fix(grep): send synonym alternations to the embedding daemon

[eastagiletracker]

This PR proposes sending synonym alternations like data loss|concurrent writer|race condition to the embedding daemon instead of falling back to lexical-only grep (Fixes #86). We include this PR work along with a full history of your repo at https://eastagiletracker.com/projects/747. You can sign in with your GitHub ID to claim ownership of the project.

What changed

patternIsSemanticFriendly counted | as a regex metacharacter and rejected any pattern with more than one, so a grep with two or more alternatives never reached EmbedClient.embed() and searchDeeplakeTables ran the lexical LIKE/ILIKE branch only. A list of synonyms is exactly the query semantic recall is meant to answer, so paraphrase matches were being lost. This follows the fix proposed in #86: | no longer counts toward the metacharacter limit, patterns with more than one other metacharacter (e.g. (foo|bar)\+) are still skipped, and alternations are capped at 8 alternatives so a many-clause pattern still goes lexical. The change is applied to both copies of the function, src/hooks/grep-direct.ts (the grep fast path used by the pre-tool-use hooks) and src/shell/grep-interceptor.ts (the virtual shell), and keeps the two in sync. No version bump is included, so you can decide whether it goes into a release.

Reproduction at current main (e054fb1): a vitest case with the embed client stubbed calls handleGrepDirect(api, "memory", "sessions", { pattern: "silent data loss|concurrent writer|race condition", ... }) and gets 0 embed() calls, with no <#> operator in the generated SQL (lexical-only).

Verification: npx vitest run before and after the change gives the same failure set (4 tests that also fail on a clean main in this environment: embeddings-client warmup auto-spawn, two plugin-cache rmSync/rename cases, and spawn-detached real-process; all spawn or filesystem permission related, none touch grep). After: 5861 passed, the 9 new tests included. tsc --noEmit and jscpd src are clean. New tests: a new grep-direct-semantic.test.ts covers handleGrepDirect (alternation embeds and produces the hybrid <#> query, the 8-alternative limit on both sides, regex-heavy patterns containing | still skip, patterns shorter than 2 characters still skip), and grep-interceptor.test.ts gains the same cases for the shell interceptor. With the source change reverted, the 4 alternation-embedding tests fail and the skip cases stay green.

How this was managed

This work was tracked as https://eastagiletracker.com/projects/747/stories/776572 on https://eastagiletracker.com/projects/747, a board imported from this repo's issues and pull requests (425 stories) and used to manage this change.

If you'd rather not receive contributions like this, reply no-more-prs on this pull request and we won't open any further ones on your repositories.

---

Lawrence W. Sinclair
CEO / East Agile
linkedin.com/in/lwsinclair/
eastagile.com

Summary by CodeRabbit

• New Features
  • Semantic search now supports grep patterns with up to eight pipe-separated alternatives, including common synonym searches.
• Bug Fixes
  • Patterns with excessive regex complexity or more than eight alternatives remain ineligible for semantic search, while simple patterns continue to work as expected.

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository: activeloopai/hivemind/.coderabbit.yaml

Review profile: CHILL

Plan: Advanced

Run ID: 3dba7fce-e674-46be-b02e-24d2e3260362

📥 Commits

Reviewing files that changed from the base of the PR and between e054fb1 and e5e5e48.

📒 Files selected for processing (4)

• src/hooks/grep-direct.ts
• src/shell/grep-interceptor.ts
• tests/claude-code/grep-direct-semantic.test.ts
• tests/claude-code/grep-interceptor.test.ts

Included review availability: This review used your included allowance. Your plan provides up to 4 included reviews per hour; 3 remain after this review.

---

📝 Walkthrough

Walkthrough

Both grep paths now allow semantic embedding for patterns with up to eight pipe-separated alternatives, subject to the existing short-pattern and metacharacter checks. Tests cover alternation limits and embedding behavior.

Changes

Semantic grep eligibility

Layer / File(s)	Summary
Update semantic eligibility and tests src/hooks/grep-direct.ts, src/shell/grep-interceptor.ts, tests/claude-code/grep-direct-semantic.test.ts, tests/claude-code/grep-interceptor.test.ts	Both paths exclude `

Priority: ➖ Normal

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Bug fix · Severity of issue fixed: Medium

Merge Risk: ⚪ Minimal · up to e5e5e

The change enables semantic search for bounded synonym patterns. No issue identified here requires resolution before merge.

Security Architecture Review

Security architecture risk: 🟡 Moderate · up to e5e5e

Bounded synonym searches can now return semantically related records. With the default semantic-output behavior, even a filename-only or count-only search can return record text. The searches remain scoped to requested paths, but this wider exposure merits design review.

Retained concerns

• Medium · security · inferred: Newly eligible alternation searches can enter the existing semantic-output branch, which returns full selected-row text instead of honoring filename-only, count-only, or line-level grep refinement. If a caller expects restricted output, additional memory or session content can reach the agent.

Security review details

Security Blast Radius

• inferred — A caller-controlled alternation can now cause selected memory or session rows within the requested search paths to be returned as text to the agent, including lines outside a literal grep match. The evidence does not establish access outside those paths or across tenants.

Security Findings and Attack Paths

• inferred — For a newly eligible alternation, a filename-only or count-only request can receive full text from retrieved rows when semantic emit-all is active. This expands reachability of a pre-existing behavior; no actual sensitive-record disclosure was verified.

Trust Boundaries and Controls

• observed — The shell checks the virtual mount before search; database queries carry a target-path filter; embedding failures fall back without a vector. These controls limit the supported attack path but do not enforce grep output flags in semantic emit-all mode.

Resilience and Maintainability Implications

• observed — The two eligibility implementations make the same routing change, while their existing semantic-output branches remain separate. Keeping their output controls aligned matters as more patterns reach those branches.

Hardening Proposals

• proposed — Preserve filename-only, count-only, and other output-limiting options after semantic retrieval, or explicitly exclude incompatible options from semantic routing in both entrypoints.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 16.67% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 6 functions across 4 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the primary change: sending grep synonym alternations to the embedding daemon.
Description check	✅ Passed	The description provides a detailed summary, explains the version-bump decision, and documents the test plan and verification results. It does not use the exact template headings or checkboxes and inc…
Linked Issues check	✅ Passed	The PR satisfies the coding requirements in [#86]. Both patternIsSemanticFriendly implementations exclude | from the metacharacter count, reject patterns with more than one remaining metacharacter…
Out of Scope Changes check	✅ Passed	The changes stay within [#86]. The source edits implement the required gating behavior in both specified files. The added tests verify that behavior. No unrelated feature, refactor, or administrative …

• Fix all pre-merge checks with AI

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create a new PR

Warning

⚠️ This pull request has been flagged as potential spam (promotional) by CodeRabbit slop detection and should be reviewed carefully.

---

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.}