A. Find events matching an interest (from the open web)
Momentra searches the organizations in your scope. To find events for an interest you haven’t indexed yet, the assistant discovers the relevant sources on the open web itself, then indexes them — there is no Momentra-side web-search tool, and none is needed.1
Host searches the web
The AI host searches the web for venues/organizations/aggregators relevant to
the interest (e.g. “family science events near Burlington VT”) and resolves
candidate domains.
2
Index the domains
index_org (batch websites) — index those domains. Poll org_status (or
the batch items) until ready.3
Surface events
search (free text) or, once you have a brief, matching_events — surface
the events.index_org →
search/match. The host brings the discovery; Momentra brings the extraction,
normalization, and matching.
B. Create a brief manually
1
Write the brief
update_brief with a brief document and status: "active" (see Brief
shape). clientName optionally labels it.2
Confirm it's stored
get_brief to confirm it’s stored and active.3
See what it surfaces
matching_events / matching_orgs to see what it surfaces; evaluate_event
to test a specific item; accept_event / reject_event to refine.C. Seed a brief from a trusted aggregator
An aggregator is a curated source you already trust — most of its events link off-domain to the real hosts (e.g. a “things to do this weekend” site). Momentra can reverse-engineer a brief from it.1
Index the aggregator
index_org the aggregator’s website → org_status until ready (its events
must be indexed and enriched).2
Seed the brief
seed_brief with the aggregator’s orgId (or website). It confirms the
org is an aggregator, then learns a brief that fits ~all of its events and
sets it as your active brief. Returns a jobId.3
Poll to ready
seed_status until state: "ready" — reports the fit percent and the
learned briefId. If the org isn’t an aggregator it returns state: "not_aggregator" and does no work.4
Review and tighten
seed_review returns borderline events (near your brief but not clearly on
it). Reject the off-brief ones with seed_reject, which records them as
negatives and returns a proposed brief tightening + the questions behind it —
apply the ones you agree with via update_brief. The proposal is guarded so
it never starts blocking the aggregator’s own events.Timing. Facet classification and vector embedding run asynchronously
after an
index_org crawl completes. A just-indexed org’s events are
searchable immediately via search/search_business_events, but
matching_events/matching_orgs/seed_brief (which rely on facets +
embeddings) may see nothing for that org until enrichment finishes. Retry after
a short delay.