Conclave Docs

For developers

The technical flows — the meeting spine, durable jobs, VFTE, permissions, deploy.

The same system from the engineering side. Each flow links to the reference page that covers it in depth. Green steps are trust / consent boundaries.

The meeting spine

The end-to-end path every meeting takes. Full detail: Data flow.

01
Capture
anonymous spans
→
02
Redis stream
transcription_segments
→
03
Conclave finalize
persist audio + transcript
→
04
DiariZen + VFTE
authoritative + identity
→
05
Enrich + KB
LLM at ingest

Durable jobs & the GPU worker

Finalize is crash-safe — the post-pass runs as a durable job an off-CVM GPU worker claims over :443. See Deployment and Capture features.

01
Finalize
enqueue
→
02
diarize_jobs stream
Redis
→
03
Worker claims
:443, no SSH
→
04
Heartbeat + result
audio by reference
→
05
Reconcile
XAUTOCLAIM on fail

VFTE identity

How identity attaches to anonymous spans — enroll, match, consent-gate, resolve at display. See VFTE features and the consent plane.

01
Enroll
audio + email + scope
→
02
identify-spans
vote-locked
→
03
Decision
MATCH / AMBIGUOUS / UNKNOWN
→
04
Consent-gated tag
propose / confirm
→
05
Resolve at display
name + visibility

Sharing & permissions

How a share resolves to exactly which artifacts a viewer receives. See Data model for the columns and gates.

01
ShareConfig
{ transcript, insights, audio }
→
02
can_see_* gates
per artifact
→
03
subject_consent edge
allow / anon / denied
→
04
Response
allowed artifacts only

Deploy

The in-place deploy to the Phala CVMs. Full recipe + gotchas: Deployment.

01
Build
--platform linux/amd64
→
02
Bump literal tag
dstack won’t re-pull vars
→
03
phala deploy
--cvm-id (never re-provision)
→
04
Warm diart
preload
→
05
Smoke test
attestation + record

For the product-side view of these journeys, see For people.

For people

The everyday journeys — record, name a speaker, own your voice, ask, share.

On this page

The meeting spineDurable jobs & the GPU workerVFTE identitySharing & permissionsDeploy