Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/bcanata/maieutic/llms.txt

Use this file to discover all available pages before exploring further.

Maieutic ships with a demo script that seeds the database with realistic student sessions, exercises, and cohort data so you can explore the instructor dashboard without running a real class. The fixtures were produced by running actual Opus sessions and saving the results to JSON — they are not synthesized.

Seeding demo data

npm run reset-demo
npm run reset-demo destroys all existing data and replaces it with the demo fixtures. Do not run it on a database that contains real student sessions.

What the script seeds

npm run reset-demo calls scripts/replay-fixtures.ts --wipe, which does the following:
  1. Drops all existing data — sessions, events, and exercises are deleted before anything is inserted.
  2. Seeds published exercises — all exercise fixtures from tests/fixtures/exercises/ are loaded. The fixture set includes exercises across all four units and all three student levels, including vowels-demo, fibonacci-demo, password-demo, and many more.
  3. Creates demo student identities — Ana (ana_student), Beto (beto_student), and Carmen (carmen_student), plus four anonymous cohort students (cohort_student_0 through cohort_student_3).
  4. Seeds sessions with realistic phase data — each session’s phase1Data, phase2Data, and phase3Data fields contain real Opus exchanges, spec iterations, divergences, and student responses captured from live API calls.
  5. Seeds live summaries and session events — the liveSummaries array and the SessionEvent audit log are populated so the dashboard has data to display.
  6. Rebases all timestamps — start times, completion times, and per-event timestamps are rewritten relative to the current clock so the live dashboard appears active.
After seeding, the demo state looks like this:
StudentExerciseStatus
AnaCount vowelsCompleted — closed session with one case-sensitivity divergence
BetoValidate a passwordCompleted — closed session, explained all divergences correctly
CarmenNth Fibonacci numberActive — stuck in phase 1, has an open help request
Cohort students 0–3Count vowelsCompleted — varied case-drift patterns for cohort analytics

When to use it

  • After initial setup — explore the instructor dashboard, session reasoning view, and cohort analytics without needing real students.
  • To reset to a known-good state — if you have been testing and want to restore the demo, run the script again.
  • Before a demo or presentation — the rebased timestamps make the dashboard look live each time you run it.

The capture-fixtures script

The JSON fixtures that npm run reset-demo replays were originally produced by npm run capture-fixtures. You do not need to run this script during normal setup.
npm run capture-fixtures
This script runs the complete Maieutic session loop — spec examination, coding phase with Opus chat, intent-diff, and post-hoc classification — for each demo student against the live Anthropic API. It then saves the results to tests/fixtures/exercises/ and tests/fixtures/sessions/. Only run capture-fixtures when Opus prompt templates change and you need fresh fixtures that reflect the updated prompts. The script takes around ten minutes and requires a valid ANTHROPIC_API_KEY.
npm run capture-fixtures makes real API calls to Anthropic and will consume a significant number of credits. Only run it when you are deliberately updating the test fixtures.

Test data coverage

The demo fixtures are designed so every instructor view has populated data:
  • Live dashboard — Carmen’s active session shows a live summary and an open help request flag; Ana and Beto appear as recently completed.
  • Session reasoning view — all three named students have full phase data, including Opus exchanges and divergence responses, so the two-column reasoning view is fully populated.
  • Cohort analytics — the four cohort students all worked the same vowels exercise with systematic case-drift, giving the cohort narrative something concrete to analyze.
  • Exercise library — all three exercises are published and have aggregate session data attached.

Build docs developers (and LLMs) love