# ChatGPT Compile — Sovereign Knowledge Base

> A compiled, versioned knowledge base of **2,489 ChatGPT conversations** (19,435
> turns) spanning 2023-03-12 → 2026-07-16, built locally with the Sovereign
> Knowledge Compiler. Everything here is derived from one person's own ChatGPT
> export — a personal corpus of work, projects, writing, and reasoning, organized
> so it can be queried, traversed, and **used directly as a reference by an LLM**.

## What this is

A static, backend-free knowledge resource. Every page is **pre-rendered to static
HTML at build time** (Next.js SSG) — an LLM agent or crawler can read the full
content of any page **without executing JavaScript**. Conversation pages include
the complete message text; theme and graph pages include full lists and static
node/edge tables; JSON-LD structured data is embedded on every page.

## Machine-readable resources

- [conversations.json](https://chatgpt-compile.vercel.app/conversations.json) — **full transcript corpus** (all conversations + turns, JSON)
- [dataset.json](https://chatgpt-compile.vercel.app/dataset.json) — aggregated view (themes, timeline, top conversations, graph)
- [asset_paths.json](https://chatgpt-compile.vercel.app/asset_paths.json) — referenced asset index
- [sitemap.xml](https://chatgpt-compile.vercel.app/sitemap.xml) — all public conversation URLs + section pages

## Pages (all pre-rendered, JS-free to read)

- [/](https://chatgpt-compile.vercel.app/) — dashboard: stats, 3D graph, themes, timeline
- [/conversations](https://chatgpt-compile.vercel.app/conversations) — index of every conversation (title, date, themes, preview)
- [/conversations/{uuid}](https://chatgpt-compile.vercel.app/conversations) — **full conversation thread** (all user/assistant turns as static text)
- [/themes](https://chatgpt-compile.vercel.app/themes) — every theme fully expanded with its linked conversations
- [/graph](https://chatgpt-compile.vercel.app/graph) — 3D graph + static top-nodes/edges tables
- [/models](https://chatgpt-compile.vercel.app/models) — model usage over time
- [/timeline](https://chatgpt-compile.vercel.app/timeline) — month-by-month counts table
- [/gallery](https://chatgpt-compile.vercel.app/gallery) — compiled images

## How an LLM should ingest this

1. Read `sitemap.xml` to get every conversation URL.
2. Fetch `/conversations/{uuid}` pages directly — the full thread is in the
   initial HTML (inside `<article class="conversation">`), so no rendering needed.
3. Use `dataset.json` for aggregate structure (themes, timeline, importance).
4. Each conversation page carries JSON-LD (`@type: Conversation`) for structured parsing.

## Note on sovereignty & privacy

Conversations are **public by default**. The owner may mark any conversation
private via `public/private-ids.json` or a per-conversation `"public": false`
flag; private conversations render metadata-only with a `noindex` robots tag and
are excluded from the sitemap. Personal/contextual content — treat as private
reference, not general fact.

## Related compiled-knowledge sites (same author, same compile-time-AI approach)

- [Kubernetes Knowledge Compiler](https://k8s-docs-compiler.vercel.app/llms.txt) — 8,198-node K8s knowledge graph
- [Sovereign Knowledge Compiler Explorer](https://skce-explorer.vercel.app/llms.txt) — Sovereign AI concept corpus

## Built with

This site follows the compile-time-AI pattern packaged as the [Discoverable Knowledge SDK](https://github.com/kliewerdaniel/discoverable-knowledge-sdk) — compile any dataset into a static, agent-discoverable knowledge site.