Curated context endpoints

The six endpoints under /v1/context/* are the project’s most carefully maintained AI-facing surface. They are handwritten narrative documents, edited under the same editorial pass as the reading site, designed for direct ingestion into an LLM system prompt with no transformation.

This page covers what each contains, when to include it, and how to compose them.

The six endpoints#

“Partial” means the endpoint touches that area as context for its main subject but isn’t the canonical document for it.

What each contains#

/v1/context/#

The root context endpoint. Returns a Context payload with:

• A short project description.
• The seven content sections of the reading site, with one-line descriptions.
• A sub_contexts array — pointers to the five narrative documents below.
• A links block to the reading site, the API manifest, and llms.txt.

Use this as a discovery endpoint. Hit it once to learn what’s available; rarely useful on its own in a system prompt.

/v1/context/hypothesis/#

The project’s central claim, stated as continuous prose. Covers:

• The five interconnected components of the working hypothesis (the Elohim are a literal plural; they are an advanced human civilization; religion is the long memory of their work; the work is organised along the precessional cycle; the current age is the age of disclosure).
• The status of the hypothesis as working — not creed, not proof.
• The interplay of source claim, comparative observation, and project synthesis.

Include this whenever you need the model to understand what the project is actually arguing. The single most load-bearing context document.

/v1/context/terminology/#

Canonical naming choices, with rationale and a do-not-use table.

Covers:

• Elohim — capitalize; plural; do not use “aliens” or “gods”.
• Yahweh — the Elohim president; do not use “God” unless quoting.
• Raëlism / Raëlian — with the diaeresis.
• Ancient astronaut theory — academic term; do not use “ancient aliens”.
• World Age / Precessional Age — distinct from “astrological age”.
• Great Year vs World Age — distinguish the two scales.
• A “do not use” table mapping common misnamings to the canonical term.

Include this in any application that surfaces the project to end users in natural language. Without it, models will reach for familiar but wrong terms (“the aliens”, “the gods”, “Raelism”).

/v1/context/timeline/#

The twelve precessional ages laid out as prose. Covers:

• A brief note on precession as an astronomical phenomenon.
• The corpus’s working dates (each age ~2,160 years, equal divisions).
• Each of the twelve ages with its dates and a one-paragraph summary of its primary content.
• A note on the corpus’s relationship to mainstream archaeology and chronology.

Include this whenever your application surfaces deep-history content. Without it, models conflate the corpus’s chronology with conventional ancient-astronaut chronologies (Sitchin’s Sumerian dates, Hancock’s Younger Dryas) or with astrological “ages”.

/v1/context/sources/#

The project’s source program — the six-tier authority model and the source-family taxonomy.

Covers:

• The six tiers: Tier 0 (Raëlian canon, primary), Tier 1 (ancient primary texts), Tier 2 (scholarly), Tier 3 (scientific/historical), Tier 4 (comparative traditions), Tier 5 (critical/skeptical).
• The source families: raelian, biblical, mesopotamian, vedic, mesoamerican, comparative, scholarly, scientific, critical, etc.
• The six-source rule — new wiki entries and articles aim for one source from each tier where applicable.
• The do-not-cite guardrails — what counts as a source and what doesn’t.

Include this in any application that exposes the project’s bibliographic backbone — search-with-citations, scholarly Q&A, research assistants. Without it, models will cite any plausible-looking source, including ones the project deliberately excludes.

/v1/context/method/#

The editorial method — how the corpus is written and what discipline it imposes.

Covers:

• Claim types (direct / framework / inferred / speculative) and their per-page labelling.
• The six-source rule in editorial terms.
• Editorial passes — the YYYY-MM date code system for tracking framing shifts (e.g. 2026-05, the modern human-civilization pass).
• Translation policy — derivation from English, divergence handling, glossary versioning.
• What the corpus is not — not a creed, not movement publication, not generic ancient-astronaut content.

Include this whenever your application asks the model to reason about the corpus’s claims, not just surface them. It encodes the discipline you want the model to inherit.

Composing them#

The five narrative endpoints are designed to be composed. Common patterns:

Minimal grounding (~2.5K tokens)#

Include only the hypothesis. The model knows what the project argues, nothing more.

ctx = httpx.get(
    "https://api.wheelofheaven.world/v1/context/hypothesis/"
).json()["data"]["body"]

Suitable for a chatbot that should be able to describe the project but not reason within it.

Standard grounding (~5K tokens)#

Include hypothesis + terminology + method.

keys = ["hypothesis", "terminology", "method"]
ctx_blocks = [
    httpx.get(f"https://api.wheelofheaven.world/v1/context/{k}/")
         .json()["data"]["body"]
    for k in keys
]
ctx = "\n\n---\n\n".join(ctx_blocks)

Suitable for a general assistant. The model can describe the project, use canonical terminology, and surface claim labels.

Full grounding (~7.5K tokens)#

All five narrative endpoints.

keys = ["hypothesis", "terminology", "timeline", "sources", "method"]
ctx_blocks = [
    httpx.get(f"https://api.wheelofheaven.world/v1/context/{k}/")
         .json()["data"]["body"]
    for k in keys
]
ctx = "\n\n---\n\n".join(ctx_blocks)

Suitable for a deep-history assistant, a research tool, or anywhere the model needs to reason about the corpus’s chronology and bibliographic discipline.

Topic-tuned grounding#

Pick the documents that match your application’s surface area.

How fresh are these?#

The context endpoints are served with a 24-hour edge TTL, the longest TTL in the API. Cloudflare purges on every deploy, so the upper bound on staleness is 24 hours or the time since last deploy, whichever is shorter.

For agents that boot once a day, a fresh fetch at boot is sufficient. For long-running agents, refetching every 24 hours is sufficient.

What these endpoints don’t contain#

• Per-page bodies. The narrative documents summarise; they don’t reproduce wiki entries. For specific entries, hit /v1/wiki/{slug}/.
• Live counts. No “we currently have 111 wiki entries”. These are editorially curated documents, not dashboards.
• Bibliographic records. The sources/ document explains the program; for the records themselves, hit /v1/sources/.
• News dispatches. Time-anchored content lives at /v1/news/. The context endpoints are evergreen.

How these endpoints are generated#

The five narrative documents are handwritten Markdown files in api.wheelofheaven.world/content/v1/context/. They are rendered through a Zola template into the JSON envelope at build time. The Markdown source is the editorial artefact; the JSON response is the machine-readable wrapper.

When the project’s editorial pass shifts, these documents are rewritten in the same pass. They are not derived from other content — they are first-class editorial output.