CATO

Data Management

Bring source evidence into a project, verify what CATO detected, and preserve a traceable path to every output.

The Data Lifecycle

  1. 1

    Ingest

    Upload source data files.

  2. 2

    Inspect

    Review type, dimensions, schema, and warnings.

  3. 3

    Organize

    Use folders and descriptive names.

  4. 4

    Analyze

    Select exact data for Chat or a workspace tool.

  5. 5

    Trace

    Keep versions, provenance, code, and generated files.

Upload Files

  1. Open Files and select Upload, then choose a file or a whole folder; or drop files or a folder onto the page; or press Ctrl/Cmd+U.
  2. Wait for storage and metadata processing to complete before starting an analysis.
  3. Open each dataset and confirm the detected format, schema, preview, and any parsing warning.
  4. Rename generic files and add project folders where that makes the study structure clearer.

A successful upload means the object was stored; it does not prove that dates, category codes, decimal separators, missing-value markers, character encoding, or scientific units were interpreted correctly.

Supported Format Families

Tables and statistics

CSV, TSV, XLSX/XLS, Parquet, Feather, SAS7BDAT, SPSS SAV, Stata DTA

Structured and documents

JSON, PDF, SQL, TXT, Markdown, logs, Python, R, and shell scripts

Genomics and single-cell

FASTA/FA, FASTQ/FQ, VCF, GFF/GFF3, GTF, BED, GenBank, GMT, Newick, H5AD, MTX, Loom

Molecular and imaging

PDB, CIF, SDF, MOL, MOL2, SMILES, NIfTI, DICOM

Proteomics and cytometry

mzML, MGF, FCS, HDF5/H5, BIOM

Scientific and graph data

MATLAB MAT, NumPy NPY/NPZ, GML, GraphML

Images

PNG, JPEG, GIF, SVG, and WebP

Archives

ZIP, TAR, TAR.GZ/TGZ, and supported compressed variants

Format recognition and interactive preview support are separate. A file may be available to analysis tools even when the browser shows metadata rather than a full visual preview.

File and Archive Limits

There is no per-file upload size limit on any plan; files above 64 MB upload directly to storage in parts.

Archives receive additional safety checks: a compressed archive is limited to 500 MB and extracted content to 2 GB. Unsafe paths, excessive expansion, unsupported encryption, corrupt members, or too many entries can cause extraction to fail.

Folders and Naming

A useful project structure

01_raw/
02_dictionary/
03_processed/
04_analysis/
05_figures/
06_manuscript/

Naming rules

  • • Use stable, descriptive names.
  • • Put dates in ISO order: YYYY-MM-DD.
  • • Encode cohort or stage, not patient identity.
  • • Keep source files immutable where possible.

Generated Files and Cross-Service Handoffs

Data created by a sandbox run (cleaned tables, model outputs, notebooks, figures, and reports) can be persisted as project artifacts. CATO records their project and conversation context so the frontend, backend, and analysis worker resolve the same authorized object.

  • Open a generated artifact from the response card or Workspace.
  • Use the selected version, not merely the latest filename, when reproducing a result.
  • Review provenance, logs, and environment information before accepting the result.
  • Promote a generated dataset to the next analysis only after validating its row count, columns, and transformations.

Artifact Versioning

Supported files, figures, notebooks, annotations, and manuscripts retain version history. A selected version can be represented in the workspace URL so collaborators can return to the intended state. Unsaved-edit guards prevent accidental version changes or tab closure while local edits are pending.

Version history is not source validation. It records what changed inside CATO. Retain authoritative source records, data dictionaries, consent and governance records, and external checksums as required.

Preview and Quality Checks

Tabular data

Rows, columns, types, missingness, duplicated keys, impossible values, category levels, and units.

Images and figures

Dimensions, orientation, labels, color accessibility, cropping, and export fidelity.

Genomics and molecular data

Reference build, coordinate convention, identifiers, sample order, chain/model selection, and units.

Documents and literature

OCR quality, page order, citation metadata, available full text, and license or access status.

Delete and Recovery

Deletion is an intentional project action and may affect conversations, workspace references, generated descendants, or reproducibility. Before deleting a source or project, export required materials and verify that downstream manuscripts and analyses no longer depend on it. Confirmation dialogs reduce accidents but are not a backup system.

Security and Privacy

  • Upload only data your account and institution are authorized to process.
  • Minimize direct identifiers and secrets before upload.
  • Analysis runs execute in isolated sandboxes with project and object authorization boundaries.
  • Data is transported over TLS and protected at rest by the production storage services.
  • Consult the current Privacy Policy and institutional agreement for processing and retention terms.