Handbook
Task `pw_extractor_synthesize_probe` v1
Synthesizes extract_questions(page) from a full page probe object (title, headings, text/HTML samples) without exemplar chunks. Non-incremental discover path.
Task pw_extractor_synthesize_probe v1
Summary
Synthesizes extract_questions(page) from a full page probe object (title, headings, text/HTML samples) without exemplar chunks. Non-incremental discover path.
Inputs
| Field | Type | Notes |
|---|---|---|
url |
string | Page URL |
operator_hints |
string | Optional; alias page_hints |
page_probe |
object | Required. Probe dict from collect_page_probe-style gather |
temperature |
number | Optional; default 0.1 |
timeout_sec |
int | Optional; default 240 |
User JSON truncated to 120000 bytes.
Output
Same as exemplar task: structure_notes, extractor_python.
Policy
Same transport and parse policy as pw_extractor_synthesize_exemplar v1, with probe-only system prompt stack.