MaveraMavera Docs
Checking…
Use or share this pageSend the page itself, print it, or open its complete context in your AI workspace.

Personas that remember

Persistent personas, hallway tests, and budget-planned research.

1 min readDocs build Updated Jul 26, 2026

Persistent personas

A persistent persona is a named synthetic person who remembers previous conversations. Interview "Sarah" today, come back next quarter, and she recalls what she told you — same identity, consistent history.

First meeting (creates the persona)
curl -X POST "$BASE/api/v1/personas/persistent" -H "x-api-key: $MAVERA_API_KEY" \
  -H "content-type: application/json" -d '{
  "personaKey": "sarah-weeknight-cooks",
  "audienceId": "aud_...",
  "topic": "how she decides what is for dinner",
  "openingQuestion": "Walk me through last Tuesday night."
}'

Later interviews use the same personaKey (no audienceId needed): the stored profile comes back with a memory digest of past sessions injected as continuity — she'll reference earlier answers naturally and stay consistent with them.

  • GET /api/v1/personas/persistent — your roster: every persona, interview count, memory digest
  • Memories are code-side summaries (capped at 12, oldest dropped) — no model invents her history
  • Identity stability: the same key always draws the same person

Hallway test — the 30-second gut check

One question past n=12 for a handful of credits. Directional only — the output says so, counts positive/negative/mixed leanings in code, and includes the one-click upgrade to a full n=50 study.

POST /api/v1/tools/hallway_test
curl -X POST "$BASE/api/v1/tools/hallway_test" -H "x-api-key: $MAVERA_API_KEY" \
  -H "content-type: application/json" -d '{
  "input": {
    "audienceId": "aud_...",
    "question": "Would you look twice at a collar that promises GPS with no subscription?",
    "stimulus": "Waggle Collar: $79, no monthly fee"
  }
}'

Research plan for a budget

Invert budgeting: give Mave a goal and a credit budget, get a sequenced study mix with the math shown.

POST /api/v1/tools/plan_research_budget
curl -X POST "$BASE/api/v1/tools/plan_research_budget" -H "x-api-key: $MAVERA_API_KEY" \
  -H "content-type: application/json" -d '{
  "input": { "goal": "launch positioning for the new collar", "budgetCredits": 2000 }
}'

The mix is model-proposed; every cost number comes from the measured price table in code (honesty.costsAre: "measured-price-table"). Over-budget mixes are trimmed, never silently exceeded. Both are also available as one-click cards on the Workflows page.

On this page