message transport for gateway tool progress
• feat(gateway): auto-delete slash-command system notices after TTL #18266 feat(gateway): auto-delete slash-command system notices after TTL

This PR is narrower than the broad auto-delete variants: Telegram-only, default-off, generation-safe, and success-only cleanup for bot-sent temporary tool-progress/status bubbles.

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• gateway/display_config.py: adds temporary_tool_progress display setting and supports the older cleanup_tool_progress alias.
• gateway/platforms/base.py: makes post-delivery callbacks composable and generation-aware so cleanup hooks do not clobber background-review, EphemeralReply, or newer-run callbacks.
• gateway/run.py: tracks Telegram tool-progress message IDs, long-running Still working... message IDs, and status-callback message IDs, then schedules best-effort cleanup only after successful final/streamed/queued delivery.
• hermes_cli/config.py: adds the default config key while preserving ephemeral_system_ttl.
• cli-config.yaml.example: documents the new display-platform override.
• tests/gateway/test_display_config.py and tests/gateway/test_run_progress_topics.py: add coverage for config resolution, callback composition/generation safety, success cleanup, failure breadcrumbs, background-review composition, queued follow-up behavior, Still working... cleanup, and status-callback cleanup.

How to Test

1. Enable the setting:

   display:
     platforms:
       telegram:
         tool_progress: all
         temporary_tool_progress: true

2. Send a Telegram message that causes a tool call or runs long enough to emit a ⏳ Still working... notice.
3. Confirm temporary progress/status bubbles are visible while the turn runs and disappear after the final response lands.
4. Confirm that failed runs leave breadcrumbs in place.

Automated verification run locally:

scripts/run_tests.sh \
  tests/gateway/test_status_command.py::test_post_delivery_callback_generation_snapshot_happens_after_bind \
  tests/gateway/test_run_progress_topics.py \
  tests/gateway/test_display_config.py \
  tests/gateway/test_ephemeral_reply.py

Result: 78 passed in 12.14s.

Broad gateway verification run locally:

ulimit -n 4096
scripts/run_tests.sh tests/gateway/ -k 'not test_send_typing and not test_closed_when_http_not_ready and not test_discord_free_channel_skips_auto_thread'

Result: 4667 passed, 10 skipped, 196 warnings in 69.23s.

Additional checks:

• git diff --check
• Python AST parse for modified Python files
• YAML parse for cli-config.yaml.example
• static scan: no obvious hardcoded secrets / unsafe eval-exec / pickle / SQL formatting patterns
• independent reviewer pass: no blockers after callback composition fix

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS / Telegram gateway

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

For New Skills

N/A — this PR does not add a skill.

Screenshots / Logs

Focused regression output:

78 passed in 12.14s

Broad gateway output, excluding known unrelated local failures:

4667 passed, 10 skipped, 196 warnings in 69.23s

Notes on unchecked full-suite item:

• Full pytest tests/ -q was not run after the May 6 rebase.
• tests/gateway/test_discord_free_response.py::test_discord_free_channel_skips_auto_thread fails isolated on clean origin/main (da6019820) too, so it is not attributed to this PR.
• The broad gateway run excludes known unrelated/local failures: test_send_typing, test_closed_when_http_not_ready, and the Discord free-response test above.
• Additional checks passed after the rebase: git diff --check, Python AST parse for modified Python files, YAML parse for cli-config.yaml.example, and a static secret/unsafe-pattern scan.

[teknium1]

Thanks for the contribution and for building this — the UX win (keep progress visible live, clean up the chat afterward) is a real improvement, and your write-up of the breadcrumb-on-failure behavior was exactly the right default.

I went ahead and merged a salvage rewrite in #21186 (merge commit bf843ad) that lands the same user-visible feature with a lighter footprint:

• Same config surface: display.cleanup_progress: true (global or per-platform under display.platforms.<plat>.cleanup_progress), default off
• Same cleanup scope: tool-progress bubble, ⏳ Still working... notices, status-callback messages
• Same correctness invariants: skip-on-failure (breadcrumbs preserved), success-only deletion, Telegram-gated via delete_message support check

The main difference is on the base.py side — instead of turning the post-delivery-callback slot into a multi-shape list with a dedicated composition helper and 4-shape normaliser, the salvage chains a new callback onto any existing one (with per-callback exception isolation) in register_post_delivery_callback itself. That keeps the existing single-callback + (gen, callback) tuple shapes and lets the cleanup callback coexist with the background-review release hook that's already registered on the same slot.

Net: ~141 LOC production vs your ~235, same behaviour, and your name is in AUTHOR_MAP + the commit's Co-authored-by trailer so future release notes pick this up correctly. Closing this in favour of the merged salvage — really appreciate the idea and the thorough test coverage, both of which informed the rewrite.