Skip to main content
Purna opens supported files in your workspace, either in a scientific component or in a general viewer. This page lists every supported file type, which viewer opens it, and what Purna’s AI can read from it. Purna reads your files in two ways:
  • Chat assistant: the model that answers you in a conversation. It reads small text files directly. For other files it gets only metadata, such as the name, type, and size.
  • Background agents: agents that run Code Execution and background jobs in a private compute workspace. Files you attach or approve are copied into that workspace, so code can open them.
File previews are read-only. Opening a file never changes it, including code, text, and notebook files. When you ask Purna to modify data, it saves the result as a new file in Files.

​
How files open

​
Supported file types

Upload shows whether you can upload the file yourself. Some formats can be displayed only when Purna generates them.

​
Structure and chemistry

​
Sequence and genome

​
Omics and imaging

​
Tables, documents, and code

​
Not accepted for upload

Purna does not accept these formats for upload yet:
  • Alignments and tracks: BAM, CRAM, bigWig, and bigBed
  • Columnar and R data: Parquet, Arrow, Feather, RDS, and Loom
  • Instrument and raw data: mzML, vendor RAW, FCS, POD5, FAST5, and AB1
  • Microscopy and volumes: ND2, CZI, LIF, MRC, CCP4 maps, NIfTI, and DICOM
  • Simulation: XTC, TRR, and DCD trajectories
  • Clinical and physiology: XPT, SAS, EDF, and NWB
  • Other: PowerPoint, GraphML, OFF meshes, and XLSB workbooks
To work with one of these formats, convert it to a supported format first. For example, export BAM regions to SAM, or Parquet to CSV.

​
What the AI can read

When the chat assistant gets only metadata for a file, it hands analysis of the contents to a background agent. Files attached to the conversation are included when Purna plans the background job. Files attached to earlier messages are passed by reference; attach a file again if you want the chat assistant to reread its text.
The chat assistant reads only part of large text files, at most the first 256 KB. For whole-file statistics or summaries, ask Purna to use a background agent.

​
Limitations

​
Viewer limits

​
What viewers do not do

  • Spreadsheets show the values saved in the file. They don’t recalculate formulas or show charts, formatting, or macros.
  • Notebooks show the outputs saved in the file. They never run code, and interactive or HTML outputs are hidden.
  • Markdown documents over 256 KB render section by section. Footnotes may not link across sections.
  • Large text files in chunked mode show line numbers without syntax highlighting.
  • Tables use fixed-width columns. Hover over a truncated cell to see its full value.
  • Phylogenetic trees, SMILES, and diagram files show their source text. They aren’t drawn as trees, molecules, or diagrams.
  • Genome annotation and variant files show as text, not as genome-browser tracks.
  • glTF files must embed all their data. Files that reference separate texture or buffer files don’t open; export them as .glb instead.
  • Plain TIFF files have no preview. OME-TIFF files open in a scientific viewer.

​
What the AI cannot do

  • The chat assistant can’t see the graphics in a viewer, such as a rendered structure, slide, or mesh. Share a selection or describe what you need.
  • Background agents get read-only copies of your files and save results as new files. They never change your originals.
  • Background agents have no specialized library for meshes, so they can work only with the raw file contents.
  • Each agent’s work is your responsibility to review. Check key results before you rely on them.
  • Files covers uploading, organizing, and downloading files.
  • Scientific Components describes each scientific viewer in detail.
  • Code Execution explains how background agents run analyses on your files.