[unsupportedpastels]

What does this PR do?

Voice-channel turns rebuild their SessionSource from the copy bound at /voice join. #133463 made that rebuilt source carry the speaker's display name and the channel's current labels, so typing and talking now render the same pinned prompt. Two problems from #127944 were not part of that fix, and this PR carries teknium's commits for both from that PR onto current main:

• Voice turns inherit the /voice join message id. A typed /voice join binds that message's source, message_id included, and every later voice turn kept it. run.py exports source.message_id as HERMES_SESSION_MESSAGE_ID, the turn's triggering message. On Discord, the background-process tool stores it as watcher_message_id, the reply anchor for completion notices, so a process started from a voice turn recorded the old /voice join message as its trigger. The cron and Kanban tools read it only for Slack and Telegram, so they aren't affected here. Voice turns now clear it. This no longer affects the prompt cache, because fix(discord): slash, /thread and voice turns carry a message turn's prompt inputs #133463 made the triggering-message note static.
• A speaker outside the member cache renders as a bare id. guild.get_member() is cache-only. A speaker it misses who isn't the /voice join invoker rendered as **User:** "<id>", so their turns still re-rendered the pinned prompt. The rebuilt source now falls back to the cached user's display name (global name or username) before the id.

Speaker	main	This PR
Member cached	display name	display name (unchanged)
/voice join invoker, member not cached	name bound at join	name bound at join (unchanged)
Other speaker, member not cached, user cached	bare id	global name / username
Cached nowhere	bare id	bare id (unchanged)
Any voice turn after a typed /voice join	join message id as message_id	no message_id

Adapted onto main. #127944 looked the name up through a new DiscordAdapter.voice_member_display_name. On main, #133463 resolves it inline in _voice_input_source, so both commits are rebuilt there, with conflict notes in the commit messages:

• The cached-user fallback is a small _cached_user_display_name helper.
• It runs after the joiner's bound name, not before as in fix(gateway): voice-channel turns render the same session prompt as text turns #127944. The bound name is the server nickname the joiner's typed messages carry; their cached global name would differ from it.
• The tests moved into tests/gateway/test_discord_turn_prompt_parity.py. test_voice_session_prompt.py isn't on main.

Known gap (out of scope). A speaker who has a server nickname, but is in neither the member cache nor the join binding, now gets their global name. That beats an id, but still differs from the nickname their typed messages carry. Matching it exactly would need a member fetch over the API on the voice path.

Related Issue

No issue. These are the remaining parts of #127944, whose name and triggering-note fixes landed in #133463.

Credit: both commits are @teknium1's review fixes on #127944, cherry-picked with -x and authorship kept. The underlying work and diagnosis are @matttmcfarland's #127944, and @andrexibiza independently reproduced the uncached-speaker fallback (F3 on that PR).

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix (undisclosed vulnerabilities in SECURITY.md §3.1 scope go to Security Advisories first, not a public PR)
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• gateway/run_voice.py: _voice_input_source clears the bound message_id, and both its branches fall back to _cached_user_display_name (new) before the bare id.
• tests/gateway/test_discord_turn_prompt_parity.py: the voice parity test binds a typed /voice join id and asserts that the voice turn doesn't carry it. The uncached case covers another speaker cached only as a user, one cached nowhere, and no binding at all. The joiner keeps their bound nickname over their cached global name.

How to Test

1. pytest tests/gateway/test_discord_turn_prompt_parity.py -q: 12 passed.
2. With main's run_voice.py, 4 of them fail: the message-id assertion in every voice case. With only the first commit's run_voice.py, 1 fails: the uncached-speaker case.
3. The Discord, voice, session, cron, prompt-freeze and pin-wiring test files (113 files) give 822 passed, 0 failed, 2 skipped. ruff passes, and ty reports the same 34 diagnostics on the two files as on main.

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 (only fix(gateway): voice-channel turns render the same session prompt as text turns #127944 has these commits)
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass: I ran the 113 related files above, not the full suite
• python scripts/check passes (the blocking lint and code-health checks CI runs)
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: unit tests only, on Linux; no live Discord voice session

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

Infographic