reference · generated from oto --help at v0.6.1

The command line

Every command as the engine describes it. --project defaults to the current directory wherever it appears. Commands run through uvx --from "oto-kg @ git+https://github.com/Cynergis/oto" oto … without an install, or as oto … once installed.

I want to…

I want to	Command	Which does
start a project	`oto init`	from an ontology, a pack, or empty; laid out for GitHub with --repo
know where I am	`oto status`	the project's state and the one next command
bring documents in	`oto ingest`	claim, extract, block personal data; complete when the graph holds the facts
record what someone said	`oto capture`	a dated source document in the speaker's words
see what the documents talk about	`oto survey`	headings, dates, recurring terms; which already resolve
decide or check the vocabulary	`oto ontology`	check, accept, rationale, widen, import; the catalog
propose facts and pass the gates	`oto curate`	start, add, check, diff, apply, undo; the ledger
let a model draft a proposal	`oto draft`	one document, to the same rules; needs an API key
derive facts from patterns	`oto rules`	check, diff, accept, explain
compile	`oto build`	every layer; Neo4j and a static site as targets
ask	`oto query`	entity, neighbors, search, history, as-of, stale, explain
give it to people and agents	`oto serve`	the explorer over HTTP; the tools over stdio for MCP
see what is on its way in	`oto preview`	the graph with every proposal and the candidate
audit citations	`oto vet`	facts citing documents the corpus no longer holds
describe what can be done	`oto actions`	list, show bound to an entity, record the response
share a vocabulary or a pack	`oto pack`	new, check, refresh, add, publish; and ontology export
use a registry	`oto registry`	add, list, refresh, check; the catalog site
ship to readers	`oto publish`	a read-only query repository; sync on their machine
measure the answers	`oto bench`	a gold set with provenance; ablations and a baseline
turn a complaint into a test	`oto feedback`	record with a name; promote into the gold set
check the install	`oto verify`	the greenfield self-test
start over	`oto clean`	delete everything generated

Every command

Command	What it does
`oto init`	create a new project (data only, no engine code)
`oto status`	where the project is in its life, and the next step
`oto ingest`	claim the inbox into a run and extract it into the corpus
`oto figures`	report figures that have no description
`oto survey`	map the corpus: documents, headings, recurring terms
`oto capture`	write what someone said into the inbox as a dated source document
`oto ontology`	the project's vocabulary, and the catalogue of ontologies to start from
`oto pack`	the full extension a person installs into Claude Code: an ontology, its skills and its views
`oto registry`	the registries this machine fetches ontologies from
`oto rules`	validate, dry-run, diff and accept the rules over the graph
`oto actions`	list and check the actions the project declares (Oto never invokes one)
`oto curate`	propose, review and apply changes to the curated graph
`oto draft`	a model drafts one document's proposal (needs the draft extra)
`oto vet`	audit the graph against the documents the corpus holds
`oto build`	compile every layer
`oto clean`	delete everything the build generated
`oto verify`	greenfield self-test: init, author, build, assert
`oto preview`	build the preview store: the live graph plus the candidate and every
`oto query`	ask the knowledge base a question
`oto serve`	run the query server over stdio (JSON-RPC 2.0), or over HTTP
`oto publish`	push the built store to a read-only query repository
`oto sync`	clone or update a published query store on this machine
`oto bench`	measure answer quality against a gold question set
`oto feedback`	record what a reader said about an answer, and make it a test

oto --help

usage: oto [-h]
           {init,status,ingest,figures,survey,capture,ontology,pack,registry,rules,actions,curate,draft,vet,build,clean,verify,preview,query,serve,publish,sync,bench,feedback,version}
           ...

Oto

positional arguments:
  {init,status,ingest,figures,survey,capture,ontology,pack,registry,rules,actions,curate,draft,vet,build,clean,verify,preview,query,serve,publish,sync,bench,feedback,version}
    init                create a new project (data only, no engine code)
    status              where the project is in its life, and the next step
    ingest              claim the inbox into a run and extract it into the corpus
    figures             report figures that have no description
    survey              map the corpus: documents, headings, recurring terms
    capture             write what someone said into the inbox as a dated source document
    ontology            the project's vocabulary, and the catalogue of ontologies to start from
    pack                the full extension a person installs into Claude Code: an ontology, its
                        skills and its views
    registry            the registries this machine fetches ontologies from
    rules               validate, dry-run, diff and accept the rules over the graph
    actions             list and check the actions the project declares (Oto never invokes one)
    curate              propose, review and apply changes to the curated graph
    draft               a model drafts one document's proposal (needs the draft extra)
    vet                 audit the graph against the documents the corpus holds
    build               compile every layer
    clean               delete everything the build generated
    verify              greenfield self-test: init, author, build, assert
    preview             build the preview store: the live graph plus the candidate and every
                        proposal
    query               ask the knowledge base a question
    serve               run the query server over stdio (JSON-RPC 2.0), or over HTTP
    publish             push the built store to a read-only query repository
    sync                clone or update a published query store on this machine
    bench               measure answer quality against a gold question set
    feedback            record what a reader said about an answer, and make it a test
    version             print the version

options:
  -h, --help            show this help message and exit