Skip to content

Gen3 Workflows Diagram Review

github-actions[bot] edited this page May 29, 2026 · 1 revision

Diagram Review Workflow

Start Here

  1. Create a dataset for diagram source files (Datasets)—PNG, PDF pages, Visio exports, or other supported diagram formats.
  2. Upload diagrams to the dataset and wait for processing to complete (Uploading and Importing Documents).
  3. Open GT Chat and choose an agent with vision or diagram-review instructions (or another agent tuned for spatial and labeling questions).
  4. Use the paperclip to attach the dataset containing your diagrams.
  5. Ask specific questions: component identification, data-flow checks, legend validation, or compliance against a written standard.

Paperclip in chat for dataset attachment during diagram review

Why this matters

Diagrams are not plain text—effective review requires vision-capable models and agents tuned for spatial and labeling questions. Keeping diagrams in a dataset lets you reuse the same corpus across multiple review sessions and share it with collaborators through Groups.

Details

Dataset and vision settings

  • When creating datasets that rely on image understanding, confirm vision or multimodal settings in Datasets and agent configuration (Building Agents).
  • Large blueprints may need to be split into readable pages or tiles before upload.

Chat practices

  • Reference diagram names or page numbers in prompts.
  • Ask the agent to list assumptions when labels are ambiguous.
  • For side-by-side comparisons, attach datasets that contain both versions and ask for a structured diff.

Collaboration

Share the dataset or agent with a Group when multiple reviewers need the same diagram corpus. Use Conversations to export review threads for records.

External automation

For bulk diagram ingestion from tools, use GT API upload keys and chat aliases when GT API is enabled for your account, instead of manual paperclip attachment.

Related workflows

Clone this wiki locally