Add similarity scores and threshold filtering to Redis semantic search

[thepatrickchin]

Description

Add similarity_score field to MemoryItem and implement threshold-based filtering in Redis search method with similarity_threshold kwarg

Closes #1433

By Submitting this PR I confirm:

• I am familiar with the Contributing Guidelines.
• We require that all contributors "sign-off" on their commits. This certifies that the contribution is your original work, or you have rights to submit it under the same license, or a compatible license.
  • Any contribution which contains commits that are not Signed-Off will not be accepted.
• When the PR is ready for review, new or existing tests cover these changes.
• When the PR is ready for review, the documentation is up to date with these changes.

Summary by CodeRabbit

• New Features

  • Added similarity-based filtering for search results via a configurable similarity threshold.
  • Search results now surface per-item similarity scores as part of returned memory items.

• Tests

  • Added tests verifying similarity_score is recorded and that threshold-based filtering excludes appropriate results.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Walkthrough

Adds similarity score handling to Redis semantic search: a similarity_threshold search kwarg is accepted and used to filter results by doc.score; each returned MemoryItem now carries an optional similarity_score field.

Changes

Cohort / File(s)	Summary
Redis Editor Implementation packages/nvidia_nat_redis/src/nat/plugins/redis/redis_editor.py	Adds similarity_threshold to search kwargs and logs it. Extracts similarity_score from doc.score, skips items failing the threshold, passes similarity_score into _create_memory_item(), and updates logging to report processed vs filtered counts. Updated _create_memory_item() signature to accept `similarity_score: float
Redis Editor Tests packages/nvidia_nat_redis/tests/test_redis_editor.py	Adds assertion that similarity_score is set for first search result. New test test_search_with_similarity_threshold_filters_results mocks docs with scores [0.2, 0.6, 0.4] and verifies threshold filtering yields only the expected subset; adjusts mocked JSON fetches accordingly.
Data Model src/nat/memory/models.py	Adds `similarity_score: float

Sequence Diagram(s)

sequenceDiagram
  participant Client
  participant RedisEditor as Editor
  participant EmbeddingService as Embeddings
  participant Redis
  participant Storage as JSONStore

  Client->>Editor: search(query, top_k, similarity_threshold?)
  Editor->>Embeddings: embed(query)
  Embeddings-->>Editor: vector
  Editor->>Redis: semantic search(vector, top_k)
  Redis-->>Editor: docs with doc.score
  Editor->>Editor: filter docs by similarity_threshold (doc.score <= threshold)
  Editor->>JSONStore: fetch JSON for kept docs
  JSONStore-->>Editor: doc JSONs
  Editor->>Client: [MemoryItem{..., similarity_score}] list

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main changes: adding similarity scores and threshold filtering to Redis semantic search, matching the PR's core objectives.
Linked Issues check	✅ Passed	The PR fully implements both requirements from issue #1433: adds similarity_score field to MemoryItem and implements similarity_threshold parameter for filtering in Redis search.
Out of Scope Changes check	✅ Passed	All changes are within scope: MemoryItem model extended with similarity_score, Redis editor implements threshold filtering and score propagation, and tests validate both features.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

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

✨ Finishing touches

• 📝 Generate docstrings

---

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.}

[willkill07]

/ok to test c6f419d

[willkill07]

Documentation build appears to have failed.

The runner failure is a known transient issue -- working on a fix for that separately.

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@packages/nvidia_nat_redis/src/nat/plugins/redis/redis_editor.py`:
- Around line 126-129: Validate and coerce similarity_threshold from kwargs
(e.g., in the block where similarity_threshold = kwargs.get(...)) by checking
isinstance/try-converting to float and setting a safe default or None on
failure; then when iterating search results (references to doc.score in the
vector search loop), ensure doc.score is not None and is numeric before
converting to float or comparing—skip or treat non-numeric scores as
below-threshold. Update the comparison logic to only compare when both floats
are valid and use a defined fallback behavior (e.g., treat invalid threshold as
no threshold or treat invalid score as failing the threshold check).

[willkill07]

/ok to test bf1292d