LocalEndpoint ConnectPrivate AI for Windows
Menu

LocalEndpoint guide

Local Model Lifecycle

Local model lifecycle documentation for reviewed downloaded models, immutable snapshots, sidecars, hardware-fit evidence, safetensors conversion boundaries, and worker allocation gates.

Desktop release JSON Runtime integration

Local model lifecycle

Downloaded model files become usable only after Desktop review.

LocalEndpoint is designed for local downloaded models, but a file on disk is not enough. The desktop path records review evidence before a model can become a selected runtime candidate.

License sidecar Inventory sidecar SHA-256 Byte count Hardware fit Selected model identity
Review

Intake is evidence-backed.

Desktop verifies model format, license sidecar, inventory sidecar, hash, byte count, revision, and hardware fit before activation.

  • Phase 3.519 package evidence records LocalModelUaixMemoryRunReadiness as Required For UAIX Memory Run Readiness so Hub runs require loaded .uaix package memory, .uai/persona.uai, and a Documents-backed wiki root before local chat readiness can pass.
  • Phase 3.519 source hardening expands Hub run readiness to nine gates: prompt text, .uaix package, persona.uai, Documents wiki root, local model identity, download, activation, runtime, and worker.
  • Phase 3.516 package evidence records LocalModelHubCatalogCandidateSelectionSummary as Required For Safe Local Model Selection so ready candidates show owner-approved local download and blocked candidates show the required different-choice action before download.
  • Phase 3.516 source hardening records Models UI SelectionSummary.StatusText, SelectionSummary.PrimaryActionLabel, and SelectionSummary.EvidenceText as Required For Hub Candidate Action Discovery so first-time users can see one candidate decision and one next action without reading raw metadata.
  • Phase 3.514 package evidence records LocalModelHubRunReadinessCard as Required For First Local Chat Action Card so Models UI exposes exactly one next action from the same display-safe readiness summary that gates the Run Local workflow.
  • Phase 3.514 source hardening records LocalEndpoint.Models.HubRun.ReadinessCard automation IDs as Required For First Local Chat Action Discovery so smoke automation can verify the compact next-action card without scraping long explanatory text.
  • Phase 3.512 source hardening records LocalModelHubRunReadinessSummary progress counts as Required For First Local Chat Readiness Progress so LocalEndpoint.Models.HubRun.ReadinessProgress shows ready required gates and total required gates from the same checklist rows as the detailed Models UI gate list.
  • Phase 3.512 source hardening records ReadyChecklistItemCount, TotalChecklistItemCount, and ProgressText as Required For Display-Safe Local Run Progress so readiness count text stays count-only and does not carry prompt text or generated text.
  • Phase 3.511 source hardening records RunHubWorkflowCommand readiness-bound CanExecute as Required For First Local Chat Command Gating so Run Local cannot execute while LocalModelHubRunReadinessSummary is blocked.
  • Phase 3.511 source hardening records RunHubWorkflowAsync readiness recheck as Required For Direct Call Fail-Closed Behavior so direct command invocation returns the display-safe readiness blocker instead of starting workflow evidence.
  • Phase 3.510 source hardening records LocalModelHubRunReadinessSummary and LocalModelHubRunReadinessChecklistItem as Required For First Local Chat Readiness so prompt text, local model identity, download gate, activation gate, runtime gate, and worker gate are visible as structured, display-safe Models UI rows.
  • Phase 3.510 source hardening records ModelIdentityReady as Required For Ready Hub Run State so approvals alone cannot mark a local run ready without a visible local-ready catalog entry, catalog candidate, or active verified local model.
  • Phase 3.509 source hardening records installed-app Models UI smoke automation evidence for LocalEndpoint.Models.FirstRunGuide.ChecklistTitle, LocalEndpoint.Models.FirstRunGuide.Checklist, and LocalEndpoint.Models.HubRun.ReadinessStatus as Required For First Local Chat Evidence Discovery so future smoke reports must prove the checklist and next-action readiness status remain visible.
  • Phase 3.508 source hardening records LocalModelFirstRunGuide and LocalModelFirstRunChecklistItem as Required For First Local Chat so prompt text, local-ready GGUF choice or active verified model reuse, systems-without-GPUs compatibility, systems-with-GPUs compatibility, download approval, activation approval, runtime approval, worker approval, and prompt/response privacy boundaries appear as a compact Models UI checklist.
  • Phase 3.506 source hardening records LocalModelHubRunWorkflowPhaseStateDefinition as Required For Local Hub Run Phase State Validation so NotStarted, Passed, Blocked, and Skipped are the only accepted progress states before Hub run evidence emission.
  • Phase 3.506 source hardening records LocalModelHubRunWorkflowService named state consumption as Required For Local Hub Run Progress State Consistency so workflow gates cannot drift from the shared phase-state contract.
  • Phase 3.505 source hardening records LocalModelHubRunWorkflowPhaseDefinition as Required For Local Hub Run Phase Identity Projection so prompt gate, Hub browse, catalog save, registry reuse, download, activation, provider start, worker stream, and terminal evidence use one named phase contract.
  • Phase 3.505 source hardening records LocalModelHubRunProgressRecorder named phase consumption as Required For Local Hub Run Progress Evidence Consistency so UI, CLI, workflow service, and smoke evidence cannot drift on phase keys or labels.
  • Phase 3.504 source hardening records LocalModelHubRunLifecycleSummary as Required For Local Model Lifecycle Status Projection so browse, download, activation, and run status text is generated from one display-safe domain contract.
  • Phase 3.504 source hardening records LocalModelHubRunTerminalSummary as Required For Local Hub Run Terminal Status Projection so outcome, message, token event count, terminal kind, progress phase count, completion state, and generated-text availability are shared by Models UI and CLI without carrying prompt or generated text content.
  • Phase 3.481 packages the local Hugging Face chat-run CLI path into the current downloadable desktop tester ZIP: browse Hub metadata, filter GGUF candidates, download locally, verify SHA-256 and byte count, process intake, activate, load .uaix memory, approve readiness, and run a compatible local GGUF through the managed worker.
  • Phase 3.482 proves the local Hugging Face path again in fresh source/dev smoke: browse Hub metadata, filter GGUF candidates, download locally, verify SHA-256 and byte count, process intake, activate, load HelpfulAssistant.uaix, approve readiness, and run a compatible local GGUF through the managed worker.
  • Hugging Face Hub browse uses metadata for local catalog projection; it does not call hosted inference, Inference Providers, Spaces, Gradio APIs, prompt upload routes, or credentialed hosted endpoints.
  • Model candidates must pass format, license sidecar, inventory sidecar, SHA-256, byte-count, revision, and hardware-fit checks before intake.
  • Accepted snapshots become immutable content-addressed local artifacts; unchanged snapshots record AlreadyStored evidence.
  • Safetensors conversion is evidence-only until converted GGUF or ONNX output re-enters local intake as a new reviewed snapshot.
  • Catalog download, intake activation, selected-model use, worker allocation, prompt tokenization, inference, and token streaming all stay behind local desktop gates.
  • LocalChatViabilityEvidence blocks non-viable runtime, model, offload, local API, memory authority, UAIX authority, or public-site boundary state before a worker envelope exists.
  • Associated GGUF package-folder enumeration failures after scan block local model intake with VerificationBlocked evidence before content-addressed storage write, intake cleanup, or runtime execution can occur.
  • Actual GPU execution claims require runtime adapter backend and device IDs to match the selected offload plan backend and device IDs before LocalEndpoint accepts `GpuObserved` evidence.
Runtime

Execution waits for identity gates.

Worker allocation, prompt tokenization, inference, and token streaming remain blocked until LocalChatViabilityEvidence, the active registry row, requested model entry, session identity, turn identity, provider request fingerprint, and runtime context agree.

  • RequestedModelEntryId prevents stale selected-model use.
  • RequestHandoffSha256 blocks start-to-stream request drift before worker handoff.
  • LocalChatViabilityEvidence records a blocked no-op before worker envelope creation when runtime, model, offload, local API, memory authority, UAIX authority, or public boundary state does not fit.
  • Registry and audit records keep prompt text and generated text out of persisted evidence.
  • Converted outputs must re-enter local intake before runtime use.
Model lifecycle boundary No model upload No hosted inference No provider API relay No public-site prompt intake No public-site model execution

Operating boundary

Public clarity, local authority.

Public product, docs, download, and validation pages do not dispatch desktop commands, probe localhost, upload files, collect runtime telemetry, or claim runtime safety certification. Account credentials are accepted only on dedicated account routes.

Website explainsNo public command dispatch Browser checksNo upload intake Your work stays localNo localhost probing Verify downloadsChecksum-backed artifacts