Ctrlk
For the complete documentation index, see llms.txt. This page is also available as Markdown.

Local Uploads

Upload files from your computer, and what each file kind becomes.

The fastest way in. Use local uploads for one-time analyses, spreadsheets from stakeholders, or exports from systems that have no connector.

Supported formats

Kind
Formats
Becomes

Tabular

CSV, TSV

A tabular source, one table per file

Spreadsheets

XLSX, XLS, ODS

XLSX becomes a workbook source kept whole; XLS and ODS are read one table per sheet. See Excel & Spreadsheets.

Columnar

Parquet

A tabular source. Recommended for large files.

Semi-structured

JSON, JSONL, NDJSON

A tabular source

Documents

PDF, Word (.docx), PowerPoint (.pptx), Markdown, text

A document source. See Documents.

Images (Beta)

PNG, JPG, JPEG

An image source

Video

MP4, MOV, AVI, MKV, WEBM

A video source

There is no fixed size cap on uploads. Files stream in chunks and an interrupted upload resumes rather than starting over. For very large tables, Parquet uploads noticeably faster than CSV.

Uploading

  1. In the Import Manager, click Local File Upload.

  2. Drag files or whole folders anywhere onto the drop zone, or use Browse files to pick individual files and Browse folder to upload everything inside a folder.

  3. Check the staged list. It shows each file's format and size plus a total. Remove a file with the cross.

  4. Choose how concepts are handled:

    • Auto-extract (default): concepts are extracted from every file and matched to your schema automatically.

    • Bind to schema: skips extraction. You bind each file's columns to the canonical schema yourself after it lands.

  5. Click Upload (n).

Local File Upload with four staged files, concept handling toggle and Upload button
Four files staged: names, formats, sizes and a total, with the Auto-extract / Bind to schema choice and the Upload button.

Unsupported files are refused by name at staging time, so a mixed folder drop keeps its valid files. Zip archives are among them: extract the archive and upload the files inside.

What happens on upload

  1. The file is stored in your organisation's data lake.

  2. Schema detection reads the columns and infers types.

  3. A preview is generated for verification.

  4. The upload creates a source, ready for datasets, ontology mapping and analysis.

Current Imports tab with per-file progress and row counts
Current Imports: each file's progress, row and column counts, and concept handling.

Replacing and extending a source

An existing file source can be replaced (new file, same source identity, downstream work intact) or extended (rows appended) from its detail page. Both modes accept a single file and keep the source's existing concept handling; extending or replacing cannot change it.

Uploading over the API

File uploads, replacements and extensions are all available over the REST API (the workspace Sources endpoints), so a nightly export job can push files without touching the UI. The Python SDK wraps the same calls.

A connected AI client uploads with upload_files: it asks for one single-use upload link per file and sends each file itself, or passes a small text file inline. Several files can go in one request, grouped under a new folder. See Uploading files.

Next steps

PreviousSources Overview
NextExcel & Spreadsheets

Last updated