You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: apps/docs/content/docs/integrations/file.mdx
+18-6Lines changed: 18 additions & 6 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -55,7 +55,7 @@ Read workspace file objects from selected files, canonical workspace file IDs, o
55
55
56
56
### File Get Content
57
57
58
-
Extract the text content of workspace files selected directly, identified by canonical file ID, or collected from one or more workspace folders.
58
+
Extract workspace file text using the same parser mode as File Search. Use the returned fileId and offset/limit to read the matching line and surrounding context. For documents and spreadsheets, line numbers refer to extracted text, not page numbers or worksheet row numbers.
59
59
60
60
#### Input
61
61
@@ -77,7 +77,7 @@ Extract the text content of workspace files selected directly, identified by can
77
77
78
78
### File Search
79
79
80
-
Search the indexed text of active workspace files for lines matching a query, and return each matching line once with its file ID and line number. By default the query is a regular expression; in exact mode it is matched verbatim and metacharacters are literal. Coverage is what the index currently holds. A term that is not found is only authoritative when "complete" is true AND "indexStatus" reports no skipped or partial files; otherwise it is unknown rather than absent, so re-check before creating something on the assumption it is missing. Narrow the search with folderPaths to confine it to one or more folder trees, which also narrows "indexStatus" to those trees.
80
+
Search the indexed text of active workspace files for lines matching a query, and return each matching line once with its file ID and line number. By default the query is a regular expression; in exact mode it is matched verbatim and metacharacters are literal. Coverage is what the index currently holds. A term that is not found is only authoritative when "complete" is true AND "indexStatus" reports no skipped files; otherwise it is unknown rather than absent, so re-check before creating something on the assumption it is missing. Narrow the search with folderPaths to confine it to one or more folder trees, which also narrows "indexStatus" to those trees.
81
81
82
82
#### Input
83
83
@@ -99,13 +99,13 @@ Search the indexed text of active workspace files for lines matching a query, an
99
99
| ↳ `text`| string | Matching line or bounded match-centered preview. |
100
100
|`count`| number | Number of returned matching lines. |
101
101
|`truncated`| boolean | Whether more matching lines exist beyond the configured hard cap. |
102
-
|`complete`| boolean | Whether indexing has no pending or failed current revisions; skipped and partial coverage is reported separately. |
102
+
|`complete`| boolean | Whether indexing has no pending or failed current revisions; excluded files are reported separately. |
103
103
|`indexStatus`| object | Current workspace search-index coverage by file status. |
104
-
| ↳ `readyFiles`| number | Files whose current revision is searchable. |
104
+
| ↳ `readyFiles`| number | Files whose entire current extracted text is searchable. |
105
105
| ↳ `pendingFiles`| number | Files still waiting to be indexed. |
106
106
| ↳ `failedFiles`| number | Files whose current indexing attempt failed. |
107
-
| ↳ `skippedFiles`| number | Files intentionally excluded because they are unsupported or oversized. |
108
-
| ↳ `partialFiles`| number |Searchable files whose extracted text was truncated by the parser or cap. |
107
+
| ↳ `skippedFiles`| number | Files excluded in full because they are oversized, unsupported, or cannot be completely extracted. |
108
+
| ↳ `partialFiles`| number |Always zero; retained for compatibility. Files are never partially indexed. |
109
109
110
110
### File Fetch
111
111
@@ -380,4 +380,16 @@ Move an existing workspace file into a folder. Moves the file itself; use Move F
380
380
|`fileId`| string | The file that was moved. |
381
381
|`folderPath`| string | The folder the file now lives in. |
382
382
383
+
{/* MANUAL-CONTENT-START:search_limits */}
384
+
## Search coverage and limits
383
385
386
+
Search indexes the complete extracted text of each eligible file. The source file and its extracted UTF-8 text must each be at most **25 MiB (26,214,400 bytes)**. Oversized files, unsupported binary formats, and documents that cannot be completely extracted within parser safety limits are excluded as whole files and counted in `skippedFiles`. Search never indexes only the first rows, lines, or characters. CSV search preserves decoded source text; spreadsheet search includes populated cells beyond the preview limits. Image-only documents require searchable text; search does not perform OCR.
387
+
388
+
Existing parser safeguards also apply to complete extraction. PDFs allow at most 10,000 pages, 20 MiB of extracted text, 250,000 characters on one page, and 60 seconds of extraction. Office archives allow at most 150 MiB expanded in total, 64 MiB for one archive entry, and 10,000 entries; malformed archives and excessive compression ratios are rejected. Hitting any of these limits excludes the whole file from search.
389
+
390
+
Updates are indexed asynchronously. `pendingFiles` and `failedFiles` indicate revisions that are not yet searchable; a new revision becomes searchable only when its full index is ready. An empty result proves absence only within the searched scope when `complete` is true and `skippedFiles` is zero.
391
+
392
+
Regex is evaluated against complete logical lines, including long lines, and cannot span line breaks. Returned lines may use a shortened preview. The result limit (up to 200 lines) and a 10-second query deadline limit an individual request, not the amount of text indexed. An expensive query fails explicitly instead of returning an apparently complete subset; narrow its literal text or folder scope and retry.
393
+
394
+
Search returns one result per matching logical line with `fileId`, 1-based `lineNumber`, and `text` (a bounded preview for long lines). An Agent can use **Search** to locate content, then **Get Content** with the returned `fileId`, `offset` near `lineNumber`, and a small `limit` to read surrounding context. These operations use the same complete-text parser mode. Line numbers for documents and spreadsheets refer to extracted text, not page numbers or worksheet row numbers. Re-run search if the file changes between calls.
0 commit comments