Skip to content

reports.md

Latest commit

 

History

History
85 lines (59 loc) · 3.91 KB

reports.md

File metadata and controls

85 lines (59 loc) · 3.91 KB

Reports

Reports let you get structured, aggregate answers from many personas (a cohort or an explicit list) for a prompt. Use them to test ideas, positioning, or questions at scale.

Generate a report

mass report generate -p "<prompt>" --cohort <cohort-id> -t <type> [-o report.html]

Or use a list of personas instead of a cohort:

mass report generate -p "<prompt>" --personas <id1>,<id2>,... -t <type> [-o report.html]

Options:

Option Description
-p, --prompt <text> The report prompt (e.g. "What do you think about product Y?"). Defaults to "What do you think?" if omitted.
--cohort <id> Cohort ID. All personas in the cohort are queried.
--personas <ids> Comma-separated persona IDs. Use instead of --cohort to target specific personas.
-t, --type <type> Report type: feedback, debate, questionnaire, or ideas. Default: feedback.
--debate-rounds <n> Debate only: max persona speaking turns. Default: 20.
--debate-duration <minutes> Debate only: stop after N minutes; 0 = no time limit. Default: 2.
-o, --out <path> Write report HTML to this path.
-w, --workspace <id> Attach the report to an existing workspace. If omitted, a new workspace is created for this run.

Example:

mass report generate -p "Views on product Y" --cohort abc-123 -t feedback -o report.html

Report types

  • feedback – Sentiment, quotes, summary, and verdict (positive / neutral / negative). Good for "what do people think?" style questions.
  • debate – For/against views, which personas take which side, and a verdict. Good for polarising or opinion-driven topics.
  • questionnaire – Questions are extracted from the prompt; each persona answers them; results are aggregated. Good for structured surveys.
  • ideas – Idea generation and aggregation. Good for brainstorming or "what would you suggest?" prompts.

Scenario examples by type

  • feedback: One clear question plus what would make them try or avoid. Example: mass report generate -p "Would you use an app that helps families plan weekly meals? What would make you try it or put you off?" --cohort <id> -t feedback -o report.html. See Examples: product feedback.
  • debate: A single proposition with a clear for/against. Example: mass report generate -p "Should remote work be the default for office jobs?" --cohort <id> -t debate -o report.html. See Examples: debate report.
  • questionnaire: Numbered or bulleted questions in the prompt so the LLM can extract and aggregate. Example: mass report generate -p "Please answer: (1) How often do you shop online? (2) What would make you switch?" --cohort <id> -t questionnaire -o report.html. See Examples: questionnaire report.
  • ideas: Open-ended request for suggestions. Example: mass report generate -p "What features would make you use a meal-planning app? List 3-5 ideas." --cohort <id> -t ideas -o report.html. See Examples: ideas report.

List, show, delete

  • List reports:

    mass report list

    Filter by workspace:

    mass report list -w <workspace-id>

  • Show a report and optionally write HTML/JSON:

    mass report show <report-id> [-o report.html]

  • Delete a report:

    mass report delete <report-id>

Output and storage

  • Report JSON is stored under the data root in reports/<report-id>.json.
  • With -o <path> on generate or show, HTML is written to that path. You can open it in a browser to read the report.

Next steps

  • Examples: in-depth walkthroughs (feedback, debate, questionnaire, ideas).
  • Workspaces: reports can be attached to a workspace via -w.
  • Testing personas: using reports to test ideas at scale.