Fix: Search methods configuration ignored (#788)

[liebertar]

I found that the search configuration issue could be resolved by making the search functions actually respect what users configure.

The problem was that both edge_search and node_search were always running all three search methods regardless of what was specified in the config. So even if you only wanted BFS, you'd still get BM25 and cosine similarity running in the background - not ideal for performance.

I've updated the code to build the search task list based on what's actually in config.search_methods. Now it only runs what you ask for. Pretty straightforward fix that should help with the performance concerns mentioned in the issue.

Ran all the tests and linting - everything looks good. Would be great if someone could review this approach and see if it makes sense.

Fixes #788

---

Important

edge_search and node_search now respect configured search methods, improving performance by executing only specified methods.

• Behavior:
  • edge_search and node_search in search.py now respect config.search_methods, executing only specified search methods (BM25, cosine similarity, BFS).
  • Previously, all search methods ran regardless of configuration, impacting performance.
• Misc:
  • Update uv.lock version from 0.18.4 to 0.18.5.

^{This description was created by for 19833bc. You can customize this summary. It will automatically update as commits are pushed.}

[ellipsis-dev]

Important

Looks good to me! 👍

Reviewed everything up to 19833bc in 1 minute and 4 seconds. Click for details.

• Reviewed 144 lines of code in 2 files
• Skipped 0 files when reviewing.
• Skipped posting 6 draft comments. View those below.
• Modify your settings and rules to customize what types of comments Ellipsis leaves. And don't forget to react with 👍 or 👎 to teach Ellipsis.

1. graphiti_core/search/search.py:177

• Draft comment:
  Good job conditionally building the search_tasks list in edge_search. Note that if config.search_methods is empty, no tasks will run and an empty result will be returned. Ensure this behavior is documented if intentional.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 80% None

2. graphiti_core/search/search.py:289

• Draft comment:
  Similarly, node_search now builds its search_tasks based on config.search_methods. If no methods are configured, it returns empty results. Consider documenting this behavior or logging a warning for clarity.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 80% None

3. graphiti_core/search/search.py:1

• Draft comment:
  Ensure that a proper copyright header is present at the top of the file.
• Reason this comment was not posted:
  Comment was not on a location in the diff, so it can't be submitted as a review comment.

4. graphiti_core/search/search.py:177

• Draft comment:
  There is a repeated pattern between edge_search and node_search for building search_tasks. Consider refactoring the duplicate logic into a shared helper function to adhere to the DRY principle.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 80% None

5. uv.lock:2

• Draft comment:
  The version bump (revision from 2 to 3 and graphiti-core version from 0.18.4 to 0.18.5) looks appropriate. Ensure downstream dependencies are updated accordingly.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 80% None

6. graphiti_core/search/search.py:219

• Draft comment:
  Line 219 appears to be incomplete. The call to 'search_results.append(' is missing an argument or closing parenthesis, which might cause a syntax error. Please review and complete the intended functionality.
• Reason this comment was not posted:
  Comment was not on a location in the diff, so it can't be submitted as a review comment.

Workflow ID: wflow_6hIP0aOdLIU3m1gb

^{You can customize by changing your verbosity settings, reacting with 👍 or 👎, replying to comments, or adding code review rules.}

[liebertar]

Hello @prasmussen15, I’ve run several tests and everything looks good on my side—hope this is helpful. When you have a moment, could you please review this and pr #816? Thank you!

[liebertar]

Hi @prasmussen15,

I've rebased the branch on the latest main to resolve the merge conflicts. The changes are ready for review.

Regarding the CI failures:

1. Claude Code Review: Failing due to OIDC token configuration issue in the GitHub Actions workflow (not related to this PR's changes)
2. Unit Tests: The voyage embedder test fails because it's missing proper optional dependency handling - this issue exists in the main branch as well

The actual changes in this PR (respecting search_methods configuration) are working correctly and all relevant tests pass locally. The PR implements the fix for #788 as intended - now edge_search and node_search only execute the search methods specified in the configuration instead of always running all three methods.

[prasmussen15]

Thanks for the contribution! I will merge this in and it will go out with the next release