Skip to main content

​
Sessions

A session contains a research request, the conversation around it, and the work Purna performs. Follow-up messages can reuse the session’s plan, files, execution history, and generated outputs. Start another session when you are beginning a separate research objective or need an independent context.

​
Workflow modes

Workflow modes change how a request is handled. Bioinformatics is a supported analysis workflow rather than a separate mode. Add the relevant data and describe the biological comparison, method, thresholds, and outputs you need.

​
Capabilities and knowledge

The controls beside the composer determine what Purna can use for the current request.
  • Capabilities add actions such as code execution, scientific image generation, and specialized biological models.
  • Skills and knowledge provide scientific sources, connected reference libraries, and source-aware guidance.
Some entries are always available, while others depend on the account plan, organization configuration, or provider access.

​
Plans and approvals

For complex tasks, Purna may ask clarifying questions and prepare a detailed plan. The plan can define:
  • Approach and ordered work
  • Assumptions and constraints
  • Required capabilities or providers
  • Validation criteria
  • Deliverables
  • Overall success criteria
Execution starts after approval. You can approve, request changes, or cancel.

​
Projects

Projects group related sessions, files, papers, jobs, and results around a study or research direction. Select a project when you want new work to share that context.

​
Files

Files include uploaded datasets and documents as well as generated scripts, tables, images, reports, and structures. Attach a file to a message or use @ to mention an existing file.

​
Compute

The Compute section of the session workspace shows execution associated with the session:
  • A private compute instance used for interactive code and data handling
  • Background jobs used for longer multi-step work
  • Deep Research runs
Use View reasoning or open a task from Compute to inspect its timeline and generated files.

​
Runtime categories

  • Compute Runtime covers general computation and data handling.
  • Biocompute Runtime covers specialized biological computation such as structure prediction and genomic, proteomic, or multi-omic analysis.
A single request can use both categories. Each operation is metered according to the runtime that performs it. See Usage and billing.

​
Outputs and artifacts

Purna can produce:
  • Cited answers and evidence summaries
  • Editable papers and reports
  • Python, R, and other code files
  • CSV, JSON, and spreadsheet outputs
  • Scientific figures and quality-control plots
  • Molecular structure files and viewers
Open outputs from the session or Files, then continue refining them with follow-up requests.

​
Sources and citations

When Purna uses external evidence, citations connect important claims to their source records. Database and literature availability depends on the enabled knowledge sources and the needs of the request. Always verify important scientific or clinical claims in the underlying source.

​
Visibility

Sessions are private unless they are shared. Projects and team settings determine who can access collaborative work. Check visibility before adding confidential or unpublished material.