Fix Unicode normalization in memory viewer search

[kaushalrog]

This PR improves memory viewer search behavior for Chinese and Japanese text by normalizing Unicode strings before matching.

Changes:

• Added Unicode normalization using NFKC
• Improved multilingual search consistency
• Better handling for full-width characters

Fixes #514

Summary by CodeRabbit

Release Notes

• Bug Fixes
  • Improved Memories tab search functionality with enhanced Unicode support and case-insensitive matching across memory titles and contents for more accurate search results.

[vercel]

@kaushalrog is attempting to deploy a commit to the rohitg00's projects Team on Vercel.

A member of the Team first needs to authorize it.

[coderabbitai]

📝 Walkthrough

Walkthrough

The PR improves memory search filtering by applying Unicode normalization (NFKC) and case-insensitive matching to both search queries and memory content. The renderMemories() function now normalizes the search term and memory title/content fields before comparison, replacing the previous direct lowercased string matching logic.

Changes

Unicode-Normalized Memory Search

Layer / File(s)	Summary
Unicode normalization in memory search filtering src/viewer/index.html	renderMemories() applies NFKC normalization and lowercasing to the search term and memory title and content fields before includes() checks, improving match accuracy for Unicode variations.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Poem

🐰 Memories now dance through Unicode's way,
NFKC normalizes the search display,
No diacritic shall hide what we seek,
Case-insensitive matches bloom so sleek!
Filtering refined for memories to keep. ✨

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Linked Issues check	⚠️ Warning	The PR addresses a Unicode normalization issue for multilingual search, but issue #514 requests documentation on scalable deployment, which are unrelated objectives.	Verify the correct issue number. If #514 is indeed the target, this PR does not fulfill the request for scalable deployment guidance. If this PR targets a different issue, update the linked issue.
Out of Scope Changes check	⚠️ Warning	The PR implements Unicode normalization (NFKC) and case-insensitive search filtering, which is within scope but does not address the linked issue #514 about scalable deployment.	Review the linked issue. The PR changes are out of scope for issue #514; either fix the issue link or ensure the PR addresses the correct scaling/deployment requirement.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The PR title accurately describes the main change: updating Unicode normalization in the memory viewer search functionality.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

Warning

Review ran into problems

🔥 Problems

Git: Failed to clone repository. Please run the @coderabbitai full review command to re-trigger a full review. If the issue persists, set path_filters to include or exclude specific files.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 2

🧹 Nitpick comments (1)

src/viewer/index.html (1)

2228-2235: ⚡ Quick win

Redundant lowercasing of search term.

The search term is already lowercased on line 2228, then normalized and lowercased again on line 2235 (inside the filter loop, repeated for every item). This is inefficient and redundant.

Consider normalizing and lowercasing the search term once before the filter:

var search = state.memories.search;
var normalizedSearch = search ? search.normalize("NFKC").toLowerCase() : '';

Then remove the normalization/lowercasing inside the filter loop.

Also note: The file consistently uses var throughout, but the new code uses const (lines 2233-2243). For consistency, consider using var to match the existing style.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/viewer/index.html` around lines 2228 - 2235, Move the normalization and
lowercasing of the search term out of the items.filter loop: compute a single
normalizedSearch once from state.memories.search (e.g., var normalizedSearch =
state.memories.search ? state.memories.search.normalize("NFKC").toLowerCase() :
'') before creating filtered, then inside the filter callback (the function used
to build filtered from items) use that normalizedSearch instead of recomputing;
also keep the file's var-style declarations (use var for normalizedSearch) to
stay consistent with existing code and remove the redundant toLowerCase() on the
already-lowercased state.memories.search.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/viewer/index.html`:
- Around line 2231-2254: The PR claims to fix issue `#514` but the diff in
items.filter (variables normalizedSearch, normalizedTitle, normalizedContent
used to implement Unicode-normalized multilingual search) is unrelated to
running agentmemory persistently; confirm the intended target and either change
the PR title/description and issue reference to the correct multilingual-search
issue (or create one) or update the code to actually address issue `#514`;
specifically, review the search-related changes inside the items.filter callback
(normalizedSearch/normalizedTitle/normalizedContent) and update the PR metadata
so the referenced issue matches these changes, or revert these search changes
and implement the persistent/scalable agentmemory fix required by `#514`.
- Around line 2245-2252: The search filter's condition incorrectly checks
normalizedContent.includes(normalizedContent) which always succeeds and prevents
content matches; update the condition inside the if in the search block so it
tests normalizedContent.includes(normalizedSearch) (i.e., compare
normalizedContent to normalizedSearch), keeping the existing variables (search,
normalizedTitle, normalizedContent, normalizedSearch) and the overall if/return
false logic intact so both title and content are properly searched.

---

Nitpick comments:
In `@src/viewer/index.html`:
- Around line 2228-2235: Move the normalization and lowercasing of the search
term out of the items.filter loop: compute a single normalizedSearch once from
state.memories.search (e.g., var normalizedSearch = state.memories.search ?
state.memories.search.normalize("NFKC").toLowerCase() : '') before creating
filtered, then inside the filter callback (the function used to build filtered
from items) use that normalizedSearch instead of recomputing; also keep the
file's var-style declarations (use var for normalizedSearch) to stay consistent
with existing code and remove the redundant toLowerCase() on the
already-lowercased state.memories.search.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 0f0bceb5-4101-4e2a-b5ad-d1f3133b3ce3

📥 Commits

Reviewing files that changed from the base of the PR and between c2f231f and 3819174.

📒 Files selected for processing (1)

• src/viewer/index.html

[rohitg00]

Thanks @kaushalrog — NFKC on both sides of the search comparison is the right call for full-width CJK input. Merged.