Public
Like
1
expected-work
Val Town is a collaborative website to build and scale JavaScript apps.
Deploy APIs, crons, & store data – all from the browser, and deployed in milliseconds.
- Full RESTful API with 8 endpoints
- New calibration analysis with 10% confidence buckets (centered: 0-10%, 10-20%, etc.)
- Enhanced performance analysis including calibration data
- OpenAPI 3.0 specification compatible with Custom GPTs
- Live API endpoint:
https://dcm31--019899c30691716293d39932f5c6be80.web.val.run/api/
openapi.yaml- OpenAI Custom GPT compatible spec (no oneOf/anyOf/allOf)CUSTOM_GPT_INSTRUCTIONS.md- Complete GPT behavior instructionsCUSTOM_GPT_SETUP_GUIDE.md- Step-by-step setup guide
POST /quick-goal- Calculate optimal goal + create Fatebook predictionPOST /fatebook-goals/{id}/resolve- Mark goals complete/incompletePOST /fatebook-goals/{id}/forecast- Update confidence during dayGET /calibration-analysis- View prediction accuracy by confidence buckets
- Morning: Ask GPT to set optimal goal based on probability assessments
- During day: Optionally update confidence levels
- Evening: Report completion status to resolve prediction
- Weekly: Review calibration analysis for improvement insights
- Go to https://chatgpt.com/create
- Use instructions from
CUSTOM_GPT_INSTRUCTIONS.md - Import schema from:
https://dcm31--019899c30691716293d39932f5c6be80.web.val.run/api/openapi.yaml - No authentication needed (API key passed in requests)
- Get Fatebook API key from https://fatebook.io/api-setup
- Start conversation: "Help me set today's work goal"
- Provide API key when prompted
- Give probability assessments for different goal amounts
- 10 confidence buckets: 0-10%, 10-20%, 20-30%, ..., 90-100%
- Centered around increments: Expected rates use bucket center points (5%, 15%, 25%, etc.)
- Only first forecasts: Clean analysis using initial EVGS predictions
- Calibration error tracking: Shows over/under-confidence patterns
{ "buckets": [ { "bucket": "70-80%", "centerPoint": 0.75, "predictions": 4, "actualSuccessRate": 0.75, "expectedSuccessRate": 0.75, "calibrationError": 0 } ], "overallCalibrationScore": 0.14, "brierScore": 0.21, "totalPredictions": 37, "perfectCalibration": false }
All endpoints tested with your actual Fatebook API key:
- ✅ Goal calculation and creation working
- ✅ Forecast updates working
- ✅ Goal resolution working
- ✅ Calibration analysis with corrected buckets working
- ✅ 37 historical predictions analyzed
All files are in the api-integration branch:
/openapi.yaml- Custom GPT compatible OpenAPI spec/CUSTOM_GPT_INSTRUCTIONS.md- GPT behavior instructions/CUSTOM_GPT_SETUP_GUIDE.md- Setup walkthrough/utils/api.ts- Enhanced API service with calibration analysis/index.ts- Main server with all endpoints
Your EVGS system now has a complete Custom GPT integration that will:
- Optimize daily goals using expected value theory
- Track predictions automatically in Fatebook
- Analyze calibration across confidence levels
- Provide personalized recommendations for improvement
The Custom GPT will act as your daily goal-setting coach, helping you develop better metacognitive skills about your own capabilities and uncertainty quantification.
Next step: Set up the Custom GPT using the provided files and start your daily goal optimization workflow!