Repository navigation
memos_search returns Chinese text as \uXXXX unicode escapes (Hermes adapter missing ensure_ascii=False) #2255
Description
Problem
When calling memos_search / memos_timeline / memos_skill_get etc. through the Hermes adapter, Chinese content in the returned JSON is rendered as \uXXXX unicode escapes instead of readable UTF-8 text.
Example — a memos_search hit snippet shows:
"reflection": "\u7528\u6237\u63d0\u4f9b\u4e86 Tushare KEY..."
instead of the actual Chinese characters. The data itself is correct in the DB; only the JSON returned to the host LLM is escaped.
Root cause
In the Hermes adapter adapters/hermes/memos_provider/__init__.py, the json.dumps(...) calls that serialize tool results to the LLM are missing ensure_ascii=False. Python's default is ensure_ascii=True, which escapes all non-ASCII characters to \uXXXX.
Affected lines (as of 2.0.15): memos_search (~1531), memos_get (~1580), memos_timeline (~1599), memos_skill_list (~1605), memos_environment (~1614 and ~1641), memos_skill_get (~1670).
Notably, a few other json.dumps calls in the same file DO already pass ensure_ascii=False (e.g. ~742, ~1015, ~1021), so this is just inconsistent — the tool-result serializers were missed.
Impact
For Chinese (and any non-ASCII) users, every memory retrieval result is hard to read and increases token/cognitive load on the host LLM. Functional correctness of retrieval is unaffected.
Suggested fix
Add ensure_ascii=False to the json.dumps(...) calls that serialize tool results back to the host (the ones listed above). One-line change each; verified it fixes the escape issue locally on 2.0.15.
Environment
- Plugin version: 2.0.15 (Hermes adapter)
- Host: Hermes Agent (Windows)
- Language: Chinese
Activity
Memtensor-AI commentedon Aug 16, 2026
🤖 AutoDev has picked up this issue and started working on it.
Task ID:
48b062ae885009c2Working branch:
bugfix/autodev-2255-20260816025022257Target branch:
latest dev* branchWorkflow: opsp (analysis → coding → testing → PR)
I will post the PR link here once done. If I need more information, I will ask in the comments.
fix(hermes-adapter): pass ensure_ascii=False in handle_tool_call json…
Memtensor-AI commentedon Aug 16, 2026
✅ AutoDev task
48b062ae885009c2completed.Summary: 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.
Base branch:
mainBranch:
bugfix/autodev-2255-20260816025022257Commit:
fa4bb36PR: #2256
Assigned to: @syzsunshine219
Reviewers: @whipser030, @hijzy
Fix #2255: memos_search returns Chinese text as \uXXXX unicode escape…
fix(hermes-adapter): pass ensure_ascii=False in handle_tool_call json…
Fix MemTensor#2255: memos_search returns Chinese text as \uXXXX unico…