Fix #2255: memos_search returns Chinese text as \uXXXX unicode escapes (Hermes adapter miss - #2256
Open
Memtensor-AI wants to merge 2 commits into
Open
Conversation
….dumps (MemTensor#2255) memos_search / memos_get / memos_timeline / memos_skill_list / memos_environment / memos_skill_get in the Hermes memory provider serialized their tool results back to the host LLM with the default ensure_ascii=True, which escaped every non-ASCII code point (notably Chinese memory content) to \uXXXX. The DB stored the correct UTF-8; only the wire JSON was mangled. This inflated tokens for Chinese users and made retrieval results unreadable when debugging. Add ensure_ascii=False to every json.dumps inside handle_tool_call (both the tool-result branches called out in the issue and the error/fallthrough branches, for a uniform pattern that matches the other json.dumps calls in this file at L742/1015/1021 that already pass the flag). Add HandleToolCallEnsureAsciiTests to apps/memos-local-plugin/tests/python/test_hermes_provider_pipeline.py with 8 regression cases (one per affected tool + both branches of memos_environment) that assert the returned JSON contains raw Chinese characters, contains no "\u" escape marker, and round-trips through json.loads. Verified the tests fail without the fix and pass with it; ruff check + ruff format clean; full test file 35/35 green. Fixes MemTensor#2255
Memtensor-AI
added
ai:generated
Collaborator
Author
🤖 Open Code ReviewTarget: PR #2256 ✅ OpenCodeReview: No comments generated. Looks good to me. Generated by cloud-assistant via Open Code Review. |
Collaborator
Author
🔧 Open Code Review requested Agent fixOpen Code Review found 2 issue(s). I have resumed the development Agent to fix them.
The Agent will push a new commit to this PR branch. OCR will recheck after the commit is pushed. |
Address PR MemTensor#2256 OCR review findings: 1. Add `test_memos_get_world_model_returns_utf8_chinese` to cover the `world_model` branch of `memos_get` (routed via `memory.get_world`). Without this test, an accidental removal of `ensure_ascii=False` from the world_model branch would go undetected — the previous tests only exercised the `trace` and `policy` kinds. 2. Remove the fragile `assertNotIn("\\u", raw)` guard from all HandleToolCallEnsureAsciiTests cases. That check was prone to false positives (any legitimate value containing a backslash followed by `u`, e.g. a Windows path, would fail the test) while providing weaker coverage than the `assertIn(_CH_..., raw)` + `json.loads` field equality assertions already present. Add a header comment explaining why the literal-in-raw check is the robust regression guard. Also add a parsed field-equality assertion to `test_memos_skill_list_returns_utf8_chinese` for parity with the other cases. All 36 tests in tests/python/test_hermes_provider_pipeline.py pass.
Collaborator
Author
✅ Automated Test Results: PASSEDAll tests passed (35/35 executed). memos_local_plugin/changed-repo-python: 35/35. Duration: 4s [advisory, non-gating] AI-generated tests on branch test/auto-gen-48b062ae885009c2-20260816111233: 194/196 passed, 2 failed — these do NOT affect the PR verdict; review the branch manually. Branch: |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Fixed #2255: Chinese memory content returned by the Hermes adapter's memos_* tools was being serialized to the host LLM as
\uXXXXescapes because everyjson.dumps(...)insideMemTensorProvider.handle_tool_call(apps/memos-local-plugin/adapters/hermes/memos_provider/init.py) relied on Python's defaultensure_ascii=True. Data in the DB was always correct; only the on-wire JSON was mangled, inflating tokens and making retrieval unreadable for Chinese users.Added
ensure_ascii=Falseto everyjson.dumpscall insidehandle_tool_call— both the tool-result branches called out in the issue (memos_search, memos_get trace/policy/world_model, memos_timeline, memos_skill_list, memos_environment list+query, memos_skill_get) and the error / fallthrough branches. The uniform pattern matches the threejson.dumpscalls elsewhere in the same file (L742/1015/1021) that already set the flag.Added
HandleToolCallEnsureAsciiTeststoapps/memos-local-plugin/tests/python/test_hermes_provider_pipeline.py— 8 regression cases that assert each affected tool returns raw Chinese characters, contains no\uescape marker, and round-trips throughjson.loads. Confirmed the tests fail without the fix and pass with it. Full test filetest_hermes_provider_pipeline.pyis 35/35 green (27 pre-existing + 8 new);ruff checkandruff format --checkboth clean on the touched files.Categorized as an opsp Bug quick-fix (no proposal/spec/design), one-line change per return statement. Committed on bugfix/autodev-2255-20260816025022257 and pushed to origin. Reviewers: @whipser030, @hijzy.
Related Issue (Required): Fixes #2255
Type of change
Please delete options that are not relevant.
How Has This Been Tested?
Automated tests are pending.
Checklist
@whipser030, @hijzy please review this PR.
Reviewer Checklist