The Ontology Layer · Reference

Sources

The kinds of knowledge Chiron retrieves, and how each behaves.

A source is one kind of knowledge Chiron can retrieve. Each source is a retriever that implements the discover/search/fetch/activity contract and owns a prefix on every ref it emits.

A source declares two facets that tell a caller how to use it: its kind (how it is searched) and its availability (where it is served from).

text

kind:          "semantic"    vector / keyword relevance ranking
               "structured"  filterable records, best served chronologically
availability:  "indexed"     served from Chiron's own store (fast)
               "live"        fetched from the provider per call (best-effort)

The sources

Source	Kind	Availability	Providers	Time-bearing
`code`	semantic	indexed	—	no
`graph`	semantic	indexed	—	no
`docs`	semantic	indexed	—	no
`company_docs`	semantic	indexed	—	no
`work_items`	structured	indexed	github, azure_devops, gitlab	yes
`commits`	structured	indexed	github, azure_devops, gitlab	yes
`prs`	structured	indexed	github, azure_devops, gitlab	yes
`wiki`	semantic	indexed	github, azure_devops, gitlab	yes
`jira`	structured	indexed	jira	yes
`database`	structured	indexed	postgresql, mysql, mssql, mongodb	no

Time-bearing sources emit events on the activity timeline; relevance-only sources (code, graph, docs) return nothing there.

Restricting a query to sources

Every search and activity call accepts a sources filter. Omit it to query everything; pass a list to narrow:

json

{ "query": "rate limiting", "sources": ["code", "docs"], "top_k": 10 }

Which sources actually have data for *your* project is what discover reports. A source that returns null from discover has nothing for that project and is simply omitted.

Next: Providers.