fix(discord): prevent duplicate user message processing on auto-thread and reconnects (#51057)

[JoaoMarcos44]

What does this PR do?

This PR prevents duplicate processing and double agent runs when a single user message is received in Discord. It addresses two distinct root causes:

1. Auto-threading Context Race (Different Snowflake IDs): When a user sends a message in a text channel and auto_thread is enabled, the bot creates a thread for that message. This triggers a second MESSAGE_CREATE event for the "thread starter message" with message.id == thread.id and type=0 (default). Since the two dispatches represent different contexts (parent channel vs. thread channel), they carry different message (snowflake) IDs. This bypasses the standard message.id deduplication and the message type filter, resulting in two parallel agent runs.
2. Adapter Recreation/Reconnect (Same Snowflake ID): When the connection drops and a new DiscordAdapter instance is initialized, its local self._dedup cache starts empty, failing to block duplicate messages replayed during the RESUME flow.

To resolve these, we implement:

• Content-Based Deduplication (Adapter-Level): A sliding window of 5 seconds to drop duplicate messages with the same author and content text (u:{author_id}:{content}) arriving in different contexts.
• Cross-Adapter Deduplication (Gateway-Level): A platform-level message ID deduplicator on the GatewayRunner which persists across adapter re-connections and instances.

Related Issue

Fixes #51057

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✅ Tests (adding or improving test coverage)

Changes Made

• plugins/platforms/discord/adapter.py: Added content-based deduplication using self._content_dedup = MessageDeduplicator(max_size=500, ttl_seconds=5). Dropped duplicate deliveries (author+content) within 5 seconds.
• gateway/run.py: Added cross-adapter deduplication using self._platform_dedup on the GatewayRunner class to prevent duplicate processing of the same message ID across adapter restarts.
• tests/gateway/test_discord_cross_adapter_dedup.py: Added comprehensive unit tests for cross-adapter and content deduplication behavior.

How to Test

1. Run the newly added test suite to verify correct deduplication logic under various scenarios:

   pytest tests/gateway/test_discord_cross_adapter_dedup.py

2. Run existing Discord adapter tests to ensure no regressions:

   pytest tests/gateway/test_discord_free_response.py

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: Windows 11

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

Screenshots / Logs

Logs of successful test run:

tests\gateway\test_discord_cross_adapter_dedup.py ..........             [100%]
============================= 10 passed in 0.86s ==============================

tests\gateway\test_discord_free_response.py ............................ [ 65%]
...............                                                          [100%]
============================= 43 passed in 1.26s ==============================

[alt-glitch]

This was generated by AI during triage.

Duplicate of #51129 — both fix the same #51057 regression where Discord auto_thread emits a second on_message under the thread context (different snowflake id) that bypasses message.id dedup. #51129 (open, earlier) takes the narrower thread-starter-dedup approach; this PR adds broader content-based + cross-adapter deduplication. Flagging for maintainer to pick the canonical approach.

[tonydwb]

Adds cross-adapter message deduplication to prevent duplicate processing when Discord RESUME replays events after reconnects. Also adds content-based dedup with 5-second TTL to catch same user message delivered in multiple Discord contexts (channel vs thread). Good fix for issue #51057. Well-scoped to 2 files.

---

Reviewed by Hermes Agent

[kshitijk4poor]

Thanks for tackling #51057 as well — closing this in favor of #51129, which fixed the same bug and was submitted first.

Both PRs correctly identified the auto-thread starter-message root cause. We went with #51129's approach because it's deterministic: it pre-seeds str(thread.id) into the existing _dedup cache, which neutralizes the exact ID the duplicate thread-starter event carries (per Discord's API, thread.id == starter message.id) with no false-positive risk.

The content-based dedup here (u:{author}:{content} with a 5s TTL) would also drop a legitimate repeated message — e.g. a user sending "yes" twice within 5s gets no response — which the deterministic approach avoids. The cross-adapter _platform_dedup on GatewayRunner addresses a different concern (RESUME-replay across reconnects) than #51057 and touches core; if that's a real issue worth fixing, it'd be welcome as its own scoped PR.

Merged fix: #51552 (#51552). Appreciate the thorough write-up and tests!