Technical referenceOfficial documentation7 min readReference data page

FastGPT File Format and Parsing Reference (25 extensions, v4.17.0)

Check FastGPT v4.17.0 file input and parser formats, required services, and the limits to verify with your own documents before ingestion.

This page lists the file parsing capability recorded in the repository: built-in parsers cover 8 extensions and the supplementary parser covers a further 17, 25 in total. Use it before uploading to confirm which parsing path a format takes, and when parsing fails to tell whether the cause is format support, a size ceiling, or an external parsing service.

How the two parsing paths differ

PathExtensions coveredNotes
Built-in parsers8Handled by parser functions inside the repository, with no external dependency
Supplementary parser17Formats the built-in parsers do not cover; converted to Markdown before ingestion

Formats covered by built-in parsers

ExtensionParser function
.csvreadCsvRawText
.docxreadDocsFile
.htmlreadHtmlRawText
.mdreadFileRawText
.pdfreadPdfFile
.pptxreadPptxRawText
.txtreadFileRawText
.xlsxreadXlsxRawText

Formats covered by the supplementary parser (17)

| Extension |

| --- |

| .doc |

| .docm |

| .epub |

| .odp |

| .ods |

| .odt |

| .pot |

| .pps |

| .ppsm |

| .ppsx |

| .ppt |

| .pptm |

| .rtf |

| .wps |

| .xls |

| .xlsb |

| .xlsm |

Extensions outside these lists are rejected, and the error message names the extension.

Embedded-image limits in the supplementary anydoc parser

ItemExpression in the definitionValue
Per embedded image10 * 1024 * 102410,485,760 bytes (10 MiB)
Embedded images in total200 * 1024 * 1024209,715,200 bytes (200 MiB)
Embedded image upload concurrency55

Environment variables related to upload and parsing

VariableDefaultNotes
FILE_TOKEN_KEY—Secret for file reading
FILE_DOMAIN— (deployment-specific)File domain (also points to FastGPT service); Assign independent domain for higher security to prevent high-risk file reading from affecting main domain content
SKIP_FILE_TYPE_CHECKfalseWhether to skip file type validation
PARSE_FILE_TIMEOUT_SECONDS600File parsing timeout (seconds)
UPLOAD_FILE_MAX_SIZE1000Max upload file size (MB)
UPLOAD_FILE_MAX_AMOUNT1000Max number of upload files
CUSTOM_PDF_PARSE_URL—Custom PDF parsing service address
CUSTOM_PDF_PARSE_KEY—Custom PDF parsing service secret key
DOC2X_KEY—Doc2x PDF parsing service secret key
DATASET_PARSE_MAX_PROCESS10Max concurrent knowledge base file parsing queue size

External document parsing services

| Service directory |

| --- |

| doc2x |

| somark |

| textin |

Scope of this reference

The table reflects the parsing paths recorded in the repository. The following cases need separate confirmation:

  • How internal differences within one extension (a scanned versus a text PDF, for example) affect the parsing result
  • Availability and quota of external parsing services, which those services control
  • Parsing capability added by the commercial edition, which is out of scope here