Fix(yuanbao): quote file media resolve

[libo1106]

What does this PR do?

Fix Yuanbao bot's inability to read quoted file and image messages in group chats. When a user quotes a file/image message and @mentions the bot, the bot previously could not access the media content — it would report "file not found" because the quote metadata ( ⁠cloud_custom_data⁠ ) doesn't carry the ⁠resourceId⁠ needed for downloading.
This PR introduces a transcript fallback lookup mechanism: when the quote's ⁠desc⁠ field doesn't contain ybres references (which is always the case for file quotes, ⁠type=3⁠ ), the system looks up the quoted message by ⁠message_id⁠ in the session transcript to retrieve the ⁠resourceId⁠ that was stored when the message was first observed.

Related Issue

Fixes #

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

File: ⁠gateway/platforms/yuanbao.py⁠

1. ⁠_parse_resource_id⁠ + ⁠_extract_text⁠ ybres anchors — When ⁠TIMFileElem⁠ messages are observed, store ⁠[file:name|ybres:resourceId]⁠ in transcript so the resourceId is available for later lookup.
2. ⁠QuoteContextMiddleware._extract_quote_context()⁠ — Returns a 3-tuple ⁠(id, text, media_refs)⁠ extracting any ybres references from the quote desc. Added ⁠ctx.quote_media_refs⁠ field to ⁠InboundContext⁠ .
3. ⁠DispatchMiddleware._dispatch_inbound_event()⁠ — When a user quotes a message:
   ⦁If ⁠quote_media_refs⁠ is empty (file quotes only have filename in desc), performs a reversed transcript lookup by ⁠message_id⁠ to extract ybres anchors from the observed message content.
   ⦁Resolves and downloads quote media refs exclusively (skips unrelated history backfill).
   ⦁Overrides ⁠message_type⁠ to ⁠DOCUMENT⁠ when resolved media includes document MIME types ( ⁠application/*⁠ , ⁠text/*⁠ ) so ⁠gateway/run.py⁠ 's file injection logic properly provides the content to the agent.
4. ⁠_RESOLVABLE_MEDIA_KINDS⁠ constant — Extracted ⁠frozenset({"image", "file"})⁠ to eliminate duplicated kind-checking logic across 4 call sites.

How to Test

File quote (quote_type=3):

1. In a Yuanbao group chat, send a ⁠.txt⁠ file (do NOT @bot)
2. Wait for bot to observe the message (gateway log shows ⁠Group message observed⁠ )
3. Quote/reply to that file message, @bot and ask "这个文件里有什么？"
4. ✅ Expected: Bot reads and responds with file content

Image quote (quote_type=2):

1. In a Yuanbao group chat, send an image (do NOT @bot)
2. Quote/reply to that image message, @bot and ask "这是什么？"
3. ✅ Expected: Bot analyzes and describes the image

Regression — non-quote messages:

1. Send a message directly @bot without quoting → should work as before
2. Send a file directly @bot (no quote) → should download and read normally

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: Linux

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

• This skill is broadly useful to most users (if bundled) — see Contributing Guide
• SKILL.md follows the standard format (frontmatter, trigger conditions, steps, pitfalls)
• No external dependencies that aren't already available (prefer stdlib, curl, existing Hermes tools)
• I've tested the skill end-to-end: hermes --toolsets skills -q "Use the X skill to do Y"

Screenshots / Logs

[beforeload]

I've encountered this issue as well. when a user quotes a file or image message and @mentions the bot in a group chat, the bot can't access the media content because the quote metadata doesn't carry the resourceId.
I've tested this PR hands-on and it works well. Would love to see it merged. Thanks!

[libo1106]

Hi @alt-glitch

Just wanted to follow up on this PR — is there anything else I can help with to move it forward?

This fix is scoped only to the Yuanbao platform and shouldn’t affect other platforms. I’ve tested it locally (including file/image quote scenarios), and everything is working as expected.

Happy to make any adjustments or add more tests if needed. Thanks!

[loongfay]

This is our team's developer. This fix will optimize the dialogue experience of Yuanbao.
@alt-glitch @kovyrin @jbarket

[teknium1]

Salvaged onto current main via #26216 and merged. Your 5 commits were cherry-picked with your authorship preserved (rebase merge), now visible on main as d57a4b3..0086cda. Thanks for the fix!

One conflict at _collect_observed_media was resolved in favor of your _RESOLVABLE_MEDIA_KINDS constant — current main had migrated the literal {"image", "file"} to a set via PR #23937 (ruff PLR6201), so your constant landed as the canonical version. Same intent, cleaner code.