---
argument-hint: [optional additional instructions]
description: Run QA script for ChunkHound
---
You are an expert quality assurance tester.
Perform structured QA on the `semantic_search` and `regex_search` tools.
TEST SUITE:
1. Pick existing file → search → verify results
2. Add new file → search → verify results
3. Edit existing file → search → verify results
- Test: add content, delete content, modify content
4. Delete file → search → verify results
5. Run tests 1-4 concurrently for all supported languages/file types
6. Language coverage:
- List all supported languages, parsers, and extensions
- Test every language systematically
7. Run multiple edits → immediate searches → verify no blocking
8. Test pagination for both semantic and regex search:
- Non-existing value (no results)
- Single occurrence (no pagination)
- Multiple occurrences → traverse all pages → validate against ripgrep
CRITICAL: STOP IMMEDIATELY ON ANY FAILURE - DOCUMENT THE EXACT FAILURE.
EXECUTION NOTES:
- Wait 2-3 seconds after changes for embedding generation
- Measure indexing latency (change → searchable)
- Create all test files in current directory for indexing
- External MCP server - don't attempt to stop it
- Use existing tools only - no helper scripts
- You are an AI agent performing the QA of the MCP server as a USER
- You have only external understanding of the tool usage and expected behavior.
You test from the outside.
DELIVERABLE:
Concise report optimized for LLM ingestion:
- Pass/fail status for each test
- Indexing latency measurements
- Any failures with exact reproduction steps
---
$ARGUMENTS
