fix(utils): recover JSON objects after leading whitespace

[VectorPeak]

Description

This PR fixes a narrow JSON recovery edge case in safe_json_loads.

What Problem This Solves

safe_json_loads already recovered a valid JSON object followed by trailing garbage, such as:

{"k":"v"} trailing

However, the same recoverable object previously failed when the content also started with JSON whitespace:

\n  {"k":"v"} trailing

The failure happens only after the primary json.loads() path rejects the trailing bytes. At that point, the fallback called JSONDecoder.raw_decode(content) from index 0. Unlike json.loads(), this fallback expects parsing to begin at the JSON document start index, so leading whitespace caused the recovery path to treat an otherwise recoverable object as completely corrupted.

This can affect chat session or mission state recovery when a JSON state file is pretty-printed, hand-edited, or partially overwritten with trailing bytes. In that case, data that could otherwise be recovered is treated as unrecoverable and returns {}.

Why This Change Was Made

The fix belongs in qwenpaw.utils.json_utils.safe_json_loads because that helper is the shared recovery boundary used by chat session loading and mission state loading. Fixing it at the utility layer keeps the behavior consistent for all callers instead of patching one session path.

The implementation intentionally stays small: it advances past leading JSON whitespace with json.decoder.WHITESPACE.match(content, 0).end() before calling raw_decode. This keeps the fallback aligned with JSON whitespace rules without allocating a trimmed copy or accepting broader non-JSON leading text.

This PR does not broaden recovery beyond the existing contract. It still leaves these behaviors unchanged:

• valid JSON continues through the normal json.loads() path
• valid dicts with only leading/trailing whitespace still parse normally
• completely invalid JSON still returns {}
• recovered non-dict values still return {}
• BOMs, arbitrary leading non-whitespace garbage, multiple concatenated objects, and non-dict JSON values are not newly accepted

Evidence

The regression test covers the exact previously missed case:

\n  {"k": "v"}garbage

Before this change, this input returned {} because the fallback raw_decode started at index 0.

After this change, the fallback skips JSON whitespace first and recovers:

{"k": "v"}

Possible Call Chain / Impact

chat or mission state file
  -> qwenpaw.app.chats.session._safe_json_loads
  -> qwenpaw.utils.json_utils.safe_json_loads
  -> json.loads(...) fails on trailing garbage
  -> raw_decode fallback previously starts at index 0
  -> leading whitespace causes recovery to fail
  -> caller receives {}

Related Issue: N/A

Security Considerations: No auth, env/config boundary, or external input trust policy changes; this narrows a local JSON recovery edge case.

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation
• Refactoring

Component(s) Affected

• Core / Backend (app, agents, config, providers, utils, local_models)
• Console (frontend web UI)
• Channels (DingTalk, Lark, QQ, Discord, iMessage, etc.)
• Skills
• CLI
• Documentation (website)
• Tests
• CI/CD
• Scripts / Deploy

Checklist

• I ran pre-commit run --all-files locally and it passes
• If pre-commit auto-fixed files, I committed those changes and reran checks
• I ran tests locally (pytest or as relevant) and they pass
• Documentation updated (if needed)
• Ready for review

For Channel Changes (DingTalk, Lark, QQ, Console, etc.)

• I ran ./scripts/check-channels.sh (or ./scripts/check-channels.sh --changed) and it passes
• Contract test exists in tests/contract/channels/test_<channel>_contract.py (REQUIRED)
• Contract test implements create_instance() with proper channel initialization
• All 19 contract verification points pass (see tests/contract/channels/__init__.py)
• Optional: Unit tests in tests/unit/channels/test_<channel>.py for complex internal logic

Testing

I ran the focused chat session recovery tests and the adjacent raw-decode utility tests to confirm the broader raw JSON recovery behavior still passes.

Local Verification Evidence

D:\ZXY\Dev\Miniforge3\condabin\mamba.bat run -n prw-qwenpaw-py3.11 pytest tests/unit/app/chats/test_session.py tests/unit/agents/utils/test_tool_message_utils.py -q
# 74 passed

git diff --check
# passed

D:\ZXY\Dev\Miniforge3\condabin\mamba.bat run -n prw-qwenpaw-py3.11 pre-commit run --files src/qwenpaw/utils/json_utils.py tests/unit/app/chats/test_session.py
# passed

pre-commit run --all-files has not been run locally for this PR.

Channel-specific checklist items are not applicable because this PR does not change any channel implementation or contract surface.

Additional Notes

The code change is limited to the raw_decode fallback start index and one regression test. It does not change the normal JSON parsing path or the {} fallback contract for unrecoverable content.

[github-actions]

Welcome to QwenPaw! 🐾

Hi @VectorPeak, thank you for your first Pull Request! 🎉

🙌 Join Developer Community

Thanks so much for your contribution! We'd love to invite you to join the official QwenPaw developer group! You can find the Discord and DingTalk group links under the "Developer Community" section on our docs page:
https://qwenpaw.agentscope.io/docs/community

We truly appreciate your enthusiasm—and look forward to your future contributions! 😊

We'll review your PR soon.

[rayrayraykk]

LGTM

[rayrayraykk]

The test fails due to ACP breaking changes, which has been fixed in #5798

[github-actions]

Welcome to QwenPaw! 🎉

Thank you @VectorPeak for your first contribution! Your PR has been merged. 🚀

We'd love to give you a shout-out in our release notes! If you're comfortable sharing, please reply to this comment with your social media handles using the format below:

discord: your_discord_handle
x: your_x_handle
xiaohongshu: your_xiaohongshu_id

Note: Only share what you're comfortable with — all fields are optional.

Thanks again for helping make QwenPaw better!