Kernel is a small HTTP API for reading and writing your projects, ACDs, the auto-curated corpus, and sources. JSON in, JSON out. Reads and writes through this API are free — only generations performed inside Kernel consume credits.
projects — the top-level container. Holds sources, ACDs, and one auto-curated corpus document.
sources — raw inputs (files or text) you push in. Classified, summarized, and folded into the corpus automatically.
ACDs — AI Context Documents. Versioned markdown documents addressed by slug. Generated inside Kernel, or pushed in from your own pipelines.
corpus — the project's single auto-curated reference document, derived from your sources.
All endpoints are under https://api.tryaware.ai/functions/v1/kernel-public-api.
Projects, ACDs, and the corpus accept either a UUID or a slug as path identifiers — pick whichever is more convenient for your caller.
Versioned reads (ACDs, corpus) return the latest version by default. Pass ?version=N to pin a specific version.
Request and response bodies are JSON unless a file upload is noted.
ACD push (markdown body or uploaded file): 25 MB cap.
Source upload (file or text): 25 MB cap.
Audio and video source files are rejected — transcribe inside Kernel instead.
Rate limiting: not enforced in v1 — please be reasonable.