Use “Search your documents” in Recued. Search every PDF in a folder for a term and record which documents matched, with the page and the matching line. Each file is first checked as readable, so a document that could not be opened is reported as unreadable rather than counted as a miss.
Search your documents
Use “Search your documents” in Recued. Search every PDF in a folder for a term and record which documents matched, with the page and the matching line. Each file is first checked as readable, so a document that could not be opened is reported as unreadable rather than counted as a miss.
Complete workflow
Document Search
4 pinned recipes
Use Document Search in Recued for PDF, retrieval, pdfgrep, and records. It includes 11 built-in actions and 3 ready-to-run workflows. Actions that change data use Recued's approval controls.
You are viewing one recipe in this workflow. Installing it opens the complete pack so its other recipes and background automations arrive together.
Use “Show past document searches” in Recued. List search runs with their counts, including how many documents could not be opened. Read-only — it re-reads no file and re-runs no search.
Use “Show what a search found” in Recued. List the documents one search run matched, with the page and the matching line, alongside the documents that could not be opened. Read-only.
Use “Records runtime” in Recued. It helps you complete this workflow using the connections and settings you choose.
Why a match is read from the OUTPUT, not the exit status
⛔⛔ item.ok does NOT mean "matched". pdfgrep follows grep — a match exits 0, no match exits 1, an unreadable file exits 2 — and the pdfgrep pack declares exit_code_handling: { success_codes: [0, 1] }, so it deliberately treats no match as a SUCCESS. A step that found nothing therefore comes back ok: true with empty stdout. Keying matched on ok marks every readable document as a hit; that was a real defect here, caught only by driving the pack live against real PDFs.
So a match is excerpt is_not_empty, and readability is a separate pdfinfo probe (exit 0 readable, 1 corrupt — verified live). The two together give three outcomes:
| pdfinfo | pdfgrep stdout | recorded as |
|---|---|---|
| :-- | :-- | :-- |
| ok | non-empty | matched — with the page and line |
| ok | empty | no match — a real, trustworthy negative |
| non-zero | — | unreadable — never counted as a miss |
⚠ A search is only as trustworthy as its unreadable count. If that number is above zero, the answer "nothing mentions this" is not yet true — those files were never actually searched. The card reports it separately and the board shows it per run for exactly that reason.
How the search runs
⚠ The foreach is what makes this safe. A foreach step collects per-item failures as {ok:false} and still returns normally, so one unreadable file among two hundred cannot end the run. A non-foreach search step would have died on the first document that did not match.
⚠ Matching is case-insensitive and treats the term as a REGULAR EXPRESSION. pdfgrep -i is used, so renew finds Renewal; but . matches any character and ( will error. To search for a literal phrase with punctuation, this is not the recipe — the fixed-string operation exists in the pdfgrep pack but is not wired here.
⚠ Results are capped: at most Max files documents are searched per run, and pdfgrep itself stops at 50 matching lines per document. Anything beyond the cap is reported on the card as skipped, never silently dropped.
⚠ Only the matching line is stored, not the document. The excerpt is what pdfgrep printed — a page number and one line.
How it works
Inspect the data fetches, transforms, gates, and output this recipe runs.
Process (38 steps)
term
trim
Trim whitespace from setting term
folder
trim
Trim whitespace from setting folder
inputs_valid
all
Check if all conditions are true
inputs_guard
guard
Stop if inputs valid equals
found
?
found_text
default
Apply default
found_trimmed
trim
Trim whitespace from found text
all_paths
split
Split found trimmed into parts
files_found
count
Count items in all paths
paths
slice
Take a subset of
files_searched
count
Count items in paths
files_skipped
math
Calculate files found minus files searched
readable_probe
?
search_probe
?
readable_rows
map
Extract from each item
search_rows
map
Extract from each item
joined
enrich_by
Apply enrich_by
unreadable_rows
filter
Filter by condition
readable_only
filter
Filter by condition
matched_rows
filter
Filter by condition
missed_rows
filter
Filter by condition
files_matched
count
Count items in matched rows
files_unreadable
count
Count items in unreadable rows
files_missed
count
Count items in missed rows
trustworthy
compare
Check if files unreadable equals 0
caveat
ternary
Apply ternary
ran_at
date_add
Add 0 seconds to
search_create
?
search_id
default
Apply default
search_recorded
compare
Check if search id has data
search_guard
guard
Stop if search recorded equals
search_ref
template
Generate text from a template
matched_with_ref
map
Extract from each item
unreadable_with_ref
map
Extract from each item
rows_to_record
flatten
Flatten nested lists in
hit_create
?
recorded_count
count
Count items in rows to record
card
to_summary
Format results as a summary card
Settings
Configurable at install. Defaults shown — change them anytime in Recued.
term
setting
=
[object Object]
folder
setting
=
[object Object]
max depth
setting
=
1
max files
setting
=
50