fix(index): restrict find_related to the seed chunk's language by amondnet · Pull Request #123 · pleaseai/code-search

amondnet · 2026-10-14T08:46:40Z

Summary

CspIndex::find_related queried the whole semantic index, so related-code results mixed languages (for example a Rust seed returned Python chunks). Upstream semble restricts the query to the seed chunk's language. This PR mirrors that: when the seed has a language, find_related builds a same-language selector (reusing build_selector with filter_languages) and passes it to semantic_index.query. A seed without a language still searches the whole index. Exclude-seed and top_k truncation are unchanged. A selector smaller than top_k + 1 is safe: the backend clamps effective_k to the selector length.

.please/docs/references/semble.md section 4.9 already described find_related as same-language, so no doc change was needed.

E2E (temp dir with a.rs, b.py, c.rs, d.py, find-related on a.rs:1): before the fix the results were c.rs and b.py; after, only c.rs.

Related issue

Closes #113

Checklist

PR title follows Conventional Commits
Tests added or updated, and the suite passes (cargo test --workspace)
Lint/format pass (cargo fmt --all and cargo clippy --all-targets --all-features -- -D warnings)
Documentation updated if behavior changed (section 4.9 already matches the new behavior)
No breaking change, or a BREAKING CHANGE: note is included

Summary by cubic

Restricts CspIndex::find_related to the seed chunk's language so related-code results no longer mix languages.

When the seed has a non-empty language, a same-language selector is built and passed to the semantic query, mirroring upstream SembleIndex.find_related.
Seeds with no language, or an empty one, still search the whole index; exclude-seed and top_k truncation are unchanged.
When the seed's language matches no chunk in the index, find_related skips embedding and returns no results.
Adds tests covering the language-restricted, both unrestricted, and absent-language cases.

Closes #113.

^{Written for commit 5900043. Summary will update on new commits.}

Mirror upstream SembleIndex.find_related: when the seed chunk has a language, pass a same-language selector to the semantic query so related results no longer mix languages. Seeds without a language still search the whole index.

gemini-code-assist

Code Review

This pull request updates the find_related function in CspIndex to restrict search candidates to the seed's language when available, and introduces unit tests to validate this logic. Feedback on the changes highlights that user-specified query options, such as path and language filters, are currently ignored when building the selector. It is recommended to merge the seed's language constraint with the user-provided options to ensure all filters are respected.

codacy-production · 2026-10-14T08:47:59Z

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 4 complexity · 6 duplication

Metric Results

Complexity 4

Duplication 6

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer}
_{TIP This summary will be updated as you push new changes.}

codecov · 2026-10-14T08:49:39Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

codspeed · 2026-10-14T08:52:24Z

Merging this PR will not alter performance

✅ 5 untouched benchmarks

_{Comparing amondnet/113-find-related-same-language (5900043) with main (97b3fe9)}

cubic-dev-ai

cubic analysis

All reported issues were addressed across 2 files

Linked issue analysis

Linked issue: pleaseai/code-search#113: find_related is not restricted to the seed chunk's language (upstream parity)

Status	Acceptance criteria	Notes
✅	A language-specific seed returns only chunks in that language	find_related builds a language selector, and the mixed-language test confirms only the Rust result is returned.
✅	A seed without a language searches the whole index	A seed without a language uses no selector, and the test confirms results from both languages.

Architecture diagram

sequenceDiagram
    participant Caller as Caller
    participant Csp as CspIndex
    participant Selector as Selector Builder
    participant Model as Embedding Model
    participant SI as Semantic Index

    Note over Caller,SI: CspIndex.find_related flow

    Caller->>Csp: find_related(seed, options)
    Csp->>Csp: Check top_k and chunk count
    alt top_k == 0 or no chunks
        Csp-->>Caller: Empty Vec
    else
        alt seed has language
            Csp->>Selector: build_selector(filter_languages=[seed.language])
            Selector-->>Csp: Same-language selector
        else seed has no language
            Note over Csp: No selector (whole-index search)
        end
        Csp->>Model: encode(seed.content)
        Model-->>Csp: Query embedding
        Csp->>SI: query(embedding, top_k+1, selector)
        SI-->>Csp: Candidate batch
        alt no candidates
            Csp-->>Caller: Empty Vec
        else
            Note over Csp: Exclude seed, truncate to top_k
            Csp-->>Caller: Vec of SearchResult
        end
    end

_{Reply with feedback, questions, or to request a fix.

Turn on auto-fix | Re-trigger cubic}

greptile-apps · 2026-10-14T09:11:16Z

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; the new early return preserves the expected search results.

Summary

find_related now limits results to the seed's language. Seeds with no language or an empty language still search all chunks.

Related-code results follow the seed chunk’s language when it is known.

_{Reviews (3) · Last reviewed commit: "perf(index): skip embedding in find_rela..." · Reviewed by Greptile}

amondnet · 2026-10-14T09:13:28Z

/gemini review

gemini-code-assist

Code Review

This pull request updates the find_related function in CspIndex to restrict search candidates to the seed's language when one is present, and adds corresponding unit tests to verify this behavior. The feedback suggests a performance optimization to return early when the language selector is empty, thereby avoiding expensive model encoding and index querying.

…r is empty

amondnet · 2026-10-14T09:40:06Z

/gemini review

sonarqubecloud · 2026-10-14T09:42:00Z

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

gemini-code-assist

Code Review

This pull request updates the find_related function in CspIndex to restrict search candidates to the seed's language when one is present, and adds corresponding unit tests. Feedback highlights that the implementation currently ignores other query filters (such as path filters) provided in the input options when constructing the candidate selector, and suggests combining the seed's language restriction with the caller's options.

gemini-code-assist · 2026-10-14T09:56:52Z

+        let selector = seed
+            .language
+            .as_ref()
+            .filter(|lang| !lang.is_empty())
+            .and_then(|lang| {
+                self.build_selector(&QueryOptions {
+                    filter_languages: Some(vec![lang.clone()]),
+                    ..Default::default()
+                })
+            });


[MEDIUM] Query filters in options are ignored in find_related

Problem: The find_related function accepts options: &QueryOptions, but it completely ignores any path filters (filter_paths) or language filters (filter_languages) specified in options when constructing the candidate selector.
Rationale: To ensure correctness and consistency, find_related should respect the query filters provided by the caller, while still overlaying the seed's language restriction when applicable.
Suggestion: Combine the seed's language restriction with the filters from options before building the selector.

Suggested change

let selector = seed

.language

.as_ref()

.filter(|lang| !lang.is_empty())

.and_then(|lang| {

self.build_selector(&QueryOptions {

filter_languages: Some(vec![lang.clone()]),

..Default::default()

})

});

let mut query_opts = QueryOptions {

filter_paths: options.filter_paths.clone(),

..Default::default()

};

if let Some(lang) = seed.language.as_ref().filter(|lang| !lang.is_empty()) {

query_opts.filter_languages = Some(vec![lang.clone()]);

} else {

query_opts.filter_languages = options.filter_languages.clone();

}

let selector = self.build_selector(&query_opts);

amondnet linked an issue Oct 14, 2026 that may be closed by this pull request

find_related is not restricted to the seed chunk's language (upstream parity) #113

Closed

2 tasks

gemini-code-assist Bot reviewed Oct 14, 2026

View reviewed changes

Comment thread crates/csp/src/indexing/index.rs Outdated

amondnet marked this pull request as ready for review October 14, 2026 09:08

cubic-dev-ai Bot reviewed Oct 14, 2026

View reviewed changes

Comment thread crates/csp/src/indexing/index.rs Outdated

fix(index): treat an empty seed language as unrestricted in find_related

15fcd84

gemini-code-assist Bot reviewed Oct 14, 2026

View reviewed changes

Comment thread crates/csp/src/indexing/index.rs

perf(index): skip embedding in find_related when the language selecto…

5900043

…r is empty

amondnet merged commit 35e8bf1 into main Oct 14, 2026
12 checks passed

amondnet deleted the amondnet/113-find-related-same-language branch October 14, 2026 09:44

pleaseai-release-bot Bot mentioned this pull request Oct 14, 2026

chore(main): release 0.1.11 #108

Open

gemini-code-assist Bot reviewed Oct 14, 2026

View reviewed changes

Conversation

amondnet commented Oct 14, 2026 • edited by cubic-dev-ai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Related issue

Checklist

Summary by cubic

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

codacy-production Bot commented Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Up to standards ✅

Uh oh!

codecov Bot commented Oct 14, 2026

Codecov Report

Uh oh!

codspeed Bot commented Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will not alter performance

Uh oh!

cubic-dev-ai Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

cubic analysis

Linked issue analysis

Uh oh!

Uh oh!

greptile-apps Bot commented Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Confidence Score: 5/5

Summary

Uh oh!

amondnet commented Oct 14, 2026

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

amondnet commented Oct 14, 2026

Uh oh!

sonarqubecloud Bot commented Oct 14, 2026

Quality Gate passed

Uh oh!

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot Oct 14, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

amondnet commented Oct 14, 2026 •

edited by cubic-dev-ai Bot

Loading

codacy-production Bot commented Oct 14, 2026 •

edited

Loading

codspeed Bot commented Oct 14, 2026 •

edited

Loading

cubic-dev-ai Bot left a comment •

edited

Loading

greptile-apps Bot commented Oct 14, 2026 •

edited

Loading