sercha search

Performs hybrid search across all indexed documents, combining keyword (BM25) and semantic (vector) search for best results.

Usage

sercha search [query] [flags]

Arguments

Argument	Required	Description
query	Yes	The search query string

Flags

Flag	Short	Default	Description
--limit	-n	10	Maximum number of results to return
--json		false	Output results as JSON

Examples

# Basic search
sercha search "kubernetes deployments"

# Limit results
sercha search "API authentication" -n 5

# JSON output for scripting
sercha search "error handling" --json

Output Formats

Table format (default)

Results:

  [1] Document Title (0.85)
      Matching snippet from the document...

  [2] Another Document (0.72)
      Another matching snippet...

JSON format

When using --json, results are output as a JSON array for use in scripts:

[
  {
    "document": {
      "id": "doc-123",
      "title": "Document Title",
      "content": "..."
    },
    "score": 0.85,
    "highlights": ["Matching snippet..."]
  }
]

Verbose Mode

Use -v for detailed search execution logs:

sercha -v search "go language"

Example verbose output:

=== Search Execution ===
[DEBUG] Query: "go language"
[DEBUG] Limit: 10, Offset: 0
[INFO] Effective search mode: Full (text + semantic + LLM)
[DEBUG] Services available: keyword=true, vector=true, embedding=true, llm=true
[DEBUG] Full search: LLM query rewrite for "go language"
[INFO] Full search: expanded query="go language OR golang OR programming"
[DEBUG] Hybrid search: running keyword and vector searches in parallel
[DEBUG] Keyword search: 3 hits
[DEBUG] Vector search: 10 hits
[DEBUG] Hybrid search: merging 3 keyword + 10 vector results with RRF
[INFO] Final results: 10
Results:
  ...

Verbose output includes:

• Search mode selection and fallback behaviour
• Query expansion by LLM (if configured)
• Keyword and vector search execution details
• Result merging statistics (RRF fusion)