Skip to content

v0.98.22

Choose a tag to compare

@github-actions github-actions released this 09 Jul 16:25
· 862 commits to main since this release

Changed

  • Dashboard front door is the Studio command centre: GET / and
    /studio/command serve the same hub-independent shell; the pre-Studio hub
    HTML moves to /classic. CLI startup prints Studio URLs first. Quick start
    gains a multi-seat golden path ending in Studio.

Added

  • SQLCipher live event store: optional synapse-channel[sqlcipher] extra
    (sqlcipher3-binary==0.6.0), EventStore(..., key_file=) / key=,
    synapse hub --db-key-file, and offline
    synapse encrypt-key migrate-sqlcipher for plaintext→encrypted migration.
    Page-level encryption for main DB + WAL; stock install stays dependency-free.
  • Studio observed_fleet section on /studio.json: multi-hub / FLEET advisory
    peer health (reachable, lagging, unreachable) from dashboard
    observed_peers. Command centre shows a peers headline and panel.
  • synapse doctor multi-seat trust checklist: --multi-seat,
    --identity-trust, and --role-grants feed check_multi_seat_posture
    (token + trust + role materials → --team-secure remedy). Auto-detects a
    multi-seat roster. New deaf-agents check warns when live agents lack a
    matching -rx waiter.
  • ACL permission verbs mailbox and role-claim: a policy can grant mailbox
    backlog replay for a non-self/non--rx monitor (target kind agent) and
    authorise heartbeat role bindings alongside the role-grant store (target kind
    role). Wired into _mailbox_recipient and permitted_role_claims. See
    docs/identity-and-acl.md.
  • synapse hub --team-secure multi-seat trust profile: requires a connect
    token, --identity-trust (forces identity binding), --role-grants (forces
    role-claim enforcement), and private directed messages. Prints recommended
    next hardening (message-auth, ACL, TLS/--paranoid, durable --db) without
    blocking startup. Composes with --paranoid for multi-seat + exposed binds.
    See docs/team-secure.md.
  • Universal receipt projections now expose release/claim evidence, delivery
    receipts, sandbox run attestations, approval/policy/verification notes,
    operator relays, cross-hub pointers, A2A validation notes, and postmortem
    notes in one read-side shape. synapse dashboard --feeds-db serves the
    projection at /receipts.json, the cockpit prefers that feed when present,
    and synapse event-query <db> "universal-receipts all" renders the same
    first-class receipt objects without changing the legacy delivery-receipt
    query.
  • The public API freeze now pins the exact synapse_channel.__all__ export
    list, and the docs include a 0.x to 1.0 migration guide covering upgrade
    order, wire-version checks, stable-surface guards, and release-cut checks.
  • The Studio /studio.json projection now includes a security-posture section
    covering sandbox grants, ACL/role visibility, dashboard exposure evidence,
    signed federation observation, and receipt evidence. /studio/command renders
    the same rows as a read-only posture panel beside the Coordination Clock.
  • The Studio command centre now has a persistent NavRail, a HeaderBar with the
    live hub id, version, verdict, and connection state, and a LiveFeed panel that
    tails /events.json when the dashboard is started with --feeds-db.
  • Declarative workflows now support step-level requires evidence predicates
    for proof-carrying steps. workflow plan and workflow run accept an
    --evidence snapshot and hold a step until its required receipt, test,
    policy, approval, sandbox, mailbox, dead-letter, or claim evidence matches.
  • synapse a2a-conformance now prints the local A2A 1.0.0 support matrix as
    Markdown or JSON, including supported, partial, unsupported, and externally
    gated rows for the bridge.
  • A2A push delivery now has an injectable WebhookDeliveryClient for local
    validation harnesses, plus real HTTPS receiver and 307 reverse-proxy redirect
    tests that keep the production default fail-closed against local/private
    webhook targets, including delivery-time DNS rebinding to a local receiver.
  • The A2A deployment threat model now records the exposed-bridge posture for
    bearer auth, TLS/proxy placement, state-file handling, webhook egress,
    logging, and receipt evidence.