Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/exegia/corpora-py/llms.txt

Use this file to discover all available pages before exploring further.

search() is the primary query tool for Context-Fabric. You write a template string that describes a structural pattern — node types with feature conditions and indentation-based containment — and the engine returns every place that pattern occurs in the corpus. In the default "results" mode the tool returns the first page of matching node tuples along with their section references and a cursor_id you can hand to search_continue() to page through the rest. Other return modes let you get a quick total count, a node-type breakdown, or rendered passage text directly. Consult search_syntax_guide() for full template syntax documentation.

​
Parameters

​
template
string
required
The query template string. Uses indentation to express containment and space-separated feature=value conditions on node lines. For example:
verse
  word pos=verb
finds every verse that contains at least one verb. See search_syntax_guide() for the complete syntax.
​
return_type
string
Controls what the tool returns. One of:
  • "results" (default) — Node tuples with section references for each matched result, plus a cursor_id for pagination when the total exceeds limit.
  • "count" — A single line with the total number of matches: "Total results: N".
  • "statistics" — A frequency breakdown of every node type present across all matched tuples.
  • "passages" — Rendered text for the first limit matched nodes, using the format specified by fmt.
​
limit
integer
Maximum number of results to include in the response. Clamped to the range 1–100. Defaults to 20. In "results" mode, remaining results are accessible via search_continue().
​
fmt
string
Text rendering format, used only when return_type="passages". Pass a format name from get_text_formats() (e.g. "text-orig-full"). When omitted, the corpus default format is used.
​
corpus
string
Corpus name. When omitted, the currently active corpus is used.

​
Returns

The shape of the return value depends on return_type:
  • "results" — A header line "Results: N total, showing M", followed by numbered lines of pipe-separated section references for each result tuple, and a cursor_id: line when further results exist. The cursor expires after 5 minutes of inactivity.
  • "count" — "Total results: N".
  • "statistics" — "Result statistics (total=N):" followed by a frequency table of node types.
  • "passages" — One [section ref] text line per result, followed by (showing M of N) when truncated.
Returns "Search error: <message>" if the template cannot be parsed or executed.

​
Example

# Find the first 5 verses containing a verb
result = search(
    template="verse\n  word pos=verb",
    return_type="results",
    limit=5,
)
print(result)
Expected output: 
Results: 4,892 total, showing 5

     1. Genesis 1 1 | Genesis 1 1
     2. Genesis 1 2 | Genesis 1 2
     3. Genesis 1 3 | Genesis 1 3
     4. Genesis 1 4 | Genesis 1 4
     5. Genesis 1 5 | Genesis 1 5

cursor_id: 3f2a1b4c-8d9e-4f0a-b1c2-d3e4f5a6b7c8

# Just get the count
result = search(template="word pos=verb", return_type="count")
print(result)
Expected output: 
Total results: 34,611

Build docs developers (and LLMs) love

Get started for freeTalk to us