apps for your life · sequential AI pipeline

A year book that can
show its work.

Omyear turns confirmed life context and deterministic symbolic calculations into a warm personal book—without letting generated prose outrun its sources.

Open Maya’s demo ↗Trace the sequence ↓

Demo generated with gpt-5.6-sol

source records72
sources cited53
source-link errors0
year chapters12

one run · seven proofs

From a birthday to a book,
in an inspectable line.

  1. 01

    INTAKE

    A person, not a prompt

    The questionnaire records confirmed facts, goals, birth data, return location and a gift message in a strict input schema.

    input.json
  2. 02

    CALCULATE

    Numbers before prose

    Python computes numerology, the 22-arcana matrix, natal and solar-return charts, Human Design activations and nearest astrocartography lines.

    calc.json
  3. 03

    REGISTER

    Every fact gets an address

    Inputs and calculations become immutable source records such as calc.period.4 or input.goal.rest. The model can cite only these IDs.

    72 source records
  4. 04

    WRITE · GPT-5.6

    Creativity inside a boundary

    GPT-5.6 receives the source registry, editorial constraints and a JSON Schema. Every interpretive object must return its exact sourceIds.

    draft.json
  5. 05

    VERIFY

    Trust, then check

    A deterministic verifier rejects unknown citations and missing sections, checks all 12 periods and practices, and flags certainty or sensitive language.

    verification.json
  6. 06

    REVIEW

    A human keeps the last word

    Cross-framework synthesis and low-confidence passages stay visible. The pipeline cannot label a draft ready for delivery without editorial approval.

    1 review flags
  7. 07

    PUBLISH

    One dataset, one living book

    The approved draft compiles into the existing Omyear book format and Astro builds the private, responsive reading experience.

    /0811

the editorial boundary

Four kinds of truth.
Never blurred together.

The interface labels what came from the person, what code calculated, what GPT‑5.6 interpreted, and what still needs a human decision.

01CONFIRMED

questionnaire facts and goals

02CALCULATED

repeatable numeric outputs

03INTERPRETED

source-linked model prose

04APPROVED

human editorial judgment

run the same path

One command. Every artifact kept.

OPENAI_API_KEY=… node pipeline/run.mjs \
  --input pipeline/examples/maya.json \
  --provider openai --model gpt-5.6-sol

The run directory preserves the input, calculations, exact prompt, structured draft, verification report, model metadata and compiled book.

Inspect the live source ledger ↗