Cerid AI — Knowledge Packs

Curator-built tarballs for the Cerid AI personal-knowledge-companion harness. Each pack is a .tar.gz archive containing a pack.json manifest plus a content/ tree of clean markdown, sourced from permissively-licensed third-party publishers and processed through Cerid's seven source adapters.

Installing a pack

The Cerid harness installs packs by id; users browse them via the Library UI, MCP tool palette, REST endpoint, or CLI:

docker exec ai-companion-mcp \
  python -m scripts.install_knowledge_pack list

docker exec ai-companion-mcp \
  python -m scripts.install_knowledge_pack install rust-book

The shipped Cerid registry already points at this repo's GitHub releases, so no extra configuration is needed for end users.

Pack catalogue

Each release carries the latest tarballs for the catalog. See the Latest release for the current sha256 + file list. Catalog metadata (license, source, adapter) lives in config/knowledge_packs.json upstream.

Provenance + licensing

Pack ID	Upstream	License	Adapter
mdn-web-docs	mdn/content	CC-BY-SA-2.5	github_zip
rust-book	rust-lang/book	MIT	github_zip
typescript-handbook	microsoft/TypeScript-Website	MIT	github_zip
python-stdlib-docs	docs.python.org	PSF-2.0	python_docs_zip
kubernetes-website	kubernetes/website	CC-BY-4.0	github_zip
helm-docs	helm/helm-www	MIT	github_zip
tldr-pages	tldr-pages/tldr	CC-BY-4.0	github_zip
apache-spark-docs	apache/spark	Apache-2.0	github_zip
learnxinyminutes	adambard/learnxinyminutes-docs	CC-BY-SA-3.0	github_zip
bogleheads-wiki	bogleheads.org/wiki	CC-BY-SA-4.0	mediawiki_api
irs-publications-curated	irs.gov/publications	US-gov-PD (CC0-1.0)	html_scrape
cfpb-ask	consumerfinance.gov/ask-cfpb	US-gov-PD (CC0-1.0)	html_scrape
18f-methods-guides	18F/methods	CC0-1.0	github_zip
chaoss-metrics	chaoss/metrics	MIT	github_zip
wikivoyage-en	dumps.wikimedia.org/enwikivoyage	CC-BY-SA-3.0	wiki_dump
gutenberg-classics-curated	gutenberg.org	Public Domain (CC0-1.0)	gutenberg
wikipedia-simple-en	wikimedia/wikipedia (HF)	CC-BY-SA-3.0	hf_dataset
cosmopedia-khanacademy	HuggingFaceTB/cosmopedia (HF)	Apache-2.0	hf_dataset
pes2o-cs-recent	allenai/peS2o (HF)	ODC-BY-1.0	hf_dataset

Cerid AI is the curator (build pipeline + tarball envelope). The content in each pack remains under its upstream license, and attribution / share-alike obligations propagate to embeddings + RAG output. The Cerid install path enforces a SPDX license-category gate that requires explicit operator opt-in for share_alike content.

Reproducibility

Tarballs are reproducible from the upstream sources via:

docker exec ai-companion-mcp \
  python -m scripts.build_catalog --all

Every release pins:

• the upstream commit/snapshot via the recipe in cerid-ai's config/knowledge_packs.json
• a sha256 of the tarball, verified at install time

License (this repo)

The build recipes and orchestration code live in Cerid-AI/cerid-ai under Apache-2.0. The tarball contents in this repo are licensed per the table above — see each pack's pack.json for the canonical SPDX identifier.