Hands-on, organized by use case. Each folder is one case — a short walkthrough plus the real artifact OpenKB produced for it (generated, not hand-written).
Test inputs live in
docs/— the sample PDFs the cases ingest. They're kept out of git on purpose (large / third-party files); drop your own documents there to follow along.
| Folder | What it shows | Real artifact inside |
|---|---|---|
configuration/ |
init, config.yaml, API keys, LiteLLM tuning (Ollama, Copilot) |
— |
commands/ |
the everyday loop: add · query · remove · recompile · lint · list · status |
a compiled sample-wiki/ |
pageindex-cloud/ |
long documents: local vs. cloud indexing, and importing cloud-indexed docs | — |
chat/ |
the interactive REPL: persistent sessions + slash commands | — |
skills/ |
distill a redistributable agent skill from your wiki | transformer-attention/SKILL.md |
slides/ |
generate a single-file HTML slide deck | attention-intro.html |
visualize/ |
render the wiki as an interactive knowledge graph | graph.html |
The compiled sample-wiki/, the skill, the deck, and the graph were all
generated by running OpenKB over a single document —
docs/attention-is-all-you-need.pdf — with
gpt-5.4:
mkdir my-kb && cd my-kb
openkb init --model gpt-5.4 --language en
openkb add /path/to/examples/docs/attention-is-all-you-need.pdf # → commands/sample-wiki/
openkb query "How does attention replace recurrence?" --save # → commands/sample-wiki/explorations/
openkb skill new transformer-attention "Reason about the Transformer…" # → skills/
openkb deck new attention-intro "A short intro deck on the Transformer…" # → slides/
openkb visualize # → visualize/Your generated text will differ run to run (LLM output isn't deterministic) — the structure is what's stable.
New to OpenKB? Read configuration/ first, then
commands/.