Build the user prompt string for a single chunk.

Wrap *body* in a Response with ETag + Cache-Control headers. The ETag is ``<db_sha256[:16]>-<app_version>`` — two deploy-wide constants, so i

Lazily create the run dir for commands that need one. Cached on ctx.

Log a summary of batch results and return the exit code.

Build an LLMExtractor whose provider is a MagicMock. Patching make_provider sidesteps eager client construction (e.g. OpenAIProvider's OpenAI

Build a debug_info dict for an option-backed match.

Run _run_diff_db and return captured log lines.

Split text into sections at markdown header lines (# or ##). Returns a list of (start_line, section_text) tuples where start_line is the 1-in

Run *query* with each params tuple, cycling through params_list for *iterations* total executions. Return per-query times in seconds.

Signal all in-progress extract() calls to stop after their current LLM request completes. Does not abort already in-flight HTTP calls,

Expand a list of files/directories into absolute .gz file paths. Paths starting with '@' are treated as files containing one path per line.

* Create key-value caches of limited size * @returns {Function(string, Object)} Returns the Object data after storing it on itself with * property n

end the most recently created command group by popping it from the group stack. groups are created by visitcommand or a nested command

Format token count for display (e.g. 878K, 1.8M).

Pick one of the bot templates uniformly at random and generate a cmd.

* For feature detection * @param {Function} fn The function to test for native support

Run LLM extraction via provider batch API. Files are finalized as soon as their batch completes (per-batch), not after all batches finish. T