fix(lark): recognize commands after leading bot mention

[Ars1027]

Fixes #9848

Problem

In Lark group chats, a message sent as @bot /stop keeps the leading bot mention in message_str. The mention wakes the bot, but wake-prefix and command matching still receive text such as @Bot /stop, so built-in and plugin commands are not recognized. An active Agent may then capture the message as a follow-up instead of stopping.

Solution

Normalize the textual projection in the Lark adapter. When the first parsed component is an At targeting the bot, omit that component from message_str while preserving it in the structured message chain.

The bot OpenID is the primary identity. The display name is used only when the mention has no identifier. Keeping this normalization in the platform adapter avoids changing shared wake or command behavior for other platforms.

Changes

• Extend _build_message_str_from_components() with optional bot identity parameters while preserving its default behavior.

• Remove only a leading bot-self mention from Lark's textual message projection.

• Preserve non-self mentions, non-leading mentions, structured At components, and existing callers.

• Add regression tests for built-in commands, plugin commands, natural-language messages, missing mention IDs, same-name users with different OpenIDs, empty input, and backward compatibility.

• No public API changes or new dependencies.

• This is NOT a breaking change.

Testing

Automated checks completed locally:

• ruff format .
• ruff check .
• python -m py_compile astrbot/core/platform/sources/lark/lark_adapter.py tests/test_lark_adapter.py
• git diff --check

The added pytest coverage exercises:

• @bot /stop -> /stop
• @bot /plugin_command -> /plugin_command
• @bot hello -> hello
• same-name mentions with a different OpenID remain unchanged
• name fallback applies only when the mention ID is missing
• calls without bot identity retain the previous behavior

Manual end-to-end verification was completed on AstrBot v4.27.4 in Ubuntu Docker:

• @bot /help invoked the built-in help handler.
• @bot /stop stopped an active Agent run.
• @bot <natural language> continued through the normal LLM path.
• QQ/OneBot /help remained functional.

Sanitized runtime evidence:

[Core][INFO] [lark] [At:<bot_open_id>] /stop
[Core][INFO] Agent execution was requested to stop by user.
[Core][INFO] Requested to stop 1 running tasks.

Notes for Reviewer

• The structured At component is intentionally preserved; only message_str is normalized.
• Only the first component can be omitted, and only when it identifies the bot.
• OpenID comparison intentionally takes precedence over display-name comparison to avoid stripping mentions of same-name users.
• Quoted-message text construction keeps the previous behavior because existing callers do not pass bot identity.

Checklist / 检查清单

• The change fixes an existing reported bug rather than introducing a new feature.
• Verification steps and sanitized runtime evidence are provided above.
• No new dependencies are introduced.
• The change does not introduce malicious code.

Summary by Sourcery

Normalize leading self-mentions in Lark messages so commands and wake-prefix handling receive the intended message text.

Bug Fixes:

• Fix Lark command and wake-prefix matching for messages that begin with a mention of the bot.
• Ensure active Agent runs can be stopped through commands sent after a leading bot mention.

Enhancements:

• Normalize Lark message text while preserving the structured mention component and existing behavior for non-self mentions and callers without bot identity.

Tests:

• Add regression coverage for bot mentions, command handling, identity matching, natural-language messages, and backward compatibility.

[sourcery-ai]

Hey - I've reviewed your changes and they look great!

Sourcery assessment

Approved.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}