fix: curator JSON parsing and CLI response extraction #9
Conversation
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
- Fix JSON syntax in curation prompt (single quotes → double quotes)
- Fix CLI response extraction for Claude Code nested message format
- Add ast.literal_eval fallback for non-compliant LLM responses
- Implement real stats gathering in /memory/stats endpoint
- Add get_stats() method to storage layer
The curator was failing because:
1. Prompt examples used Python dict syntax, causing Claude to return
invalid JSON with single quotes
2. Claude Code CLI outputs {"type":"assistant","message":{"content":[...]}}
but extraction looked for message["content"] instead of
message["message"]["content"]
3. No fallback when LLM ignored format instructions
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Author
|
I had an issue while testing on my Mac and hope this helps |
jespo2021
added a commit
to jespo2021/memory
that referenced
this pull request
Dec 13, 2025
Tests added: - test_cli_response_extraction: Validates nested Claude Code message format handling (message.message.content) - test_json_parsing_fallback: Validates ast.literal_eval fallback for single-quote JSON from LLM - test_storage_stats: Validates get_stats() returns real database counts Code changes: - Formatted all source files with black - Fixed ruff linting issues (unused imports, import ordering, trailing whitespace) - Removed f-string prefix from strings without placeholders - Removed unused variable assignments All tests verified to fail on main branch and pass on this branch.
Tests added: - test_cli_response_extraction: Validates nested Claude Code message format handling (message.message.content) - test_json_parsing_fallback: Validates ast.literal_eval fallback for single-quote JSON from LLM - test_storage_stats: Validates get_stats() returns real database counts All tests verified to fail on main branch and pass on this branch.
jespo2021
force-pushed
the
fix/curator-json-parsing
branch
from
2fd3d92 to
7894a1d
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Summary
Fixes memory curation failing silently due to JSON parsing errors and incorrect CLI response extraction.
Changes
message["message"]["content"]instead ofmessage["content"]ast.literal_evalwhen LLM returns Python dict syntax despite instructions/memory/statsnow returns actual counts instead of placeholder zerosRoot Cause
The curator was returning 0 memories because:
'key': 'value'examples, so Claude followed that format and returned invalid JSON{"type":"assistant","message":{"content":[...]}}but the code looked formessage["content"]directlyTesting
Regression tests added to verify the fixes:
test_cli_response_extraction.py: Validates nested Claude Code message format handlingtest_json_parsing_fallback.py: Validates ast.literal_eval fallback for single-quote JSONtest_storage_stats.py: Validates get_stats() returns real database countsAll tests verified to fail on main branch and pass on this branch.
🤖 Generated with Claude Code