Spaces:
Running
Running
Commit History
-
GridStack: harvest Next.js route capabilities Running
k0880kUnverified signatureIndex Next.js App Router exported HTTP handlers as capability cells and cover path inference with regression tests.ae5ef37
-
test(retina): prove genome convergence resolution gaps and verification
k0880k7bd6597
-
Update PostMeet recovery contract for Meeting Lens v5
k0880k06b340f -
Test full Meeting Lens hydration instead of vlog-card hydration
k0880kc90ee3d
-
test(postmeet): cover YouTube source and Vlog template hydration
k0880k7c8ed40 -
test(migration): assert explicit participant inventory
k0880ka31ad8a -
test(migration): lock builder ownership and explicit inventory
k0880k2c59669 -
test(ci): lock automatic HF fallback and PostMeet gate
k0880k1dabe7a -
fix(postmeet): mount v4 route contract in live editor
k0880kUnverified signature* fix(postmeet): make route seam safe to mount at startup * fix(postmeet): mount v4 URL contract from editor startup * test(postmeet): lock startup route mount and GPU readinessd5867b3 -
test(postmeet): lock GPU health and v4 publish contract
k0880kbdacabf -
test(postmeet): lock real signalmesh-gpu identity
k0880k1ebb62b -
test(postmeet): prevent GPU and Vlog pipeline regression
k0880ka6718ad
-
test(bighelp): cover direct provider binding
k0880k1fe702c -
test(bighelp): prove stable aliases require real edge evidence
k0880kfd676ca -
feat(capability): typed .fq driver contract, Studio adapter, evidence-based closure
Wraps PostMeet as a resident capability and gives Studio a generic way to operate any typed .fq driver pack, with closure state computed from evidence rather than asserted by whoever wrote the document. Capability contract (core/capability_fq.py) Seven roles carried by a typed .fq: manifest, health, inventory, execute, status, evidence, receipt. Registration refuses a pack that advertises a role with no runtime behind it, and health is graded by the contract from reported checks — a driver cannot mark itself healthy. Studio adapter (core/studio_fq.py, /studio/fq) Renders any capability from describe() alone: identity, health, closure, actions, live jobs, evidence, receipts. Job phases and progress fields come from the manifest, so no product-specific branch exists — asserted by a test that greps the source for product names. Accepted is never success; only capability.receipt can report verified. Driver auth renders as status only, never as a secret. Promotion aut…249db15
-
test(operator): align browser harness input contract
k0880k8f07721 -
test(operator): cover fuzzy intent, prerequisites and universal widget
k0880ka1571b3 -
feat(qm): add device-bound resident token claims
882f141 -
test(brand): enforce SignalMesh-only lockup
k0880k1d3ae18
-
feat: durable resident creation and chaining control plane
k0880kUnverified signatureAdd persistent resident identities, private tokens, leases, explicit chains, agent-native APIs, WorldState evidence, and the GridStack control surface.08b6964 -
feat: private QM resident integration
k0880kUnverified signatureAdd the protected QM transport, canonical hydration, guarded outbound broadcast, two-scope proof, and sanitized GridStack workplace UI.9839950 -
test: prove canonical brand replacement leaves functional SVGs intact
k0880ke31fafc -
test: prove retrospective change stays read-only
k0880k76764a3 -
test: prove Mesh.gg signed receiver and authority boundary
k0880kc0bdc62 -
test: prove multi-party ceremony spindle and secret invariants
k0880kbde18f1 -
test(maintenance): keep last error pinned to manifested root
k0880kc81ce8a -
test: prove visual mitosis surface is mounted
k0880k703bb7e -
test: prove nested living visual diffs
k0880k714f1f1 -
test: assert schema-visible marketing route in OpenAPI
k0880k94695d5 -
test: require OpenAPI and XSTRAPP on unified product surface
k0880k9150ff8 -
test: cover XSTRAPP mitosis, hidden spindles, and projections
k0880k4bdbe7a -
test: validate CI telemetry becomes resident state
k0880k98f607c -
test: validate permanent developer maintenance residency
k0880k7fb693b -
test: validate portable marketing fq boundary
k0880k15bbf92 -
Merge unified meshMarketingPortal
k0880kUnverified signaturefeat: unify meshMarketingPortal with canonical portal/outreach62bbda5 -
test: prove unified deployable surface mounts workspace and marketing
k0880k08c1063 -
test: cover team auth, department lens, and setup persistence
k0880k4df83f2 -
test: cover artifact workspace isolation and safety
k0880k4003f58 -
test(fq): prove fq pointer hydration stays tuned and bounded
k0880k99b601f -
test(fq): pin catalog playlist and reconnect semantics
k0880k5d02bae -
test: pin runtime genome across storage ingestion layers
k0880kacd8cee -
test: pin parcel and opaque ref boundaries
k0880k5402290 -
test: runtime genome on every accepted signal
k0880k4e94fdf
-
test(marketing): never restart a completed Gmail sequence
k0880k5daeaa9 -
test(marketing): resolve scheduled social as job owner
k0880k5aa1632 -
test(marketing): pin unified portal invariants
k0880kfa8d3e3 -
fix(deploy): leave the matrix API tests with the work they test
core/matrix.py is a roster module and polymesh reads it for capacity. Its tests exercise /ui/matrix routes and chat seat selection, which live in an signalmesh_api.py that is not part of this deploy — so they fail here for a true reason: the thing they test is not shipping yet. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>c81fa65 -
feat(intake): a wall on the way in, and the portal spine beside it
22,487 journalled signals held 2,255 distinct things. One typewriter post arrived 168 times, and a run of dev.to posts sold aged Gmail accounts. Asking an honest question returned the loudest thing rather than the truest one, because arrival was the only thing being counted. Intake refuses three facts anyone can check — already here, nothing said, selling — and names which one. A duplicate still touches the copy already held, because a feed re-sending its front page is alive and refusing it silently would make the frequency read as dead. The fingerprint rides on the signal, or a registry keeps refusing content it no longer holds. Filler can no longer learn a trail: a trail is permanent and was read back with no threshold, so one lucky score for "also" answered for "also" forever. The portal spine rides along inert — nothing imports it yet. orgs, portal, polymesh and outreach are additive, and signalmesh_api is untouched, so this deploy changes exactly one live path: what is allowed o…ea2bedd
-
feat(world): the mesh writes the documents, and the city has people in it
layout.generate — the inverse of layout.docbank. The annotator recovers labels from a document somebody else wrote, so it infers columns and merges lines to guess where a region ends. Generation knows every block before the page exists, so it colours per BLOCK INSTANCE rather than per label: a block's region is the union of its own words, exact by construction, one region per page across a column break, no tolerance to tune. Same two compiles and the same fidelity gate — a run whose glyphs moved emits no labels at all. Four LaTeX drift bugs the gate caught and refused: \textcolor's \leavevmode starting a paragraph on an empty hbox instead of a \parindent box; a font change grouped on one side only; leading `~` glue discarded at a line start in the clean render but shielded by a colour whatsit in the annotated one; and list bullets extracting as unlabelled, clipping every list region off its own markers. All eight block kinds now compile at 0.000pt. drivemesh — a Drive folder is a pla…dff1081 -
test: compile every shipped module on the Python the Space actually runs
The outage that just happened could not be caught by any test running on this machine's interpreter, because the code was valid here and invalid there. The obvious guard does not work. ast.parse(src, feature_version=(3, 11)) parses the exact file that crashed the Space — feature_version gates a few AST features, not the f-string grammar. A test built on it would have stayed green through the whole outage, which is worse than having no test. So this shells out to a real python3.11 and compiles each module. It skips loudly if that interpreter is missing rather than quietly claiming to have checked. And it carries a regression test that feeds the guard the exact construct that took the site down, because a guard nobody has watched fail is a guard nobody should trust. Verified against dd35f14, the commit that broke production: flagged, with the same SyntaxError the Space reported. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XGo…db5917b