agent-hitch 0.2.9 → 0.2.10
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +206 -137
- package/README.zh-CN.md +193 -119
- package/dist/scripts/canary-native-training.js +221 -0
- package/dist/scripts/canary-native-training.js.map +1 -0
- package/dist/scripts/canary-remote-training.js +246 -0
- package/dist/scripts/canary-remote-training.js.map +1 -0
- package/dist/scripts/canary-remote-verifier.js +124 -0
- package/dist/scripts/canary-remote-verifier.js.map +1 -0
- package/dist/scripts/canary-worker-host.js +113 -0
- package/dist/scripts/canary-worker-host.js.map +1 -0
- package/dist/scripts/canary-worker-peer.js +188 -0
- package/dist/scripts/canary-worker-peer.js.map +1 -0
- package/dist/scripts/check-architecture.js +4 -3
- package/dist/scripts/check-architecture.js.map +1 -1
- package/dist/src/adapters/catalog.js +2 -1
- package/dist/src/adapters/catalog.js.map +1 -1
- package/dist/src/adapters/providers/codex.js +65 -2
- package/dist/src/adapters/providers/codex.js.map +1 -1
- package/dist/src/adapters/providers/model-call.js +17 -1
- package/dist/src/adapters/providers/model-call.js.map +1 -1
- package/dist/src/adapters/providers/training-tool.js +29 -0
- package/dist/src/adapters/providers/training-tool.js.map +1 -0
- package/dist/src/artifacts/index.js +1 -1
- package/dist/src/artifacts/index.js.map +1 -1
- package/dist/src/artifacts/store.js +15 -10
- package/dist/src/artifacts/store.js.map +1 -1
- package/dist/src/backends/harbor/backend.js +24 -15
- package/dist/src/backends/harbor/backend.js.map +1 -1
- package/dist/src/backends/harbor/bridge-environment.js +12 -0
- package/dist/src/backends/harbor/bridge-environment.js.map +1 -0
- package/dist/src/backends/harbor/environment-config.js +2 -2
- package/dist/src/backends/harbor/environment-config.js.map +1 -1
- package/dist/src/backends/harbor/host-credential-helper.js +81 -0
- package/dist/src/backends/harbor/host-credential-helper.js.map +1 -0
- package/dist/src/backends/harbor/index.js +3 -0
- package/dist/src/backends/harbor/index.js.map +1 -1
- package/dist/src/backends/harbor/model-proxy-config.js +18 -1
- package/dist/src/backends/harbor/model-proxy-config.js.map +1 -1
- package/dist/src/backends/harbor/observation.js +61 -0
- package/dist/src/backends/harbor/observation.js.map +1 -0
- package/dist/src/backends/harbor/portable-regrade.js +150 -0
- package/dist/src/backends/harbor/portable-regrade.js.map +1 -0
- package/dist/src/backends/harbor/process.js +26 -5
- package/dist/src/backends/harbor/process.js.map +1 -1
- package/dist/src/backends/harbor/regrade.js +4 -1
- package/dist/src/backends/harbor/regrade.js.map +1 -1
- package/dist/src/backends/harbor/supervisor.js +22 -0
- package/dist/src/backends/harbor/supervisor.js.map +1 -1
- package/dist/src/backends/harbor/tools.js +2 -0
- package/dist/src/backends/harbor/tools.js.map +1 -1
- package/dist/src/backends/index.js +3 -0
- package/dist/src/backends/index.js.map +1 -1
- package/dist/src/cli/arguments.js +33 -0
- package/dist/src/cli/arguments.js.map +1 -1
- package/dist/src/cli/commands/capabilities.js +14 -1
- package/dist/src/cli/commands/capabilities.js.map +1 -1
- package/dist/src/cli/commands/daemon.js +19 -0
- package/dist/src/cli/commands/daemon.js.map +1 -1
- package/dist/src/cli/commands/eval.js +38 -10
- package/dist/src/cli/commands/eval.js.map +1 -1
- package/dist/src/cli/commands/local.js +130 -0
- package/dist/src/cli/commands/local.js.map +1 -0
- package/dist/src/cli/commands/model-node.js +26 -0
- package/dist/src/cli/commands/model-node.js.map +1 -0
- package/dist/src/cli/commands/models.js +72 -0
- package/dist/src/cli/commands/models.js.map +1 -0
- package/dist/src/cli/commands/run.js +36 -1
- package/dist/src/cli/commands/run.js.map +1 -1
- package/dist/src/cli/commands/training.js +32 -0
- package/dist/src/cli/commands/training.js.map +1 -0
- package/dist/src/cli/commands/worker.js +30 -2
- package/dist/src/cli/commands/worker.js.map +1 -1
- package/dist/src/cli/main.js +8 -0
- package/dist/src/cli/main.js.map +1 -1
- package/dist/src/cli/output.js +24 -2
- package/dist/src/cli/output.js.map +1 -1
- package/dist/src/control-plane/eval-records.js +2 -0
- package/dist/src/control-plane/eval-records.js.map +1 -1
- package/dist/src/control-plane/eval-recovery.js +7 -1
- package/dist/src/control-plane/eval-recovery.js.map +1 -1
- package/dist/src/control-plane/eval-scheduler.js +24 -50
- package/dist/src/control-plane/eval-scheduler.js.map +1 -1
- package/dist/src/control-plane/eval-submission-persist.js +30 -0
- package/dist/src/control-plane/eval-submission-persist.js.map +1 -0
- package/dist/src/control-plane/execution-observation-contract.js +60 -0
- package/dist/src/control-plane/execution-observation-contract.js.map +1 -0
- package/dist/src/control-plane/index.js +10 -0
- package/dist/src/control-plane/index.js.map +1 -1
- package/dist/src/control-plane/inference-manager.js +357 -0
- package/dist/src/control-plane/inference-manager.js.map +1 -0
- package/dist/src/control-plane/managed-gateway-receipt.js +26 -0
- package/dist/src/control-plane/managed-gateway-receipt.js.map +1 -0
- package/dist/src/control-plane/managed-service-recovery.js +158 -0
- package/dist/src/control-plane/managed-service-recovery.js.map +1 -0
- package/dist/src/control-plane/model-capture-planning.js +15 -1
- package/dist/src/control-plane/model-capture-planning.js.map +1 -1
- package/dist/src/control-plane/ordered-eval-control.js +221 -0
- package/dist/src/control-plane/ordered-eval-control.js.map +1 -0
- package/dist/src/control-plane/remote-backend-result.js +18 -0
- package/dist/src/control-plane/remote-backend-result.js.map +1 -0
- package/dist/src/control-plane/remote-execution-lease.js +15 -0
- package/dist/src/control-plane/remote-execution-lease.js.map +1 -0
- package/dist/src/control-plane/remote-execution-observation.js +108 -0
- package/dist/src/control-plane/remote-execution-observation.js.map +1 -0
- package/dist/src/control-plane/remote-model-routes.js +107 -0
- package/dist/src/control-plane/remote-model-routes.js.map +1 -0
- package/dist/src/control-plane/remote-observation-reporter.js +37 -0
- package/dist/src/control-plane/remote-observation-reporter.js.map +1 -0
- package/dist/src/control-plane/remote-offer-contract.js +48 -0
- package/dist/src/control-plane/remote-offer-contract.js.map +1 -0
- package/dist/src/control-plane/remote-rerun-journal.js +61 -0
- package/dist/src/control-plane/remote-rerun-journal.js.map +1 -0
- package/dist/src/control-plane/remote-result-transport.js +55 -9
- package/dist/src/control-plane/remote-result-transport.js.map +1 -1
- package/dist/src/control-plane/remote-verifier-import.js +52 -0
- package/dist/src/control-plane/remote-verifier-import.js.map +1 -0
- package/dist/src/control-plane/remote-verifier-result.js +61 -0
- package/dist/src/control-plane/remote-verifier-result.js.map +1 -0
- package/dist/src/control-plane/remote-verifier-source-transport.js +68 -0
- package/dist/src/control-plane/remote-verifier-source-transport.js.map +1 -0
- package/dist/src/control-plane/remote-work-coordinator.js +103 -47
- package/dist/src/control-plane/remote-work-coordinator.js.map +1 -1
- package/dist/src/control-plane/remote-work-inputs.js +26 -5
- package/dist/src/control-plane/remote-work-inputs.js.map +1 -1
- package/dist/src/control-plane/remote-work-recovery.js +123 -34
- package/dist/src/control-plane/remote-work-recovery.js.map +1 -1
- package/dist/src/control-plane/remote-worker-artifacts.js +10 -7
- package/dist/src/control-plane/remote-worker-artifacts.js.map +1 -1
- package/dist/src/control-plane/remote-worker-cleanup.js +104 -0
- package/dist/src/control-plane/remote-worker-cleanup.js.map +1 -0
- package/dist/src/control-plane/remote-worker-client.js +156 -14
- package/dist/src/control-plane/remote-worker-client.js.map +1 -1
- package/dist/src/control-plane/remote-worker-codecs.js +160 -0
- package/dist/src/control-plane/remote-worker-codecs.js.map +1 -0
- package/dist/src/control-plane/remote-worker-generation-recovery.js +67 -0
- package/dist/src/control-plane/remote-worker-generation-recovery.js.map +1 -0
- package/dist/src/control-plane/remote-worker-ownership.js +51 -0
- package/dist/src/control-plane/remote-worker-ownership.js.map +1 -0
- package/dist/src/control-plane/remote-worker-protocol.js +85 -119
- package/dist/src/control-plane/remote-worker-protocol.js.map +1 -1
- package/dist/src/control-plane/remote-worker-runner.js +95 -18
- package/dist/src/control-plane/remote-worker-runner.js.map +1 -1
- package/dist/src/control-plane/remote-workers.js +33 -5
- package/dist/src/control-plane/remote-workers.js.map +1 -1
- package/dist/src/control-plane/rerun-recovery.js +89 -0
- package/dist/src/control-plane/rerun-recovery.js.map +1 -0
- package/dist/src/control-plane/rerun-scheduler.js +70 -96
- package/dist/src/control-plane/rerun-scheduler.js.map +1 -1
- package/dist/src/control-plane/rerun-source.js +57 -0
- package/dist/src/control-plane/rerun-source.js.map +1 -0
- package/dist/src/control-plane/worker-model-relay.js +76 -0
- package/dist/src/control-plane/worker-model-relay.js.map +1 -0
- package/dist/src/controller-runtime/hash.js +3 -0
- package/dist/src/controller-runtime/hash.js.map +1 -1
- package/dist/src/controller-runtime/index.js +2 -0
- package/dist/src/controller-runtime/index.js.map +1 -1
- package/dist/src/controller-runtime/observation.js +29 -0
- package/dist/src/controller-runtime/observation.js.map +1 -0
- package/dist/src/controller-runtime/resident.js +22 -0
- package/dist/src/controller-runtime/resident.js.map +1 -0
- package/dist/src/daemon/client.js +34 -1
- package/dist/src/daemon/client.js.map +1 -1
- package/dist/src/daemon/execution-observation.js +38 -0
- package/dist/src/daemon/execution-observation.js.map +1 -0
- package/dist/src/daemon/http.js +57 -0
- package/dist/src/daemon/http.js.map +1 -0
- package/dist/src/daemon/inference-prepare.js +47 -0
- package/dist/src/daemon/inference-prepare.js.map +1 -0
- package/dist/src/daemon/scheduler.js +4 -1
- package/dist/src/daemon/scheduler.js.map +1 -1
- package/dist/src/daemon/server.js +80 -82
- package/dist/src/daemon/server.js.map +1 -1
- package/dist/src/daemon/worker-observation.js +34 -0
- package/dist/src/daemon/worker-observation.js.map +1 -0
- package/dist/src/daemon/worker-ownership.js +71 -0
- package/dist/src/daemon/worker-ownership.js.map +1 -0
- package/dist/src/daemon/worker-routes.js +83 -15
- package/dist/src/daemon/worker-routes.js.map +1 -1
- package/dist/src/domain/execution-observation.js +2 -0
- package/dist/src/domain/execution-observation.js.map +1 -0
- package/dist/src/domain/index.js +1 -0
- package/dist/src/domain/index.js.map +1 -1
- package/dist/src/domain/inference.js +2 -0
- package/dist/src/domain/inference.js.map +1 -0
- package/dist/src/domain/runs.js.map +1 -1
- package/dist/src/domain/training.js +2 -0
- package/dist/src/domain/training.js.map +1 -0
- package/dist/src/domain/validation.js +11 -0
- package/dist/src/domain/validation.js.map +1 -1
- package/dist/src/domain/verifier-source.js +2 -0
- package/dist/src/domain/verifier-source.js.map +1 -0
- package/dist/src/domain/worker-recovery.js +2 -0
- package/dist/src/domain/worker-recovery.js.map +1 -0
- package/dist/src/evals/eval-logical-plan.js +8 -1
- package/dist/src/evals/eval-logical-plan.js.map +1 -1
- package/dist/src/evals/execution-leases.js +15 -2
- package/dist/src/evals/execution-leases.js.map +1 -1
- package/dist/src/evals/execution-plan.js +11 -1
- package/dist/src/evals/execution-plan.js.map +1 -1
- package/dist/src/evals/harbor-bridge-error.js +8 -0
- package/dist/src/evals/harbor-bridge-error.js.map +1 -1
- package/dist/src/evals/index.js +19 -0
- package/dist/src/evals/index.js.map +1 -1
- package/dist/src/evals/model-capture-plan.js +12 -0
- package/dist/src/evals/model-capture-plan.js.map +1 -1
- package/dist/src/evals/model-capture-runtime.js +25 -0
- package/dist/src/evals/model-capture-runtime.js.map +1 -1
- package/dist/src/evals/physical-work-plan.js +57 -0
- package/dist/src/evals/physical-work-plan.js.map +1 -0
- package/dist/src/evals/preparation-rerun.js +6 -0
- package/dist/src/evals/preparation-rerun.js.map +1 -1
- package/dist/src/evals/recovery.js +8 -1
- package/dist/src/evals/recovery.js.map +1 -1
- package/dist/src/evals/regrade-evidence.js +4 -4
- package/dist/src/evals/regrade-evidence.js.map +1 -1
- package/dist/src/evals/remote-execution-ownership.js +50 -0
- package/dist/src/evals/remote-execution-ownership.js.map +1 -0
- package/dist/src/evals/remote-infrastructure-retry.js +2 -0
- package/dist/src/evals/remote-infrastructure-retry.js.map +1 -1
- package/dist/src/evals/remote-lease-release.js +58 -0
- package/dist/src/evals/remote-lease-release.js.map +1 -0
- package/dist/src/evals/remote-rerun-execution.js +85 -0
- package/dist/src/evals/remote-rerun-execution.js.map +1 -0
- package/dist/src/evals/remote-trial.js +6 -0
- package/dist/src/evals/remote-trial.js.map +1 -0
- package/dist/src/evals/remote-verifier-assessment.js +85 -0
- package/dist/src/evals/remote-verifier-assessment.js.map +1 -0
- package/dist/src/evals/remote-verifier-contract.js +43 -0
- package/dist/src/evals/remote-verifier-contract.js.map +1 -0
- package/dist/src/evals/remote-verifier-execution.js +105 -0
- package/dist/src/evals/remote-verifier-execution.js.map +1 -0
- package/dist/src/evals/remote-verifier-journal.js +21 -0
- package/dist/src/evals/remote-verifier-journal.js.map +1 -0
- package/dist/src/evals/remote-verifier-outcome.js +51 -0
- package/dist/src/evals/remote-verifier-outcome.js.map +1 -0
- package/dist/src/evals/remote-verifier-rerun.js +134 -0
- package/dist/src/evals/remote-verifier-rerun.js.map +1 -0
- package/dist/src/evals/remote-verifier-selection.js +61 -0
- package/dist/src/evals/remote-verifier-selection.js.map +1 -0
- package/dist/src/evals/remote-verifier-source.js +33 -0
- package/dist/src/evals/remote-verifier-source.js.map +1 -0
- package/dist/src/evals/remote-worker-cleanup.js +46 -0
- package/dist/src/evals/remote-worker-cleanup.js.map +1 -0
- package/dist/src/evals/request.js +58 -1
- package/dist/src/evals/request.js.map +1 -1
- package/dist/src/evals/rerun-completion.js +39 -0
- package/dist/src/evals/rerun-completion.js.map +1 -0
- package/dist/src/evals/rerun-resume.js +70 -0
- package/dist/src/evals/rerun-resume.js.map +1 -0
- package/dist/src/evals/rerun-state.js +25 -0
- package/dist/src/evals/rerun-state.js.map +1 -0
- package/dist/src/evals/rerun-types.js.map +1 -1
- package/dist/src/evals/rerun.js +90 -46
- package/dist/src/evals/rerun.js.map +1 -1
- package/dist/src/evals/service.js +59 -5
- package/dist/src/evals/service.js.map +1 -1
- package/dist/src/evals/verifier-only-rerun.js +2 -16
- package/dist/src/evals/verifier-only-rerun.js.map +1 -1
- package/dist/src/evals/verifier-rerun-result.js +24 -0
- package/dist/src/evals/verifier-rerun-result.js.map +1 -0
- package/dist/src/evals/verifier-source-import.js +48 -0
- package/dist/src/evals/verifier-source-import.js.map +1 -0
- package/dist/src/evals/verifier-source.js +139 -0
- package/dist/src/evals/verifier-source.js.map +1 -0
- package/dist/src/foundation/base64.js +19 -0
- package/dist/src/foundation/base64.js.map +1 -0
- package/dist/src/foundation/config.js +10 -0
- package/dist/src/foundation/config.js.map +1 -1
- package/dist/src/foundation/credential-redaction.js +22 -9
- package/dist/src/foundation/credential-redaction.js.map +1 -1
- package/dist/src/foundation/index.js +1 -0
- package/dist/src/foundation/index.js.map +1 -1
- package/dist/src/foundation/process-identity.js +13 -9
- package/dist/src/foundation/process-identity.js.map +1 -1
- package/dist/src/inference/daemon-capacity.js +41 -0
- package/dist/src/inference/daemon-capacity.js.map +1 -0
- package/dist/src/inference/device-reservation.js +58 -0
- package/dist/src/inference/device-reservation.js.map +1 -0
- package/dist/src/inference/doctor.js +95 -0
- package/dist/src/inference/doctor.js.map +1 -0
- package/dist/src/inference/generation-recovery.js +83 -0
- package/dist/src/inference/generation-recovery.js.map +1 -0
- package/dist/src/inference/index.js +24 -0
- package/dist/src/inference/index.js.map +1 -0
- package/dist/src/inference/lock.js +331 -0
- package/dist/src/inference/lock.js.map +1 -0
- package/dist/src/inference/managed-launcher.js +35 -0
- package/dist/src/inference/managed-launcher.js.map +1 -0
- package/dist/src/inference/manifest.js +262 -0
- package/dist/src/inference/manifest.js.map +1 -0
- package/dist/src/inference/materialize.js +66 -0
- package/dist/src/inference/materialize.js.map +1 -0
- package/dist/src/inference/model-location.js +34 -0
- package/dist/src/inference/model-location.js.map +1 -0
- package/dist/src/inference/model-store.js +439 -0
- package/dist/src/inference/model-store.js.map +1 -0
- package/dist/src/inference/node-client.js +167 -0
- package/dist/src/inference/node-client.js.map +1 -0
- package/dist/src/inference/node-registry.js +79 -0
- package/dist/src/inference/node-registry.js.map +1 -0
- package/dist/src/inference/observation.js +79 -0
- package/dist/src/inference/observation.js.map +1 -0
- package/dist/src/inference/preflight.js +132 -0
- package/dist/src/inference/preflight.js.map +1 -0
- package/dist/src/inference/process-sglang.js +169 -0
- package/dist/src/inference/process-sglang.js.map +1 -0
- package/dist/src/inference/protocol-probe.js +29 -0
- package/dist/src/inference/protocol-probe.js.map +1 -0
- package/dist/src/inference/remote-model-store.js +24 -0
- package/dist/src/inference/remote-model-store.js.map +1 -0
- package/dist/src/inference/runtime-catalog.js +41 -0
- package/dist/src/inference/runtime-catalog.js.map +1 -0
- package/dist/src/inference/runtime-store.js +93 -0
- package/dist/src/inference/runtime-store.js.map +1 -0
- package/dist/src/inference/service-attachment.js +46 -0
- package/dist/src/inference/service-attachment.js.map +1 -0
- package/dist/src/inference/service-inspection.js +38 -0
- package/dist/src/inference/service-inspection.js.map +1 -0
- package/dist/src/inference/sglang.js +279 -0
- package/dist/src/inference/sglang.js.map +1 -0
- package/dist/src/inference/supervisor.js +460 -0
- package/dist/src/inference/supervisor.js.map +1 -0
- package/dist/src/model-access/environment.js +28 -0
- package/dist/src/model-access/environment.js.map +1 -0
- package/dist/src/model-access/index.js +4 -0
- package/dist/src/model-access/index.js.map +1 -1
- package/dist/src/model-access/local-gateway-request.js +139 -0
- package/dist/src/model-access/local-gateway-request.js.map +1 -0
- package/dist/src/model-access/local-gateway.js +323 -0
- package/dist/src/model-access/local-gateway.js.map +1 -0
- package/dist/src/model-access/proxy.js +90 -5
- package/dist/src/model-access/proxy.js.map +1 -1
- package/dist/src/model-access/remote-model.js +58 -0
- package/dist/src/model-access/remote-model.js.map +1 -0
- package/dist/src/model-access/training.js +115 -0
- package/dist/src/model-access/training.js.map +1 -0
- package/dist/src/runs/adapter-process.js +2 -2
- package/dist/src/runs/adapter-process.js.map +1 -1
- package/dist/src/runs/executor-types.js +2 -0
- package/dist/src/runs/executor-types.js.map +1 -0
- package/dist/src/runs/executor.js +54 -34
- package/dist/src/runs/executor.js.map +1 -1
- package/dist/src/runs/index.js +1 -0
- package/dist/src/runs/index.js.map +1 -1
- package/dist/src/runs/local-inference-environment.js +59 -0
- package/dist/src/runs/local-inference-environment.js.map +1 -0
- package/dist/src/runs/local-inference-run.js +53 -0
- package/dist/src/runs/local-inference-run.js.map +1 -0
- package/dist/src/runs/outcome.js +22 -0
- package/dist/src/runs/outcome.js.map +1 -1
- package/dist/src/runs/records.js +9 -0
- package/dist/src/runs/records.js.map +1 -1
- package/dist/src/runs/request.js +58 -3
- package/dist/src/runs/request.js.map +1 -1
- package/dist/src/runs/training-evidence.js +22 -0
- package/dist/src/runs/training-evidence.js.map +1 -0
- package/dist/src/runs/training-runtime.js +20 -0
- package/dist/src/runs/training-runtime.js.map +1 -0
- package/dist/src/workers/execution-observation.js +32 -0
- package/dist/src/workers/execution-observation.js.map +1 -0
- package/dist/src/workers/index.js +2 -0
- package/dist/src/workers/index.js.map +1 -1
- package/dist/src/workers/remote-harbor-ownership.js +266 -0
- package/dist/src/workers/remote-harbor-ownership.js.map +1 -0
- package/dist/src/workers/remote-harbor-verifier.js +76 -0
- package/dist/src/workers/remote-harbor-verifier.js.map +1 -0
- package/dist/src/workers/remote-harbor-work-spec.js +66 -10
- package/dist/src/workers/remote-harbor-work-spec.js.map +1 -1
- package/dist/src/workers/remote-harbor-worker.js +56 -49
- package/dist/src/workers/remote-harbor-worker.js.map +1 -1
- package/dist/src/workers/remote-worker-host.js +33 -0
- package/dist/src/workers/remote-worker-host.js.map +1 -0
- package/docs/local-model-inference-spec.zh-CN.md +666 -0
- package/docs/schemas/controller-runtime-observation.schema.json +21 -0
- package/docs/schemas/eval-execution-plan.schema.json +1 -0
- package/docs/schemas/eval-request.schema.json +3 -1
- package/docs/schemas/execution-lease.schema.json +95 -1
- package/docs/schemas/execution-provider-status.schema.json +162 -28
- package/docs/schemas/inference-execution-evidence.schema.json +487 -0
- package/docs/schemas/inference-lock.schema.json +487 -0
- package/docs/schemas/inference-runtime-manifest.schema.json +156 -0
- package/docs/schemas/inference-service-record.schema.json +227 -0
- package/docs/schemas/local-inference-selection.schema.json +98 -0
- package/docs/schemas/local-model-manifest.schema.json +43 -0
- package/docs/schemas/model-node-binding.schema.json +34 -0
- package/docs/schemas/model-node-generation-release.schema.json +46 -0
- package/docs/schemas/ordered-eval-control.schema.json +39 -0
- package/docs/schemas/remote-execution-lease-grant.schema.json +45 -0
- package/docs/schemas/remote-harbor-work-spec.schema.json +325 -0
- package/docs/schemas/remote-model-binding.schema.json +21 -0
- package/docs/schemas/remote-physical-execution.schema.json +108 -0
- package/docs/schemas/remote-result-envelope.schema.json +9 -3
- package/docs/schemas/remote-tree-envelope.schema.json +17 -3
- package/docs/schemas/remote-verifier-result.schema.json +203 -0
- package/docs/schemas/remote-verifier-source.schema.json +52 -0
- package/docs/schemas/remote-verifier-work.schema.json +125 -0
- package/docs/schemas/remote-work-offer.schema.json +170 -34
- package/docs/schemas/remote-worker-execution-admission.schema.json +214 -0
- package/docs/schemas/remote-worker-generation-cleanup.schema.json +199 -0
- package/docs/schemas/remote-worker-registration.schema.json +226 -25
- package/docs/schemas/run-manifest.schema.json +211 -44
- package/docs/schemas/run-request.schema.json +2 -1
- package/docs/schemas/training-external-binding.schema.json +26 -0
- package/integrations/harbor/hitch_harbor_agent.py +241 -13
- package/integrations/harbor/hitch_harbor_environment.py +22 -1
- package/integrations/harbor/hitch_host_credentials.py +354 -0
- package/integrations/harbor/hitch_private_exec.py +150 -0
- package/integrations/model-call/cli.js +13 -5
- package/integrations/training-tool/cli.js +72 -0
- package/package.json +8 -2
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"node-client.js","sourceRoot":"","sources":["../../../src/inference/node-client.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,KAAK,EAAE,MAAM,oBAAoB,CAAC;AAC3C,OAAO,EAAE,UAAU,EAAE,MAAM,aAAa,CAAC;AACzC,OAAO,EAAE,gBAAgB,EAAE,MAAM,SAAS,CAAC;AAC3C,OAAO,IAAI,MAAM,WAAW,CAAC;AAC7B,OAAO,EAAE,KAAK,EAAE,KAAK,EAAE,MAAM,kBAAkB,CAAC;AAChD,OAAO,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AACvC,OAAO,EAAE,QAAQ,EAAE,MAAM,sBAAsB,CAAC;AAChD,OAAO,EAAE,UAAU,EAAE,eAAe,EAAE,WAAW,EAAE,QAAQ,EAAE,UAAU,EAAE,UAAU,EAAE,UAAU,EAAE,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAE9I,OAAO,EAAE,gBAAgB,EAAE,MAAM,kBAAkB,CAAC;AACpD,OAAO,EAAE,iBAAiB,EAAE,MAAM,qBAAqB,CAAC;AACxD,OAAO,EAAE,uBAAuB,EAAE,MAAM,eAAe,CAAC;AAoBxD,MAAM,KAAK,GAAG,CAAC,KAAa,EAAU,EAAE,CAAC,IAAI,KAAK,CAAC,UAAU,CAAC,GAAG,EAAE,OAAO,CAAC,GAAG,CAAC;AAE/E,iFAAiF;AACjF,MAAM,OAAO,yBAAyB;IACf;IAA8C;IAA2C;IAA9G,YAAqB,UAAmC,EAAW,IAAgC,EAAW,MAAyB,OAAO,CAAC,GAAG;QAA7H,eAAU,GAAV,UAAU,CAAyB;QAAW,SAAI,GAAJ,IAAI,CAA4B;QAAW,QAAG,GAAH,GAAG,CAAiC;QAChJ,IAAI,CAAC,UAAU,CAAC,MAAM,CAAC,MAAM,IAAI,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,OAAO,GAAG,KAAK,QAAQ,IAAI,CAAC,GAAG,IAAI,GAAG,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;eAChH,CAAC,IAAI,CAAC,KAAK,CAAC,UAAU,CAAC,UAAU,CAAC,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,IAAI,CAAC,IAAI,CAAC,UAAU;eACjF,CAAC,UAAU,CAAC,OAAO,CAAC,SAAS,EAAE,UAAU,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,IAAI,CAAC,IAAI,IAAI,GAAG,CAAC,IAAI,IAAI,GAAG,KAAK,CAAC;eAC7H,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,KAAK,KAAK,IAAI,CAAC,oCAAoC,CAAC,IAAI,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,EAAE,CAAC;YACpH,MAAM,IAAI,SAAS,CAAC,wCAAwC,CAAC,CAAC;QAChE,CAAC;IACH,CAAC;IAED,OAAO,CAAC,MAA4B;QAClC,MAAM,OAAO,GAAG,CAAC,GAAG,IAAI,CAAC,UAAU,CAAC,MAAM,EAAE,IAAI,EAAE,oBAAoB,EAAE,MAAM,EAAE,UAAU,EAAE,IAAI,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;QACxH,OAAO,IAAI,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,KAAK,OAAO,CAAC,CAAC,CAAC,OAAO;YACzD,CAAC,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,IAAI,EAAE,eAAe,EAAE,IAAI,EAAE,mBAAmB,EAAE,IAAI,EAAE,IAAI,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,EAAE,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC;IAC1I,CAAC;IAEO,QAAQ,CAAC,SAAiB,EAAE,OAAgB;QAClD,OAAO,EAAE,aAAa,EAAE,CAAC,EAAE,SAAS,EAAE,UAAU,EAAE,EAAE,IAAI,EAAE,IAAI,CAAC,IAAI,EAAE,SAAS,EAAE,WAAW,EAAE,UAAU,CAAC,OAAO,CAAC,EAAE,OAAO,EAAE,CAAC;IAC9H,CAAC;IAEO,QAAQ,CAAC,KAAc,EAAE,QAAkB;QACjD,MAAM,IAAI,GAAG,KAAwD,CAAC;QACtE,IAAI,CAAC,IAAI,IAAI,IAAI,CAAC,aAAa,KAAK,CAAC,IAAI,IAAI,CAAC,SAAS,KAAK,QAAQ,CAAC,SAAS,IAAI,IAAI,CAAC,WAAW,KAAK,QAAQ,CAAC,WAAW;eACtH,UAAU,CAAC,IAAI,CAAC,IAAI,IAAI,IAAI,CAAC,KAAK,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,QAAQ,IAAI,IAAI,CAAC,EAAE,CAAC;YACpF,MAAM,IAAI,UAAU,CAAC,iDAAiD,EAAE,EAAE,IAAI,EAAE,4BAA4B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;QAChI,CAAC;QACD,OAAO,IAAI,CAAC,MAAM,CAAC;IACrB,CAAC;IAEO,KAAK,CAAC,QAAQ,CAAC,MAA4B,EAAE,QAAkB,EAAE,IAAa;QACpF,MAAM,OAAO,GAAG,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;QACrC,MAAM,KAAK,GAAG,KAAK,CAAC,OAAO,CAAC,CAAC,CAAE,EAAE,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,EAAE,GAAG,EAAE,IAAI,CAAC,GAAG,EAAE,KAAK,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC,CAAC;QACvG,KAAK,CAAC,MAAM,CAAC,MAAM,EAAE,CAAC;QAAC,KAAK,CAAC,KAAK,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;QACzD,MAAM,KAAK,GAAG,UAAU,CAAC,GAAG,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC;QAClF,MAAM,SAAS,GAAG,IAAI,OAAO,CAAgB,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;YAC/D,KAAK,CAAC,IAAI,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC;YAC5B,KAAK,CAAC,IAAI,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;QAC/B,CAAC,CAAC,CAAC;QACH,KAAK,SAAS,CAAC,KAAK,CAAC,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;QAC/B,IAAI,MAAM,GAAG,EAAE,CAAC;QAAC,IAAI,MAAM,GAAG,CAAC,CAAC;QAChC,MAAM,IAAI,GAAG,KAAK,IAAI,EAAE;YACtB,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,KAAK,CAAC,MAAM,EAAE,CAAC;gBACvC,MAAM,IAAK,KAAgB,CAAC,UAAU,CAAC;gBACvC,IAAI,MAAM,GAAG,CAAC,GAAG,IAAI,GAAG,IAAI;oBAAE,MAAM,IAAI,SAAS,CAAC,oDAAoD,CAAC,CAAC;gBACxG,MAAM,IAAI,KAAK,CAAC,QAAQ,EAAE,CAAC;YAC7B,CAAC;QACH,CAAC,CAAC;QACF,KAAK,SAAS,CAAC,CAAC,KAAK,KAAK,MAAM,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,GAAG,IAAI,CAAC,CAAC,CAAC,IAAI,IAAI;YAAE,KAAK,CAAC,CAAC,gBAAgB,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;QACxH,IAAI,CAAC;YACH,MAAM,CAAC,EAAE,AAAD,EAAG,IAAI,CAAC,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,QAAQ,CAAC,IAAI,CAAC,KAAK,EAAE,CAAC,EAAE,KAAK,CAAC,KAAK,CAAC,EAAE,IAAI,EAAE,EAAE,SAAS,CAAC,CAAC,CAAC;YACzG,IAAI,IAAI,KAAK,CAAC,EAAE,CAAC;gBACf,IAAI,MAAM,GAAG,kBAAkB,CAAC;gBAChC,IAAI,CAAC;oBAAC,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;oBAAC,IAAI,0BAA0B,CAAC,IAAI,CAAC,MAAM,CAAC,KAAK,EAAE,IAAI,CAAC;wBAAE,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC;gBAAC,CAAC;gBAAC,MAAM,CAAC,CAAA,CAAC;gBACxI,MAAM,IAAI,UAAU,CAAC,cAAc,QAAQ,CAAC,SAAS,YAAY,MAAM,0CAA0C,EAAE,EAAE,IAAI,EAAE,6BAA6B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;YAC5K,CAAC;YACD,OAAO,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,EAAE,QAAQ,CAAC,CAAC;QACrD,CAAC;gBAAS,CAAC;YAAC,YAAY,CAAC,KAAK,CAAC,CAAC;YAAC,IAAI,KAAK,CAAC,QAAQ,KAAK,IAAI;gBAAE,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;QAAC,CAAC;IACxF,CAAC;IAED,KAAK,CAAC,IAAI,CAAC,SAAiB,EAAE,OAAgB,IAAsB,OAAO,IAAI,CAAC,QAAQ,CAAC,KAAK,EAAE,IAAI,CAAC,QAAQ,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC;IAC7H,OAAO,CAAC,KAAwB;QACtC,OAAO,EAAE,SAAS,EAAE,KAAK,CAAC,SAAS,EAAE,OAAO,EAAE,WAAW,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO,EAAE,IAAI,EAAE,KAAK,CAAC,IAAI,EAAE,CAAC;IACxI,CAAC;IACD,KAAK,CAAC,OAAO,CAAC,KAAwB;QACpC,MAAM,WAAW,GAAG,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,MAAM,iBAAiB,CAAC,KAAK,CAAC,IAAI,EAAE,KAAK,CAAC,KAAK,EAAE,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC;QAC3H,IAAI,WAAW,EAAE,CAAC;YAChB,MAAM,QAAQ,GAAG,uBAAuB,CAAC,MAAM,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,EAAE,WAAW,EAAE,CAAC,CAAC,CAAC;YAC7F,IAAI,QAAQ,CAAC,QAAQ,KAAK,KAAK,CAAC,KAAK,CAAC,QAAQ;gBAAE,MAAM,IAAI,SAAS,CAAC,sDAAsD,CAAC,CAAC;YAC5H,MAAM,IAAI,CAAC,IAAI,CAAC,mBAAmB,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC;YAC1D,OAAO;QACT,CAAC;QACD,MAAM,gBAAgB,CAAC,KAAK,CAAC,IAAI,EAAE,KAAK,CAAC,KAAK,CAAC,CAAC;QAChD,KAAK,MAAM,IAAI,IAAI,KAAK,CAAC,KAAK,CAAC,KAAK,EAAE,CAAC;YACrC,MAAM,KAAK,GAAG,MAAM,IAAI,CAAC,IAAI,CAAC,UAAU,EAAE,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM,EAAE,CAAwC,CAAC;YAC1G,IAAI,KAAK,CAAC,OAAO,IAAI,KAAK,CAAC,IAAI,KAAK,IAAI,CAAC,IAAI;gBAAE,SAAS;YACxD,IAAI,KAAK,CAAC,OAAO;gBAAE,MAAM,IAAI,SAAS,CAAC,wDAAwD,CAAC,CAAC;YACjG,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,QAAQ,CAAC,YAAY,EAAE,IAAI,CAAC,QAAQ,CAAC,YAAY,EAAE,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM,EAAE,IAAI,EAAE,IAAI,CAAC,IAAI,EAAE,CAAC,EACpH,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,UAAU,EAAE,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAuC,CAAC;YAC5G,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,MAAM,CAAC,IAAI,KAAK,IAAI,CAAC,IAAI;gBAAE,MAAM,IAAI,SAAS,CAAC,+CAA+C,CAAC,CAAC;QACzH,CAAC;QACD,MAAM,IAAI,CAAC,IAAI,CAAC,mBAAmB,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC;IAC5D,CAAC;IACD,KAAK,CAAC,KAAwB,IAAsB,OAAO,IAAI,CAAC,IAAI,CAAC,iBAAiB,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC;IAC/G,OAAO,CAAC,KAAyB,IAAsB,OAAO,IAAI,CAAC,IAAI,CAAC,mBAAmB,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC;IACtG,MAAM,CAAC,KAAwB;QAC7B,MAAM,QAAQ,GAAG,EAAE,SAAS,EAAE,KAAK,CAAC,MAAM,CAAC,UAAU,EAAE,OAAO,EAAE,WAAW,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO,EAAE,IAAI,EAAE,KAAK,CAAC,IAAI,EAAE,CAAC;QACxJ,OAAO,IAAI,CAAC,IAAI,CAAC,kBAAkB,EAAE,EAAE,SAAS,EAAE,QAAQ,CAAC,SAAS,EAAE,OAAO,EAAE,QAAQ,CAAC,OAAO;YAC7F,WAAW,EAAE,KAAK,CAAC,IAAI,CAAC,YAAY,EAAE,WAAW,EAAE,UAAU,CAAC,QAAQ,CAAC,EAAE,cAAc,EAAE,KAAK,CAAC,MAAM,CAAC,cAAc,EAAE,CAAC,CAAC;IAC5H,CAAC;IACD,IAAI,CAAC,KAAyB,IAAsB,OAAO,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC;IAEhG,KAAK,CAAC,KAAK,CAAC,IAAY,EAAE,QAAgB;QACxC,MAAM,EAAE,OAAO,EAAE,SAAS,EAAE,GAAG,IAAI,CAAC,UAAU,CAAC;QAC/C,IAAI,QAAQ,KAAK,OAAO,CAAC,QAAQ,IAAI,CAAC,SAAS,CAAC,IAAI,KAAK,OAAO,IAAI,OAAO,CAAC,SAAS,KAAK,OAAO,CAAC,QAAQ,CAAC,EAAE,CAAC;YAC5G,MAAM,IAAI,SAAS,CAAC,2DAA2D,CAAC,CAAC;QACnF,CAAC;QACD,IAAI,SAAS,CAAC,IAAI,KAAK,KAAK,EAAE,CAAC;YAC7B,MAAM,SAAS,GAAG,UAAU,CAAC,IAAI,CAAC,CAAC,uBAAuB,CAAC;YAC3D,MAAM,QAAQ,GAAG,UAAU,CAAC,EAAE,IAAI,EAAE,IAAI,CAAC,IAAI,EAAE,UAAU,EAAE,IAAI,CAAC,UAAU,EAAE,CAAC,CAAC;YAC9E,wEAAwE;YACxE,MAAM,eAAe,GAAG,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,cAAc,OAAO,CAAC,MAAM,EAAE,EAAE,IAAI,MAAM,IAAI,WAAW,CAAC,IAAI,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC;YACxH,MAAM,KAAK,CAAC,eAAe,EAAE,EAAE,IAAI,EAAE,KAAK,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;YAC/D,MAAM,WAAW,GAAG,MAAM,KAAK,CAAC,eAAe,CAAC,CAAC;YACjD,IAAI,CAAC,WAAW,CAAC,WAAW,EAAE,IAAI,WAAW,CAAC,cAAc,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI,GAAG,KAAK,CAAC,KAAK,CAAC;mBAC7F,CAAC,OAAO,CAAC,MAAM,IAAI,WAAW,CAAC,GAAG,KAAK,OAAO,CAAC,MAAM,EAAE,CAAC;gBAAE,MAAM,IAAI,SAAS,CAAC,gDAAgD,CAAC,CAAC;YACrI,MAAM,MAAM,GAAG,IAAI,CAAC,IAAI,CAAC,eAAe,EAAE,GAAG,QAAQ,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,OAAO,CAAC,CAAC;YACzE,MAAM,UAAU,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,QAAQ,CAAC,KAAK,CAAC,CAAC,CAAC,GAAG,cAAc,CAAC,CAAC;YAC5E,MAAM,YAAY,CAAC,SAAS,EAAE,QAAQ,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,KAAK,IAAI,EAAE;gBAC1D,IAAI,KAAyC,CAAC;gBAC9C,IAAI,CAAC;oBAAC,KAAK,GAAG,MAAM,QAAQ,CAAC,UAAU,CAAC,CAAC;gBAAC,CAAC;gBAAC,OAAO,KAAK,EAAE,CAAC;oBAAC,IAAK,KAA+B,CAAC,IAAI,KAAK,QAAQ;wBAAE,MAAM,KAAK,CAAC;gBAAC,CAAC;gBAClI,IAAI,KAAK,IAAI,KAAK,CAAC,QAAQ,KAAK,QAAQ;oBAAE,MAAM,IAAI,SAAS,CAAC,qCAAqC,CAAC,CAAC;gBACrG,MAAM,MAAM,GAAG,CAAC,IAAc,EAAE,EAAE,CAAC,UAAU,CAAC,KAAK,EAAE,IAAI,EAAE,EAAE,GAAG,EAAE,IAAI,CAAC,GAAG,EAAE,SAAS,EAAE,MAAM,EAAE,WAAW,EAAE,6BAA6B,EAAE,eAAe,EAAE,EAAE,EAAE,CAAC,CAAC;gBAClK,MAAM,MAAM,GAAG,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,eAAe,EAAE,IAAI,EAAE,mBAAmB,CAAC,CAAC;gBAChF,IAAI,KAAK,EAAE,CAAC;oBAAC,IAAI,CAAC;wBAAC,MAAM,MAAM,CAAC,CAAC,GAAG,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC;wBAAC,OAAO;oBAAC,CAAC;oBAAC,MAAM,CAAC,CAAA,CAAC;gBAAC,CAAC;gBACvG,MAAM,eAAe,CAAC,UAAU,EAAE,EAAE,QAAQ,EAAE,IAAI,EAAE,IAAI,CAAC,IAAI,EAAE,UAAU,EAAE,IAAI,CAAC,UAAU,EAAE,CAAC,CAAC;gBAC9F,MAAM,MAAM,CAAC,CAAC,GAAG,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,0BAA0B,EAAE,IAAI,EAAE,wBAAwB,EAAE,IAAI,EAAE,uBAAuB;oBACpI,IAAI,EAAE,aAAa,OAAO,CAAC,SAAS,cAAc,OAAO,CAAC,QAAQ,EAAE,EAAE,IAAI,EAAE,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC;YACjG,CAAC,CAAC,CAAC;QACL,CAAC;QACD,OAAO,oBAAoB,OAAO,CAAC,SAAS,EAAE,CAAC;IACjD,CAAC;CACF"}
|
|
@@ -0,0 +1,79 @@
|
|
|
1
|
+
import path from "node:path";
|
|
2
|
+
import { HitchError, atomicWriteJSON, readJSON, sha256JSON, statePaths, withFileLock } from "../foundation/index.js";
|
|
3
|
+
import { parseInferenceRuntimeManifest, parseModelNodeBinding } from "./manifest.js";
|
|
4
|
+
import { PythonInferenceNodeClient } from "./node-client.js";
|
|
5
|
+
export function parseModelNodeRegistration(value) {
|
|
6
|
+
const record = exact(value, ["schema_version", "binding", "connection"]);
|
|
7
|
+
if (record.schema_version !== "2")
|
|
8
|
+
throw new TypeError("model node registration requires schema version 2");
|
|
9
|
+
const binding = parseModelNodeBinding(record.binding);
|
|
10
|
+
const connection = exact(record.connection, ["transport", "python", "configPath", "gateway"]);
|
|
11
|
+
const transport = exact(connection.transport, ["type", "host"]);
|
|
12
|
+
if (transport.type !== "local" && transport.type !== "ssh" || transport.type === "local" && transport.host !== undefined)
|
|
13
|
+
throw new TypeError("invalid model node transport");
|
|
14
|
+
const gateway = exact(connection.gateway, ["localPort", "nodePort"]);
|
|
15
|
+
if (!Array.isArray(connection.python) || typeof connection.configPath !== "string")
|
|
16
|
+
throw new TypeError("invalid model node connection");
|
|
17
|
+
const result = { schema_version: "2", binding, connection: { transport, python: connection.python, configPath: connection.configPath, gateway } };
|
|
18
|
+
// Shared command/port/SSH validation, without launching a process.
|
|
19
|
+
new PythonInferenceNodeClient(result.connection, { nodeId: binding.node_id, generation: binding.generation });
|
|
20
|
+
if (transport.type === "local" && gateway.localPort !== gateway.nodePort)
|
|
21
|
+
throw new TypeError("local model node route ports must match");
|
|
22
|
+
return result;
|
|
23
|
+
}
|
|
24
|
+
export async function observeModelNode(client, binding) {
|
|
25
|
+
const value = await client.call("probe", {});
|
|
26
|
+
if (!value || value.nodeId !== binding.node_id || value.generation !== binding.generation
|
|
27
|
+
|| value.runtimeDigest !== binding.runtime_digest || sha256JSON(value.runtime) !== binding.runtime_digest
|
|
28
|
+
|| !Array.isArray(value.gpuUuids) || value.gpuUuids.some(uuid => typeof uuid !== "string")
|
|
29
|
+
|| !Array.isArray(value.launchers) || !value.launchers.includes("process")
|
|
30
|
+
|| value.capabilities?.binaryCas !== true || value.capabilities?.durableInferenceProcesses !== true) {
|
|
31
|
+
throw new HitchError("model-node generation, Python runtime or process capability changed", { code: "inference_runtime_mismatch", exitCode: 12 });
|
|
32
|
+
}
|
|
33
|
+
return value;
|
|
34
|
+
}
|
|
35
|
+
export function runtimeFromModelNode(observation) {
|
|
36
|
+
const packages = observation.runtime.packages.filter(item => item.name === "sglang");
|
|
37
|
+
if (packages.length !== 1)
|
|
38
|
+
throw new TypeError("model node must observe exactly one SGLang distribution");
|
|
39
|
+
const packageInfo = packages[0];
|
|
40
|
+
const body = { schema_version: "2", engine: "sglang", sglang_version: packageInfo.version, sglang_commit: packageInfo.commit,
|
|
41
|
+
backend: "cuda", package: { kind: "python-env", environment_digest: observation.runtimeDigest,
|
|
42
|
+
python_version: observation.runtime.pythonVersion, packages_digest: observation.runtime.packagesDigest }, compatibility_profile: "gear-model-node-v2" };
|
|
43
|
+
return parseInferenceRuntimeManifest({ ...body, runtime_id: sha256JSON(body) });
|
|
44
|
+
}
|
|
45
|
+
export async function registerModelNode(root, value, env = process.env) {
|
|
46
|
+
const registration = parseModelNodeRegistration(value);
|
|
47
|
+
const { binding, connection } = registration;
|
|
48
|
+
const client = new PythonInferenceNodeClient(connection, { nodeId: binding.node_id, generation: binding.generation }, env);
|
|
49
|
+
const observation = await observeModelNode(client, binding);
|
|
50
|
+
const runtime = runtimeFromModelNode(observation);
|
|
51
|
+
await withFileLock(statePaths(root).inferenceOperationLocks, sha256JSON(binding), async () => {
|
|
52
|
+
// Explicit re-registration may repair connection coordinates after probing
|
|
53
|
+
// the same frozen identity. Old generation records remain recoverable.
|
|
54
|
+
await atomicWriteJSON(registrationPath(root, binding), registration);
|
|
55
|
+
await atomicWriteJSON(path.join(statePaths(root).inferenceRuntimes, runtime.runtime_id.slice(7), "manifest.json"), runtime);
|
|
56
|
+
});
|
|
57
|
+
return { schema_version: "2", binding, runtime, gpu_uuids: observation.gpuUuids };
|
|
58
|
+
}
|
|
59
|
+
export async function loadModelNodeClient(root, binding, env = process.env) {
|
|
60
|
+
const record = parseModelNodeRegistration(await readJSON(registrationPath(root, parseModelNodeBinding(binding))));
|
|
61
|
+
if (sha256JSON(record.binding) !== sha256JSON(binding))
|
|
62
|
+
throw new TypeError("model node registration identity changed");
|
|
63
|
+
return new PythonInferenceNodeClient(record.connection, { nodeId: binding.node_id, generation: binding.generation }, env);
|
|
64
|
+
}
|
|
65
|
+
function registrationPath(root, binding) {
|
|
66
|
+
return path.join(statePaths(root).inferenceNodes, sha256JSON(binding).slice(7) + ".json");
|
|
67
|
+
}
|
|
68
|
+
function exact(value, keys) {
|
|
69
|
+
if (!value || typeof value !== "object" || Array.isArray(value) || Object.keys(value).some(key => !keys.includes(key)))
|
|
70
|
+
throw new TypeError("invalid model node registration fields");
|
|
71
|
+
return value;
|
|
72
|
+
}
|
|
73
|
+
export async function readModelNodeRegistration(file) {
|
|
74
|
+
return parseModelNodeRegistration(await readJSON(file));
|
|
75
|
+
}
|
|
76
|
+
export async function readModelNodeBinding(file) {
|
|
77
|
+
return parseModelNodeBinding(await readJSON(file));
|
|
78
|
+
}
|
|
79
|
+
//# sourceMappingURL=node-registry.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"node-registry.js","sourceRoot":"","sources":["../../../src/inference/node-registry.ts"],"names":[],"mappings":"AAAA,OAAO,IAAI,MAAM,WAAW,CAAC;AAE7B,OAAO,EAAE,UAAU,EAAE,eAAe,EAAE,QAAQ,EAAE,UAAU,EAAE,UAAU,EAAE,YAAY,EAAE,MAAM,wBAAwB,CAAC;AACrH,OAAO,EAAE,6BAA6B,EAAE,qBAAqB,EAAE,MAAM,eAAe,CAAC;AACrF,OAAO,EAAE,yBAAyB,EAAE,MAAM,kBAAkB,CAAC;AAe7D,MAAM,UAAU,0BAA0B,CAAC,KAAc;IACvD,MAAM,MAAM,GAAG,KAAK,CAAC,KAAK,EAAE,CAAC,gBAAgB,EAAE,SAAS,EAAE,YAAY,CAAC,CAAC,CAAC;IACzE,IAAI,MAAM,CAAC,cAAc,KAAK,GAAG;QAAE,MAAM,IAAI,SAAS,CAAC,mDAAmD,CAAC,CAAC;IAC5G,MAAM,OAAO,GAAG,qBAAqB,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC;IACtD,MAAM,UAAU,GAAG,KAAK,CAAC,MAAM,CAAC,UAAU,EAAE,CAAC,WAAW,EAAE,QAAQ,EAAE,YAAY,EAAE,SAAS,CAAC,CAAC,CAAC;IAC9F,MAAM,SAAS,GAAG,KAAK,CAAC,UAAU,CAAC,SAAS,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,CAAC;IAChE,IAAI,SAAS,CAAC,IAAI,KAAK,OAAO,IAAI,SAAS,CAAC,IAAI,KAAK,KAAK,IAAI,SAAS,CAAC,IAAI,KAAK,OAAO,IAAI,SAAS,CAAC,IAAI,KAAK,SAAS;QAAE,MAAM,IAAI,SAAS,CAAC,8BAA8B,CAAC,CAAC;IAC9K,MAAM,OAAO,GAAG,KAAK,CAAC,UAAU,CAAC,OAAO,EAAE,CAAC,WAAW,EAAE,UAAU,CAAC,CAAC,CAAC;IACrE,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,OAAO,UAAU,CAAC,UAAU,KAAK,QAAQ;QAAE,MAAM,IAAI,SAAS,CAAC,+BAA+B,CAAC,CAAC;IACzI,MAAM,MAAM,GAAG,EAAE,cAAc,EAAE,GAAG,EAAE,OAAO,EAAE,UAAU,EAAE,EAAE,SAAS,EAAE,MAAM,EAAE,UAAU,CAAC,MAAM,EAAE,UAAU,EAAE,UAAU,CAAC,UAAU,EAAE,OAAO,EAAE,EAA6B,CAAC;IAC7K,mEAAmE;IACnE,IAAI,yBAAyB,CAAC,MAAM,CAAC,UAAU,EAAE,EAAE,MAAM,EAAE,OAAO,CAAC,OAAO,EAAE,UAAU,EAAE,OAAO,CAAC,UAAU,EAAE,CAAC,CAAC;IAC9G,IAAI,SAAS,CAAC,IAAI,KAAK,OAAO,IAAI,OAAO,CAAC,SAAS,KAAK,OAAO,CAAC,QAAQ;QAAE,MAAM,IAAI,SAAS,CAAC,yCAAyC,CAAC,CAAC;IACzI,OAAO,MAAM,CAAC;AAChB,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,gBAAgB,CAAC,MAAiC,EAAE,OAA2B;IACnG,MAAM,KAAK,GAAG,MAAM,MAAM,CAAC,IAAI,CAAC,OAAO,EAAE,EAAE,CAA2B,CAAC;IACvE,IAAI,CAAC,KAAK,IAAI,KAAK,CAAC,MAAM,KAAK,OAAO,CAAC,OAAO,IAAI,KAAK,CAAC,UAAU,KAAK,OAAO,CAAC,UAAU;WACpF,KAAK,CAAC,aAAa,KAAK,OAAO,CAAC,cAAc,IAAI,UAAU,CAAC,KAAK,CAAC,OAAO,CAAC,KAAK,OAAO,CAAC,cAAc;WACtG,CAAC,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,QAAQ,CAAC,IAAI,KAAK,CAAC,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,CAAC;WACvF,CAAC,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,SAAS,CAAC,IAAI,CAAC,KAAK,CAAC,SAAS,CAAC,QAAQ,CAAC,SAAS,CAAC;WACvE,KAAK,CAAC,YAAY,EAAE,SAAS,KAAK,IAAI,IAAI,KAAK,CAAC,YAAY,EAAE,yBAAyB,KAAK,IAAI,EAAE,CAAC;QACtG,MAAM,IAAI,UAAU,CAAC,qEAAqE,EAAE,EAAE,IAAI,EAAE,4BAA4B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;IACpJ,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC;AAED,MAAM,UAAU,oBAAoB,CAAC,WAAmC;IACtE,MAAM,QAAQ,GAAG,WAAW,CAAC,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC;IACrF,IAAI,QAAQ,CAAC,MAAM,KAAK,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,yDAAyD,CAAC,CAAC;IAC1G,MAAM,WAAW,GAAG,QAAQ,CAAC,CAAC,CAAE,CAAC;IACjC,MAAM,IAAI,GAAG,EAAE,cAAc,EAAE,GAAY,EAAE,MAAM,EAAE,QAAiB,EAAE,cAAc,EAAE,WAAW,CAAC,OAAO,EAAE,aAAa,EAAE,WAAW,CAAC,MAAM;QAC5I,OAAO,EAAE,MAAe,EAAE,OAAO,EAAE,EAAE,IAAI,EAAE,YAAqB,EAAE,kBAAkB,EAAE,WAAW,CAAC,aAAa;YAC7G,cAAc,EAAE,WAAW,CAAC,OAAO,CAAC,aAAa,EAAE,eAAe,EAAE,WAAW,CAAC,OAAO,CAAC,cAAc,EAAE,EAAE,qBAAqB,EAAE,oBAAoB,EAAE,CAAC;IAC5J,OAAO,6BAA6B,CAAC,EAAE,GAAG,IAAI,EAAE,UAAU,EAAE,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;AAClF,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,iBAAiB,CAAC,IAAY,EAAE,KAAc,EAAE,MAAyB,OAAO,CAAC,GAAG;IACxG,MAAM,YAAY,GAAG,0BAA0B,CAAC,KAAK,CAAC,CAAC;IACvD,MAAM,EAAE,OAAO,EAAE,UAAU,EAAE,GAAG,YAAY,CAAC;IAC7C,MAAM,MAAM,GAAG,IAAI,yBAAyB,CAAC,UAAU,EAAE,EAAE,MAAM,EAAE,OAAO,CAAC,OAAO,EAAE,UAAU,EAAE,OAAO,CAAC,UAAU,EAAE,EAAE,GAAG,CAAC,CAAC;IAC3H,MAAM,WAAW,GAAG,MAAM,gBAAgB,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC5D,MAAM,OAAO,GAAG,oBAAoB,CAAC,WAAW,CAAC,CAAC;IAClD,MAAM,YAAY,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC,uBAAuB,EAAE,UAAU,CAAC,OAAO,CAAC,EAAE,KAAK,IAAI,EAAE;QAC3F,2EAA2E;QAC3E,uEAAuE;QACvE,MAAM,eAAe,CAAC,gBAAgB,CAAC,IAAI,EAAE,OAAO,CAAC,EAAE,YAAY,CAAC,CAAC;QACrE,MAAM,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC,iBAAiB,EAAE,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,eAAe,CAAC,EAAE,OAAO,CAAC,CAAC;IAC9H,CAAC,CAAC,CAAC;IACH,OAAO,EAAE,cAAc,EAAE,GAAG,EAAE,OAAO,EAAE,OAAO,EAAE,SAAS,EAAE,WAAW,CAAC,QAAQ,EAAE,CAAC;AACpF,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,mBAAmB,CAAC,IAAY,EAAE,OAA2B,EAAE,MAAyB,OAAO,CAAC,GAAG;IACvH,MAAM,MAAM,GAAG,0BAA0B,CAAC,MAAM,QAAQ,CAAC,gBAAgB,CAAC,IAAI,EAAE,qBAAqB,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC;IAClH,IAAI,UAAU,CAAC,MAAM,CAAC,OAAO,CAAC,KAAK,UAAU,CAAC,OAAO,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,0CAA0C,CAAC,CAAC;IACxH,OAAO,IAAI,yBAAyB,CAAC,MAAM,CAAC,UAAU,EAAE,EAAE,MAAM,EAAE,OAAO,CAAC,OAAO,EAAE,UAAU,EAAE,OAAO,CAAC,UAAU,EAAE,EAAE,GAAG,CAAC,CAAC;AAC5H,CAAC;AAED,SAAS,gBAAgB,CAAC,IAAY,EAAE,OAA2B;IACjE,OAAO,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC,cAAc,EAAE,UAAU,CAAC,OAAO,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,GAAG,OAAO,CAAC,CAAC;AAC5F,CAAC;AACD,SAAS,KAAK,CAAC,KAAc,EAAE,IAAc;IAC3C,IAAI,CAAC,KAAK,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,wCAAwC,CAAC,CAAC;IACtL,OAAO,KAAgC,CAAC;AAC1C,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,yBAAyB,CAAC,IAAY;IAC1D,OAAO,0BAA0B,CAAC,MAAM,QAAQ,CAAC,IAAI,CAAC,CAAC,CAAC;AAC1D,CAAC;AACD,MAAM,CAAC,KAAK,UAAU,oBAAoB,CAAC,IAAY;IACrD,OAAO,qBAAqB,CAAC,MAAM,QAAQ,CAAC,IAAI,CAAC,CAAC,CAAC;AACrD,CAAC"}
|
|
@@ -0,0 +1,79 @@
|
|
|
1
|
+
import { HitchError } from "../foundation/index.js";
|
|
2
|
+
/** /server_info contains credentials. Persist only this explicit projection. */
|
|
3
|
+
export function validateRuntimeObservation(value, imageId, gpuUuid, lock, runtime) {
|
|
4
|
+
const { info, tokens } = validateEngineInfo(value, gpuUuid, lock, runtime);
|
|
5
|
+
if (runtime.package.kind !== "oci" || !/^sha256:[a-f0-9]{64}$/.test(imageId))
|
|
6
|
+
throw mismatch("container image ID is missing or runtime is not OCI");
|
|
7
|
+
return {
|
|
8
|
+
schema_version: "1", ...observationFields(info, tokens, gpuUuid, lock), container_image_id: imageId,
|
|
9
|
+
};
|
|
10
|
+
}
|
|
11
|
+
/** Process observations record the node environment; no fictitious inner OCI image. */
|
|
12
|
+
export function validateProcessRuntimeObservation(value, observed, gpuUuid, lock, runtime, expectedNode) {
|
|
13
|
+
if (runtime.schema_version !== "2" || runtime.package.kind !== "python-env" || observed.kind !== "python-env"
|
|
14
|
+
|| observed.node_id !== expectedNode.node_id || observed.generation !== expectedNode.generation
|
|
15
|
+
|| observed.environment_digest !== runtime.package.environment_digest || observed.python_version !== runtime.package.python_version
|
|
16
|
+
|| observed.packages_digest !== runtime.package.packages_digest
|
|
17
|
+
|| (observed.outer_image_digest !== null && !/^sha256:[a-f0-9]{64}$/.test(observed.outer_image_digest)))
|
|
18
|
+
throw mismatch("model-node Python environment or generation changed");
|
|
19
|
+
const { info, tokens } = validateEngineInfo(value, gpuUuid, lock, runtime);
|
|
20
|
+
return { schema_version: "2", ...observationFields(info, tokens, gpuUuid, lock), runtime: {
|
|
21
|
+
kind: "python-env", node_id: observed.node_id, generation: observed.generation,
|
|
22
|
+
environment_digest: observed.environment_digest, python_version: observed.python_version,
|
|
23
|
+
packages_digest: observed.packages_digest, outer_image_digest: observed.outer_image_digest,
|
|
24
|
+
} };
|
|
25
|
+
}
|
|
26
|
+
function validateEngineInfo(value, gpuUuid, lock, runtime) {
|
|
27
|
+
if (!value || typeof value !== "object" || Array.isArray(value))
|
|
28
|
+
throw mismatch("invalid /server_info response");
|
|
29
|
+
const info = value;
|
|
30
|
+
const expected = {
|
|
31
|
+
version: runtime.sglang_version, device: lock.execution.platform.backend,
|
|
32
|
+
dtype: lock.execution.dtype, kv_cache_dtype: lock.execution.kv_cache_dtype,
|
|
33
|
+
attention_backend: lock.execution.attention_backend, sampling_backend: lock.execution.sampling_backend,
|
|
34
|
+
context_length: lock.execution.context_tokens_per_request, max_running_requests: lock.execution.max_running_requests,
|
|
35
|
+
tp_size: 1, dp_size: 1, pp_size: 1,
|
|
36
|
+
disable_radix_cache: lock.execution.prefix_cache.mode === "disabled",
|
|
37
|
+
disable_overlap_schedule: !lock.execution.platform.overlap_schedule,
|
|
38
|
+
};
|
|
39
|
+
for (const [field, wanted] of Object.entries(expected)) {
|
|
40
|
+
const actual = info[field];
|
|
41
|
+
if (field === "dtype" || field === "kv_cache_dtype") {
|
|
42
|
+
if (normalizeDtype(actual) === normalizeDtype(wanted))
|
|
43
|
+
continue;
|
|
44
|
+
}
|
|
45
|
+
if (actual !== wanted)
|
|
46
|
+
throw mismatch(`${field}: expected ${String(wanted)}, engine reported ${String(actual)}`);
|
|
47
|
+
}
|
|
48
|
+
const tokens = info.max_total_num_tokens;
|
|
49
|
+
if (!Number.isSafeInteger(tokens) || tokens < lock.execution.max_total_tokens) {
|
|
50
|
+
throw mismatch(`engine token pool ${String(tokens)} is smaller than locked budget ${lock.execution.max_total_tokens}`);
|
|
51
|
+
}
|
|
52
|
+
if (lock.execution.platform.backend === "cuda") {
|
|
53
|
+
if (info.mem_fraction_static !== lock.execution.platform.mem_fraction_static)
|
|
54
|
+
throw mismatch("CUDA memory fraction changed");
|
|
55
|
+
if (!gpuUuid || gpuUuid !== lock.execution.platform.device_constraint)
|
|
56
|
+
throw mismatch("container GPU UUID differs from the lock");
|
|
57
|
+
}
|
|
58
|
+
else if (gpuUuid !== null)
|
|
59
|
+
throw mismatch("CPU service unexpectedly exposes a GPU");
|
|
60
|
+
return { info, tokens: tokens };
|
|
61
|
+
}
|
|
62
|
+
function observationFields(info, tokens, gpuUuid, lock) {
|
|
63
|
+
return {
|
|
64
|
+
observed_at: new Date().toISOString(), version: String(info.version),
|
|
65
|
+
device: lock.execution.platform.backend, dtype: String(info.dtype),
|
|
66
|
+
kv_cache_dtype: String(info.kv_cache_dtype), attention_backend: String(info.attention_backend),
|
|
67
|
+
sampling_backend: String(info.sampling_backend), context_length: info.context_length,
|
|
68
|
+
max_total_num_tokens: tokens, max_running_requests: info.max_running_requests,
|
|
69
|
+
gpu_uuid: gpuUuid,
|
|
70
|
+
probe: { api: lock.protocol.api, max_output_tokens: 8, streaming: true },
|
|
71
|
+
};
|
|
72
|
+
}
|
|
73
|
+
function normalizeDtype(value) {
|
|
74
|
+
return { bf16: "bfloat16", half: "float16", fp16: "float16", float: "float32" }[String(value)] ?? value;
|
|
75
|
+
}
|
|
76
|
+
function mismatch(message) {
|
|
77
|
+
return new HitchError(`SGLang runtime observation mismatch: ${message}`, { code: "inference_runtime_mismatch", exitCode: 12 });
|
|
78
|
+
}
|
|
79
|
+
//# sourceMappingURL=observation.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"observation.js","sourceRoot":"","sources":["../../../src/inference/observation.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,UAAU,EAAE,MAAM,wBAAwB,CAAC;AAEpD,gFAAgF;AAChF,MAAM,UAAU,0BAA0B,CACxC,KAAc,EAAE,OAAe,EAAE,OAAsB,EAAE,IAAqB,EAAE,OAAmC;IAEnH,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,GAAG,kBAAkB,CAAC,KAAK,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,CAAC,CAAC;IAC3E,IAAI,OAAO,CAAC,OAAO,CAAC,IAAI,KAAK,KAAK,IAAI,CAAC,uBAAuB,CAAC,IAAI,CAAC,OAAO,CAAC;QAAE,MAAM,QAAQ,CAAC,qDAAqD,CAAC,CAAC;IACpJ,OAAO;QACL,cAAc,EAAE,GAAG,EAAE,GAAG,iBAAiB,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,kBAAkB,EAAE,OAAO;KACpG,CAAC;AACJ,CAAC;AAED,uFAAuF;AACvF,MAAM,UAAU,iCAAiC,CAC/C,KAAc,EAAE,QAAoC,EAAE,OAAsB,EAAE,IAAqB,EACnG,OAAmC,EAAE,YAAqD;IAE1F,IAAI,OAAO,CAAC,cAAc,KAAK,GAAG,IAAI,OAAO,CAAC,OAAO,CAAC,IAAI,KAAK,YAAY,IAAI,QAAQ,CAAC,IAAI,KAAK,YAAY;WACxG,QAAQ,CAAC,OAAO,KAAK,YAAY,CAAC,OAAO,IAAI,QAAQ,CAAC,UAAU,KAAK,YAAY,CAAC,UAAU;WAC5F,QAAQ,CAAC,kBAAkB,KAAK,OAAO,CAAC,OAAO,CAAC,kBAAkB,IAAI,QAAQ,CAAC,cAAc,KAAK,OAAO,CAAC,OAAO,CAAC,cAAc;WAChI,QAAQ,CAAC,eAAe,KAAK,OAAO,CAAC,OAAO,CAAC,eAAe;WAC5D,CAAC,QAAQ,CAAC,kBAAkB,KAAK,IAAI,IAAI,CAAC,uBAAuB,CAAC,IAAI,CAAC,QAAQ,CAAC,kBAAkB,CAAC,CAAC;QAAE,MAAM,QAAQ,CAAC,qDAAqD,CAAC,CAAC;IACjL,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,GAAG,kBAAkB,CAAC,KAAK,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,CAAC,CAAC;IAC3E,OAAO,EAAE,cAAc,EAAE,GAAG,EAAE,GAAG,iBAAiB,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,OAAO,EAAE;YACxF,IAAI,EAAE,YAAY,EAAE,OAAO,EAAE,QAAQ,CAAC,OAAO,EAAE,UAAU,EAAE,QAAQ,CAAC,UAAU;YAC9E,kBAAkB,EAAE,QAAQ,CAAC,kBAAkB,EAAE,cAAc,EAAE,QAAQ,CAAC,cAAc;YACxF,eAAe,EAAE,QAAQ,CAAC,eAAe,EAAE,kBAAkB,EAAE,QAAQ,CAAC,kBAAkB;SAC3F,EAAE,CAAC;AACN,CAAC;AAED,SAAS,kBAAkB,CAAC,KAAc,EAAE,OAAsB,EAAE,IAAqB,EAAE,OAAmC;IAC5H,IAAI,CAAC,KAAK,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC;QAAE,MAAM,QAAQ,CAAC,+BAA+B,CAAC,CAAC;IACjH,MAAM,IAAI,GAAG,KAAgC,CAAC;IAC9C,MAAM,QAAQ,GAA4B;QACxC,OAAO,EAAE,OAAO,CAAC,cAAc,EAAE,MAAM,EAAE,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO;QACxE,KAAK,EAAE,IAAI,CAAC,SAAS,CAAC,KAAK,EAAE,cAAc,EAAE,IAAI,CAAC,SAAS,CAAC,cAAc;QAC1E,iBAAiB,EAAE,IAAI,CAAC,SAAS,CAAC,iBAAiB,EAAE,gBAAgB,EAAE,IAAI,CAAC,SAAS,CAAC,gBAAgB;QACtG,cAAc,EAAE,IAAI,CAAC,SAAS,CAAC,0BAA0B,EAAE,oBAAoB,EAAE,IAAI,CAAC,SAAS,CAAC,oBAAoB;QACpH,OAAO,EAAE,CAAC,EAAE,OAAO,EAAE,CAAC,EAAE,OAAO,EAAE,CAAC;QAClC,mBAAmB,EAAE,IAAI,CAAC,SAAS,CAAC,YAAY,CAAC,IAAI,KAAK,UAAU;QACpE,wBAAwB,EAAE,CAAC,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,gBAAgB;KACpE,CAAC;IACF,KAAK,MAAM,CAAC,KAAK,EAAE,MAAM,CAAC,IAAI,MAAM,CAAC,OAAO,CAAC,QAAQ,CAAC,EAAE,CAAC;QACvD,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC;QAC3B,IAAI,KAAK,KAAK,OAAO,IAAI,KAAK,KAAK,gBAAgB,EAAE,CAAC;YACpD,IAAI,cAAc,CAAC,MAAM,CAAC,KAAK,cAAc,CAAC,MAAM,CAAC;gBAAE,SAAS;QAClE,CAAC;QACD,IAAI,MAAM,KAAK,MAAM;YAAE,MAAM,QAAQ,CAAC,GAAG,KAAK,cAAc,MAAM,CAAC,MAAM,CAAC,qBAAqB,MAAM,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC;IACnH,CAAC;IACD,MAAM,MAAM,GAAG,IAAI,CAAC,oBAAoB,CAAC;IACzC,IAAI,CAAC,MAAM,CAAC,aAAa,CAAC,MAAM,CAAC,IAAK,MAAiB,GAAG,IAAI,CAAC,SAAS,CAAC,gBAAgB,EAAE,CAAC;QAC1F,MAAM,QAAQ,CAAC,qBAAqB,MAAM,CAAC,MAAM,CAAC,kCAAkC,IAAI,CAAC,SAAS,CAAC,gBAAgB,EAAE,CAAC,CAAC;IACzH,CAAC;IACD,IAAI,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,MAAM,EAAE,CAAC;QAC/C,IAAI,IAAI,CAAC,mBAAmB,KAAK,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,mBAAmB;YAAE,MAAM,QAAQ,CAAC,8BAA8B,CAAC,CAAC;QAC7H,IAAI,CAAC,OAAO,IAAI,OAAO,KAAK,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,iBAAiB;YAAE,MAAM,QAAQ,CAAC,0CAA0C,CAAC,CAAC;IACpI,CAAC;SAAM,IAAI,OAAO,KAAK,IAAI;QAAE,MAAM,QAAQ,CAAC,wCAAwC,CAAC,CAAC;IACtF,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,MAAgB,EAAE,CAAC;AAC5C,CAAC;AAED,SAAS,iBAAiB,CAAC,IAA6B,EAAE,MAAc,EAAE,OAAsB,EAAE,IAAqB;IACrH,OAAO;QACL,WAAW,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE,EAAE,OAAO,EAAE,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC;QACpE,MAAM,EAAE,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAyB,EAAE,KAAK,EAAE,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC;QACpF,cAAc,EAAE,MAAM,CAAC,IAAI,CAAC,cAAc,CAAC,EAAE,iBAAiB,EAAE,MAAM,CAAC,IAAI,CAAC,iBAAiB,CAAC;QAC9F,gBAAgB,EAAE,MAAM,CAAC,IAAI,CAAC,gBAAgB,CAAC,EAAE,cAAc,EAAE,IAAI,CAAC,cAAwB;QAC9F,oBAAoB,EAAE,MAAgB,EAAE,oBAAoB,EAAE,IAAI,CAAC,oBAA8B;QACjG,QAAQ,EAAE,OAAO;QACjB,KAAK,EAAE,EAAE,GAAG,EAAE,IAAI,CAAC,QAAQ,CAAC,GAAG,EAAE,iBAAiB,EAAE,CAAC,EAAE,SAAS,EAAE,IAAI,EAAE;KACzE,CAAC;AACJ,CAAC;AACD,SAAS,cAAc,CAAC,KAAc;IACpC,OAAQ,EAAE,IAAI,EAAE,UAAU,EAAE,IAAI,EAAE,SAAS,EAAE,IAAI,EAAE,SAAS,EAAE,KAAK,EAAE,SAAS,EAA6B,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,IAAI,KAAK,CAAC;AACtI,CAAC;AACD,SAAS,QAAQ,CAAC,OAAe;IAC/B,OAAO,IAAI,UAAU,CAAC,wCAAwC,OAAO,EAAE,EAAE,EAAE,IAAI,EAAE,4BAA4B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;AACjI,CAAC"}
|
|
@@ -0,0 +1,132 @@
|
|
|
1
|
+
import { HitchError } from "../foundation/index.js";
|
|
2
|
+
import { resolveLocalInferenceDevice } from "./doctor.js";
|
|
3
|
+
import { reservedInferenceDevices } from "./device-reservation.js";
|
|
4
|
+
import { buildInferenceLock, loadInferenceLock, persistInferenceLock } from "./lock.js";
|
|
5
|
+
import { resolveLocalModel, verifyLocalModel } from "./model-store.js";
|
|
6
|
+
import { loadInferenceRuntime, prepareInferenceRuntime } from "./runtime-store.js";
|
|
7
|
+
import { loadModelNodeClient, observeModelNode, runtimeFromModelNode } from "./node-registry.js";
|
|
8
|
+
import { parseModelNodeBinding } from "./manifest.js";
|
|
9
|
+
import { sha256JSON } from "../foundation/index.js";
|
|
10
|
+
export async function prepareLocalInference(options) {
|
|
11
|
+
const model = await resolveLocalModel(options.root, options.selection.model);
|
|
12
|
+
if (options.selection.model_node) {
|
|
13
|
+
const { verifyModelOnNode } = await import("./remote-model-store.js");
|
|
14
|
+
await verifyModelOnNode(options.root, model, options.selection.model_node);
|
|
15
|
+
return prepareModelNodeInference(options, model);
|
|
16
|
+
}
|
|
17
|
+
await verifyLocalModel(options.root, model);
|
|
18
|
+
const reusable = options.reusableLocks?.find((lock) => lock.model_id === model.model_id
|
|
19
|
+
&& !lock.model_node
|
|
20
|
+
&& lock.protocol.api === (options.harnessRef?.startsWith("training-tool@") ? "chat-completions" : "responses")
|
|
21
|
+
&& lock.profile === options.selection.profile
|
|
22
|
+
&& (options.selection.device === "auto" || options.selection.device === lock.execution.platform.backend));
|
|
23
|
+
const inferenceId = options.selection.inference_id ?? reusable?.inference_id;
|
|
24
|
+
if (inferenceId) {
|
|
25
|
+
const lock = await loadInferenceLock(options.root, inferenceId);
|
|
26
|
+
if (lock.model_node)
|
|
27
|
+
throw new HitchError("a v2 inference lock requires its explicit model-node selection", { code: "inference_lock_mismatch", exitCode: 2 });
|
|
28
|
+
if (lock.model_id !== model.model_id) {
|
|
29
|
+
throw new HitchError("prepared inference identity refers to a different model", { code: "inference_lock_mismatch", exitCode: 2 });
|
|
30
|
+
}
|
|
31
|
+
const resolved = await resolveLocalInferenceDevice(lock.execution.platform.backend, {
|
|
32
|
+
...options.doctor,
|
|
33
|
+
...(lock.execution.platform.backend === "cuda" && lock.execution.platform.device_constraint
|
|
34
|
+
? { deviceConstraint: lock.execution.platform.device_constraint } : {}),
|
|
35
|
+
});
|
|
36
|
+
assertHarnessCompatibility(options.harnessRef, lock);
|
|
37
|
+
return {
|
|
38
|
+
doctor: resolved.doctor,
|
|
39
|
+
model,
|
|
40
|
+
runtime: await loadInferenceRuntime(options.root, lock.runtime_id),
|
|
41
|
+
lock,
|
|
42
|
+
runtime_cache_hit: true,
|
|
43
|
+
};
|
|
44
|
+
}
|
|
45
|
+
const weights = model.files.filter((file) => file.path.endsWith(".safetensors")).reduce((sum, file) => sum + file.size, 0);
|
|
46
|
+
const resolved = await resolveLocalInferenceDevice(options.selection.device, {
|
|
47
|
+
...options.doctor, requiredMemoryMiB: Math.ceil((weights * 1.25 + 1024 ** 3) / 1024 ** 2),
|
|
48
|
+
excludedDeviceUuids: [...await reservedInferenceDevices(), ...(options.doctor?.excludedDeviceUuids ?? [])],
|
|
49
|
+
});
|
|
50
|
+
if (resolved.backend === "metal") {
|
|
51
|
+
throw new HitchError("Metal local inference is not available in P0", { code: "inference_device_unsupported", exitCode: 3 });
|
|
52
|
+
}
|
|
53
|
+
const prepared = await prepareInferenceRuntime({
|
|
54
|
+
...options.runtime,
|
|
55
|
+
root: options.root,
|
|
56
|
+
backend: resolved.backend,
|
|
57
|
+
offline: options.selection.offline,
|
|
58
|
+
...(options.onProgress ? { onProgress: options.onProgress } : {}),
|
|
59
|
+
});
|
|
60
|
+
const lock = buildInferenceLock(model, prepared.manifest, {
|
|
61
|
+
backend: resolved.backend,
|
|
62
|
+
profile: options.selection.profile,
|
|
63
|
+
...(options.harnessRef?.startsWith("training-tool@") ? { api: "chat-completions" } : {}),
|
|
64
|
+
...(resolved.doctor.gpu?.uuid ? { deviceConstraint: resolved.doctor.gpu.uuid } : {}),
|
|
65
|
+
});
|
|
66
|
+
assertHarnessCompatibility(options.harnessRef, lock);
|
|
67
|
+
await persistInferenceLock(options.root, lock);
|
|
68
|
+
return {
|
|
69
|
+
model,
|
|
70
|
+
runtime: prepared.manifest,
|
|
71
|
+
lock,
|
|
72
|
+
doctor: resolved.doctor,
|
|
73
|
+
runtime_cache_hit: prepared.cache_hit,
|
|
74
|
+
};
|
|
75
|
+
}
|
|
76
|
+
async function prepareModelNodeInference(options, model) {
|
|
77
|
+
const binding = parseModelNodeBinding(options.selection.model_node);
|
|
78
|
+
if (!["auto", "cuda"].includes(options.selection.device))
|
|
79
|
+
throw new HitchError("model-node inference requires CUDA", { code: "inference_device_unsupported", exitCode: 3 });
|
|
80
|
+
const observation = await observeModelNode(await loadModelNodeClient(options.root, binding), binding);
|
|
81
|
+
const runtime = runtimeFromModelNode(observation);
|
|
82
|
+
const persistedRuntime = await loadInferenceRuntime(options.root, runtime.runtime_id);
|
|
83
|
+
if (sha256JSON(persistedRuntime) !== sha256JSON(runtime))
|
|
84
|
+
throw new HitchError("registered model-node runtime changed", { code: "inference_runtime_mismatch", exitCode: 12 });
|
|
85
|
+
const reusable = options.reusableLocks?.find(lock => lock.model_id === model.model_id && lock.profile === options.selection.profile
|
|
86
|
+
&& sha256JSON(lock.model_node ?? null) === sha256JSON(binding)
|
|
87
|
+
&& lock.protocol.api === (options.harnessRef?.startsWith("training-tool@") ? "chat-completions" : "responses"));
|
|
88
|
+
const inferenceId = options.selection.inference_id ?? reusable?.inference_id;
|
|
89
|
+
const device = options.doctor?.deviceConstraint;
|
|
90
|
+
let lock;
|
|
91
|
+
if (inferenceId) {
|
|
92
|
+
lock = await loadInferenceLock(options.root, inferenceId);
|
|
93
|
+
if (sha256JSON(lock.model_node ?? null) !== sha256JSON(binding) || lock.model_id !== model.model_id || lock.runtime_id !== runtime.runtime_id) {
|
|
94
|
+
throw new HitchError("inference lock differs from the frozen model node, runtime or model", { code: "inference_lock_mismatch", exitCode: 2 });
|
|
95
|
+
}
|
|
96
|
+
if (device && (lock.execution.platform.backend !== "cuda" || lock.execution.platform.device_constraint !== device))
|
|
97
|
+
throw new HitchError("inference device changed", { code: "inference_lock_mismatch", exitCode: 2 });
|
|
98
|
+
}
|
|
99
|
+
else {
|
|
100
|
+
if (!device)
|
|
101
|
+
throw new HitchError("plan model-node inference with an explicit --gpu before submitting it", { code: "inference_device_unavailable", exitCode: 3 });
|
|
102
|
+
lock = buildInferenceLock(model, runtime, { backend: "cuda", profile: options.selection.profile, modelNode: binding, deviceConstraint: device,
|
|
103
|
+
...(options.harnessRef?.startsWith("training-tool@") ? { api: "chat-completions" } : {}) });
|
|
104
|
+
}
|
|
105
|
+
if (lock.execution.platform.backend !== "cuda" || !observation.gpuUuids.includes(lock.execution.platform.device_constraint ?? "")) {
|
|
106
|
+
throw new HitchError("locked physical GPU is absent on the model node", { code: "inference_device_unavailable", exitCode: 3 });
|
|
107
|
+
}
|
|
108
|
+
assertHarnessCompatibility(options.harnessRef, lock);
|
|
109
|
+
await persistInferenceLock(options.root, lock);
|
|
110
|
+
return { model, runtime, lock, runtime_cache_hit: true, node_observation: observation };
|
|
111
|
+
}
|
|
112
|
+
function assertHarnessCompatibility(harnessRef, lock) {
|
|
113
|
+
if (!harnessRef)
|
|
114
|
+
return;
|
|
115
|
+
const harness = harnessRef.split("@", 1)[0];
|
|
116
|
+
if (harness !== "codex" && harness !== "model-call" && harness !== "training-tool") {
|
|
117
|
+
throw new HitchError(`local inference is not supported in the preview for ${harness}`, { code: "inference_harness_unsupported", exitCode: 2 });
|
|
118
|
+
}
|
|
119
|
+
if (harness === "training-tool" && lock.protocol.api !== "chat-completions")
|
|
120
|
+
throw new HitchError("training-tool requires an explicit Chat Completions inference lock", { code: "inference_protocol_unsupported", exitCode: 2 });
|
|
121
|
+
if ((harness === "codex" || harness === "training-tool") && !lock.protocol.tool_calls) {
|
|
122
|
+
throw new HitchError("this model type has no configured SGLang tool-call parser for Codex", {
|
|
123
|
+
code: "inference_protocol_unsupported", exitCode: 2,
|
|
124
|
+
});
|
|
125
|
+
}
|
|
126
|
+
if (harness === "codex" && harnessRef !== "codex@version:0.145.0") {
|
|
127
|
+
throw new HitchError("local inference supports only codex@version:0.145.0", {
|
|
128
|
+
code: "inference_harness_unsupported", exitCode: 2,
|
|
129
|
+
});
|
|
130
|
+
}
|
|
131
|
+
}
|
|
132
|
+
//# sourceMappingURL=preflight.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"preflight.js","sourceRoot":"","sources":["../../../src/inference/preflight.ts"],"names":[],"mappings":"AAMA,OAAO,EAAE,UAAU,EAAE,MAAM,wBAAwB,CAAC;AAEpD,OAAO,EAAE,2BAA2B,EAAE,MAAM,aAAa,CAAC;AAC1D,OAAO,EAAE,wBAAwB,EAAE,MAAM,yBAAyB,CAAC;AACnE,OAAO,EAAE,kBAAkB,EAAE,iBAAiB,EAAE,oBAAoB,EAAE,MAAM,WAAW,CAAC;AACxF,OAAO,EAAE,iBAAiB,EAAE,gBAAgB,EAAE,MAAM,kBAAkB,CAAC;AACvE,OAAO,EAAE,oBAAoB,EAAE,uBAAuB,EAAE,MAAM,oBAAoB,CAAC;AAEnF,OAAO,EAAE,mBAAmB,EAAE,gBAAgB,EAAE,oBAAoB,EAAE,MAAM,oBAAoB,CAAC;AACjG,OAAO,EAAE,qBAAqB,EAAE,MAAM,eAAe,CAAC;AACtD,OAAO,EAAE,UAAU,EAAE,MAAM,wBAAwB,CAAC;AAsBpD,MAAM,CAAC,KAAK,UAAU,qBAAqB,CAAC,OAAuC;IACjF,MAAM,KAAK,GAAG,MAAM,iBAAiB,CAAC,OAAO,CAAC,IAAI,EAAE,OAAO,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC;IAC7E,IAAI,OAAO,CAAC,SAAS,CAAC,UAAU,EAAE,CAAC;QACjC,MAAM,EAAE,iBAAiB,EAAE,GAAG,MAAM,MAAM,CAAC,yBAAyB,CAAC,CAAC;QACtE,MAAM,iBAAiB,CAAC,OAAO,CAAC,IAAI,EAAE,KAAK,EAAE,OAAO,CAAC,SAAS,CAAC,UAAU,CAAC,CAAC;QAC3E,OAAO,yBAAyB,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;IACnD,CAAC;IACD,MAAM,gBAAgB,CAAC,OAAO,CAAC,IAAI,EAAE,KAAK,CAAC,CAAC;IAC5C,MAAM,QAAQ,GAAG,OAAO,CAAC,aAAa,EAAE,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,QAAQ,KAAK,KAAK,CAAC,QAAQ;WAClF,CAAC,IAAI,CAAC,UAAU;WAChB,IAAI,CAAC,QAAQ,CAAC,GAAG,KAAK,CAAC,OAAO,CAAC,UAAU,EAAE,UAAU,CAAC,gBAAgB,CAAC,CAAC,CAAC,CAAC,kBAAkB,CAAC,CAAC,CAAC,WAAW,CAAC;WAC3G,IAAI,CAAC,OAAO,KAAK,OAAO,CAAC,SAAS,CAAC,OAAO;WAC1C,CAAC,OAAO,CAAC,SAAS,CAAC,MAAM,KAAK,MAAM,IAAI,OAAO,CAAC,SAAS,CAAC,MAAM,KAAK,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC;IAC5G,MAAM,WAAW,GAAG,OAAO,CAAC,SAAS,CAAC,YAAY,IAAI,QAAQ,EAAE,YAAY,CAAC;IAC7E,IAAI,WAAW,EAAE,CAAC;QAChB,MAAM,IAAI,GAAG,MAAM,iBAAiB,CAAC,OAAO,CAAC,IAAI,EAAE,WAAW,CAAC,CAAC;QAChE,IAAI,IAAI,CAAC,UAAU;YAAE,MAAM,IAAI,UAAU,CAAC,gEAAgE,EAAE,EAAE,IAAI,EAAE,yBAAyB,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;QAC9J,IAAI,IAAI,CAAC,QAAQ,KAAK,KAAK,CAAC,QAAQ,EAAE,CAAC;YACrC,MAAM,IAAI,UAAU,CAAC,yDAAyD,EAAE,EAAE,IAAI,EAAE,yBAAyB,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;QACpI,CAAC;QACD,MAAM,QAAQ,GAAG,MAAM,2BAA2B,CAAC,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,EAAE;YAClF,GAAG,OAAO,CAAC,MAAM;YACjB,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,MAAM,IAAI,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,iBAAiB;gBACzF,CAAC,CAAC,EAAE,gBAAgB,EAAE,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,iBAAiB,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;SAC1E,CAAC,CAAC;QACH,0BAA0B,CAAC,OAAO,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC;QACrD,OAAO;YACL,MAAM,EAAE,QAAQ,CAAC,MAAM;YACvB,KAAK;YACL,OAAO,EAAE,MAAM,oBAAoB,CAAC,OAAO,CAAC,IAAI,EAAE,IAAI,CAAC,UAAU,CAAC;YAClE,IAAI;YACJ,iBAAiB,EAAE,IAAI;SACxB,CAAC;IACJ,CAAC;IACD,MAAM,OAAO,GAAG,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,cAAc,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,IAAI,EAAE,EAAE,CAAC,GAAG,GAAG,IAAI,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC;IAC3H,MAAM,QAAQ,GAAG,MAAM,2BAA2B,CAAC,OAAO,CAAC,SAAS,CAAC,MAAM,EAAE;QAC3E,GAAG,OAAO,CAAC,MAAM,EAAE,iBAAiB,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC,OAAO,GAAG,IAAI,GAAG,IAAI,IAAI,CAAC,CAAC,GAAG,IAAI,IAAI,CAAC,CAAC;QACzF,mBAAmB,EAAE,CAAC,GAAG,MAAM,wBAAwB,EAAE,EAAE,GAAG,CAAC,OAAO,CAAC,MAAM,EAAE,mBAAmB,IAAI,EAAE,CAAC,CAAC;KAC3G,CAAC,CAAC;IACH,IAAI,QAAQ,CAAC,OAAO,KAAK,OAAO,EAAE,CAAC;QACjC,MAAM,IAAI,UAAU,CAAC,8CAA8C,EAAE,EAAE,IAAI,EAAE,8BAA8B,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;IAC9H,CAAC;IACD,MAAM,QAAQ,GAAG,MAAM,uBAAuB,CAAC;QAC7C,GAAG,OAAO,CAAC,OAAO;QAClB,IAAI,EAAE,OAAO,CAAC,IAAI;QAClB,OAAO,EAAE,QAAQ,CAAC,OAAO;QACzB,OAAO,EAAE,OAAO,CAAC,SAAS,CAAC,OAAO;QAClC,GAAG,CAAC,OAAO,CAAC,UAAU,CAAC,CAAC,CAAC,EAAE,UAAU,EAAE,OAAO,CAAC,UAAU,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;KAClE,CAAC,CAAC;IACH,MAAM,IAAI,GAAG,kBAAkB,CAAC,KAAK,EAAE,QAAQ,CAAC,QAAQ,EAAE;QACxD,OAAO,EAAE,QAAQ,CAAC,OAAO;QACzB,OAAO,EAAE,OAAO,CAAC,SAAS,CAAC,OAAO;QAClC,GAAG,CAAC,OAAO,CAAC,UAAU,EAAE,UAAU,CAAC,gBAAgB,CAAC,CAAC,CAAC,CAAC,EAAE,GAAG,EAAE,kBAA2B,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACjG,GAAG,CAAC,QAAQ,CAAC,MAAM,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC,CAAC,EAAE,gBAAgB,EAAE,QAAQ,CAAC,MAAM,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;KACrF,CAAC,CAAC;IACH,0BAA0B,CAAC,OAAO,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC;IACrD,MAAM,oBAAoB,CAAC,OAAO,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC;IAC/C,OAAO;QACL,KAAK;QACL,OAAO,EAAE,QAAQ,CAAC,QAAQ;QAC1B,IAAI;QACJ,MAAM,EAAE,QAAQ,CAAC,MAAM;QACvB,iBAAiB,EAAE,QAAQ,CAAC,SAAS;KACtC,CAAC;AACJ,CAAC;AAED,KAAK,UAAU,yBAAyB,CAAC,OAAuC,EAAE,KAA2B;IAC3G,MAAM,OAAO,GAAG,qBAAqB,CAAC,OAAO,CAAC,SAAS,CAAC,UAAU,CAAC,CAAC;IACpE,IAAI,CAAC,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,QAAQ,CAAC,OAAO,CAAC,SAAS,CAAC,MAAM,CAAC;QAAE,MAAM,IAAI,UAAU,CAAC,oCAAoC,EAAE,EAAE,IAAI,EAAE,8BAA8B,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;IAC5K,MAAM,WAAW,GAAG,MAAM,gBAAgB,CAAC,MAAM,mBAAmB,CAAC,OAAO,CAAC,IAAI,EAAE,OAAO,CAAC,EAAE,OAAO,CAAC,CAAC;IACtG,MAAM,OAAO,GAAG,oBAAoB,CAAC,WAAW,CAAC,CAAC;IAClD,MAAM,gBAAgB,GAAG,MAAM,oBAAoB,CAAC,OAAO,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC,CAAC;IACtF,IAAI,UAAU,CAAC,gBAAgB,CAAC,KAAK,UAAU,CAAC,OAAO,CAAC;QAAE,MAAM,IAAI,UAAU,CAAC,uCAAuC,EAAE,EAAE,IAAI,EAAE,4BAA4B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;IAC9K,MAAM,QAAQ,GAAG,OAAO,CAAC,aAAa,EAAE,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,QAAQ,KAAK,KAAK,CAAC,QAAQ,IAAI,IAAI,CAAC,OAAO,KAAK,OAAO,CAAC,SAAS,CAAC,OAAO;WAC9H,UAAU,CAAC,IAAI,CAAC,UAAU,IAAI,IAAI,CAAC,KAAK,UAAU,CAAC,OAAO,CAAC;WAC3D,IAAI,CAAC,QAAQ,CAAC,GAAG,KAAK,CAAC,OAAO,CAAC,UAAU,EAAE,UAAU,CAAC,gBAAgB,CAAC,CAAC,CAAC,CAAC,kBAAkB,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC;IAClH,MAAM,WAAW,GAAG,OAAO,CAAC,SAAS,CAAC,YAAY,IAAI,QAAQ,EAAE,YAAY,CAAC;IAC7E,MAAM,MAAM,GAAG,OAAO,CAAC,MAAM,EAAE,gBAAgB,CAAC;IAChD,IAAI,IAAqB,CAAC;IAC1B,IAAI,WAAW,EAAE,CAAC;QAChB,IAAI,GAAG,MAAM,iBAAiB,CAAC,OAAO,CAAC,IAAI,EAAE,WAAW,CAAC,CAAC;QAC1D,IAAI,UAAU,CAAC,IAAI,CAAC,UAAU,IAAI,IAAI,CAAC,KAAK,UAAU,CAAC,OAAO,CAAC,IAAI,IAAI,CAAC,QAAQ,KAAK,KAAK,CAAC,QAAQ,IAAI,IAAI,CAAC,UAAU,KAAK,OAAO,CAAC,UAAU,EAAE,CAAC;YAC9I,MAAM,IAAI,UAAU,CAAC,qEAAqE,EAAE,EAAE,IAAI,EAAE,yBAAyB,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;QAChJ,CAAC;QACD,IAAI,MAAM,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,MAAM,IAAI,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,iBAAiB,KAAK,MAAM,CAAC;YAAE,MAAM,IAAI,UAAU,CAAC,0BAA0B,EAAE,EAAE,IAAI,EAAE,yBAAyB,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;IACzN,CAAC;SAAM,CAAC;QACN,IAAI,CAAC,MAAM;YAAE,MAAM,IAAI,UAAU,CAAC,uEAAuE,EAAE,EAAE,IAAI,EAAE,8BAA8B,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;QAClK,IAAI,GAAG,kBAAkB,CAAC,KAAK,EAAE,OAAO,EAAE,EAAE,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,OAAO,CAAC,SAAS,CAAC,OAAO,EAAE,SAAS,EAAE,OAAO,EAAE,gBAAgB,EAAE,MAAM;YAC3I,GAAG,CAAC,OAAO,CAAC,UAAU,EAAE,UAAU,CAAC,gBAAgB,CAAC,CAAC,CAAC,CAAC,EAAE,GAAG,EAAE,kBAA2B,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC;IACzG,CAAC;IACD,IAAI,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,MAAM,IAAI,CAAC,WAAW,CAAC,QAAQ,CAAC,QAAQ,CAAC,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,iBAAiB,IAAI,EAAE,CAAC,EAAE,CAAC;QAClI,MAAM,IAAI,UAAU,CAAC,iDAAiD,EAAE,EAAE,IAAI,EAAE,8BAA8B,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;IACjI,CAAC;IACD,0BAA0B,CAAC,OAAO,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC;IACrD,MAAM,oBAAoB,CAAC,OAAO,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC;IAC/C,OAAO,EAAE,KAAK,EAAE,OAAO,EAAE,IAAI,EAAE,iBAAiB,EAAE,IAAI,EAAE,gBAAgB,EAAE,WAAW,EAAE,CAAC;AAC1F,CAAC;AAED,SAAS,0BAA0B,CAAC,UAA8B,EAAE,IAAqB;IACvF,IAAI,CAAC,UAAU;QAAE,OAAO;IACxB,MAAM,OAAO,GAAG,UAAU,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IAC5C,IAAI,OAAO,KAAK,OAAO,IAAI,OAAO,KAAK,YAAY,IAAI,OAAO,KAAK,eAAe,EAAE,CAAC;QACnF,MAAM,IAAI,UAAU,CAAC,uDAAuD,OAAO,EAAE,EAAE,EAAE,IAAI,EAAE,+BAA+B,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;IACjJ,CAAC;IACD,IAAI,OAAO,KAAK,eAAe,IAAI,IAAI,CAAC,QAAQ,CAAC,GAAG,KAAK,kBAAkB;QAAE,MAAM,IAAI,UAAU,CAAC,oEAAoE,EAAE,EAAE,IAAI,EAAE,gCAAgC,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;IACjO,IAAI,CAAC,OAAO,KAAK,OAAO,IAAI,OAAO,KAAK,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,UAAU,EAAE,CAAC;QACtF,MAAM,IAAI,UAAU,CAAC,qEAAqE,EAAE;YAC1F,IAAI,EAAE,gCAAgC,EAAE,QAAQ,EAAE,CAAC;SACpD,CAAC,CAAC;IACL,CAAC;IACD,IAAI,OAAO,KAAK,OAAO,IAAI,UAAU,KAAK,uBAAuB,EAAE,CAAC;QAClE,MAAM,IAAI,UAAU,CAAC,qDAAqD,EAAE;YAC1E,IAAI,EAAE,+BAA+B,EAAE,QAAQ,EAAE,CAAC;SACnD,CAAC,CAAC;IACL,CAAC;AACH,CAAC"}
|
|
@@ -0,0 +1,169 @@
|
|
|
1
|
+
import { HitchError, delay, hitchRootId, sha256JSON } from "../foundation/index.js";
|
|
2
|
+
import { parseInferenceRuntimeManifest, parseInferenceServiceHandle, parseInferenceServiceRecord, parseLocalModelManifest } from "./manifest.js";
|
|
3
|
+
import { validateInferenceLockShape } from "./lock.js";
|
|
4
|
+
import { validateProcessRuntimeObservation } from "./observation.js";
|
|
5
|
+
import { probeSGLangProtocol } from "./protocol-probe.js";
|
|
6
|
+
export class ProcessSGLangLauncher {
|
|
7
|
+
client;
|
|
8
|
+
request;
|
|
9
|
+
constructor(client, request = fetch) {
|
|
10
|
+
this.client = client;
|
|
11
|
+
this.request = request;
|
|
12
|
+
}
|
|
13
|
+
status(value, owner, expected) {
|
|
14
|
+
const status = value;
|
|
15
|
+
if (!status || status.schemaVersion !== 2 || typeof status.resourcesReleased !== "boolean"
|
|
16
|
+
|| !["admitting", "starting", "ready", "stopping", "stopped", "failed"].includes(status.state))
|
|
17
|
+
throw new TypeError("invalid model-node service status");
|
|
18
|
+
if (owner.inferenceId && status.inferenceId !== owner.inferenceId)
|
|
19
|
+
throw new HitchError("model-node service inference lock changed", { code: "inference_runtime_mismatch", exitCode: 12 });
|
|
20
|
+
if (status.handle) {
|
|
21
|
+
const handle = parseInferenceServiceHandle(status.handle);
|
|
22
|
+
if (handle.kind !== "process" || handle.service_id !== owner.serviceId || handle.node_id !== this.client.node.nodeId
|
|
23
|
+
|| handle.generation !== this.client.node.generation || (expected && sha256JSON(handle) !== sha256JSON(expected))) {
|
|
24
|
+
throw new HitchError("model-node process identity changed", { code: "inference_runtime_mismatch", exitCode: 12 });
|
|
25
|
+
}
|
|
26
|
+
}
|
|
27
|
+
else if (expected || status.state === "ready")
|
|
28
|
+
throw new TypeError("model-node process handle is missing");
|
|
29
|
+
return status;
|
|
30
|
+
}
|
|
31
|
+
async start(input) {
|
|
32
|
+
validateInput(input);
|
|
33
|
+
const owner = { serviceId: input.serviceId, ownerId: hitchRootId(input.root), inferenceId: input.lock.inference_id };
|
|
34
|
+
await this.client.prepare(input);
|
|
35
|
+
let handle;
|
|
36
|
+
try {
|
|
37
|
+
let status = this.status(await this.client.start(input), owner);
|
|
38
|
+
const deadline = Date.now() + input.lock.execution.startup_timeout_ms;
|
|
39
|
+
while (true) {
|
|
40
|
+
if (status.handle && !handle) {
|
|
41
|
+
handle = status.handle;
|
|
42
|
+
await input.onServiceCreated?.(handle);
|
|
43
|
+
}
|
|
44
|
+
if (input.signal?.aborted)
|
|
45
|
+
throw new HitchError("SGLang startup cancelled", { code: "cancelled", exitCode: 9 });
|
|
46
|
+
if (status.state === "ready")
|
|
47
|
+
break;
|
|
48
|
+
if (status.state === "failed" || status.state === "stopped" || status.state === "stopping")
|
|
49
|
+
throw new HitchError("model-node process did not become ready", { code: "inference_process_exited", exitCode: 12 });
|
|
50
|
+
if (Date.now() >= deadline)
|
|
51
|
+
throw new HitchError("model-node startup timed out", { code: "inference_start_timeout", exitCode: 12 });
|
|
52
|
+
await delay(200);
|
|
53
|
+
status = this.status(await this.client.inspect(owner), owner, handle);
|
|
54
|
+
}
|
|
55
|
+
if (!handle)
|
|
56
|
+
throw new TypeError("ready model-node process lacks its handle");
|
|
57
|
+
return await this.connect(input, owner, status, handle);
|
|
58
|
+
}
|
|
59
|
+
catch (error) {
|
|
60
|
+
try {
|
|
61
|
+
await this.stopConfirmed(owner, handle);
|
|
62
|
+
}
|
|
63
|
+
catch {
|
|
64
|
+
throw new HitchError("model-node startup cleanup is unconfirmed; device ownership retained", { code: "inference_recovery_ambiguous", exitCode: 12, cause: error });
|
|
65
|
+
}
|
|
66
|
+
throw error;
|
|
67
|
+
}
|
|
68
|
+
}
|
|
69
|
+
async attach(input) {
|
|
70
|
+
const { signal, ...evidence } = input;
|
|
71
|
+
input = { ...structuredClone(evidence), ...(signal ? { signal } : {}) };
|
|
72
|
+
validateInput(input);
|
|
73
|
+
const record = parseInferenceServiceRecord(input.record);
|
|
74
|
+
const handle = record.service_handle;
|
|
75
|
+
if (!this.client.attach || !input.observation || input.observation.schema_version !== "2"
|
|
76
|
+
|| record.state !== "ready" || record.container_id || handle?.kind !== "process"
|
|
77
|
+
|| record.inference_id !== input.lock.inference_id
|
|
78
|
+
|| sha256JSON(record.model_node ?? null) !== sha256JSON(input.lock.model_node ?? null)
|
|
79
|
+
|| input.lock.model_node && (input.lock.model_node.node_id !== this.client.node.nodeId || input.lock.model_node.generation !== this.client.node.generation)
|
|
80
|
+
|| handle.service_id !== record.service_id || handle.node_id !== this.client.node.nodeId || handle.generation !== this.client.node.generation) {
|
|
81
|
+
throw new HitchError("service attachment requires the original ready process and model node", { code: "inference_recovery_ambiguous", exitCode: 12 });
|
|
82
|
+
}
|
|
83
|
+
input.signal?.throwIfAborted();
|
|
84
|
+
const owner = { serviceId: record.service_id, ownerId: hitchRootId(input.root), inferenceId: input.lock.inference_id };
|
|
85
|
+
// Failure must never enter startup cleanup: another live owner may still be
|
|
86
|
+
// generating. No prepare, start, token probe, cache flush or stop is allowed.
|
|
87
|
+
const status = this.status(await this.client.attach(input), owner, handle);
|
|
88
|
+
if (status.inputDigest !== sha256JSON({ serviceId: owner.serviceId, ownerId: owner.ownerId,
|
|
89
|
+
model: input.model, runtime: input.runtime, lock: input.lock }))
|
|
90
|
+
throw new TypeError("attachment start request digest changed");
|
|
91
|
+
return this.connect({ ...input, serviceId: record.service_id }, owner, status, handle, { observation: input.observation });
|
|
92
|
+
}
|
|
93
|
+
async connect(input, owner, status, handle, recovery) {
|
|
94
|
+
const access = status.access;
|
|
95
|
+
const actual = status.runtime;
|
|
96
|
+
if (status.state !== "ready" || status.resourcesReleased || !access || !actual
|
|
97
|
+
|| !/^[a-f0-9]{64}$/.test(access.engineToken) || !/^[a-f0-9]{64}$/.test(access.adminToken)
|
|
98
|
+
|| access.wireModel !== `hitch-${input.model.model_id.slice(7, 23)}`)
|
|
99
|
+
throw new TypeError("ready model-node process lacks its private connection/runtime");
|
|
100
|
+
const baseUrl = await this.client.route(input.root, access.port);
|
|
101
|
+
const headers = { Authorization: `Bearer ${access.engineToken}` };
|
|
102
|
+
const models = await this.request(`${baseUrl}/v1/models`, { headers, redirect: "error", signal: AbortSignal.timeout(5_000) });
|
|
103
|
+
const catalogue = models.ok ? await models.json() : null;
|
|
104
|
+
if (!catalogue?.data?.some(model => model.id === access.wireModel))
|
|
105
|
+
throw new HitchError("model-node route serves another model", { code: "inference_runtime_mismatch", exitCode: 12 });
|
|
106
|
+
const observed = { kind: "python-env", node_id: this.client.node.nodeId, generation: this.client.node.generation,
|
|
107
|
+
environment_digest: sha256JSON(actual), python_version: actual.pythonVersion, packages_digest: actual.packagesDigest, outer_image_digest: actual.outerImageDigest };
|
|
108
|
+
const currentObservation = validateProcessRuntimeObservation(status.serverInfo, observed, status.gpuUuid ?? null, input.lock, input.runtime, { node_id: this.client.node.nodeId, generation: this.client.node.generation });
|
|
109
|
+
let observation = currentObservation;
|
|
110
|
+
if (recovery) {
|
|
111
|
+
const prior = recovery.observation;
|
|
112
|
+
if (!prior || typeof prior.observed_at !== "string" || !Number.isFinite(Date.parse(prior.observed_at)) || Date.parse(prior.observed_at) > Date.now()
|
|
113
|
+
|| sha256JSON({ ...currentObservation, observed_at: prior.observed_at }) !== sha256JSON(prior)) {
|
|
114
|
+
throw new HitchError("live process differs from its original protocol/runtime evidence", { code: "inference_runtime_mismatch", exitCode: 12 });
|
|
115
|
+
}
|
|
116
|
+
// Retain the original probe timestamp; attachment performs no inference.
|
|
117
|
+
observation = structuredClone(prior);
|
|
118
|
+
}
|
|
119
|
+
else {
|
|
120
|
+
await probeSGLangProtocol(this.request, baseUrl, access.engineToken, access.wireModel, input.lock.protocol.api);
|
|
121
|
+
const cleared = await this.request(`${baseUrl}/flush_cache`, { headers: { Authorization: `Bearer ${access.adminToken}` }, signal: AbortSignal.timeout(10_000) });
|
|
122
|
+
if (!cleared.ok)
|
|
123
|
+
throw new HitchError("model-node warmup cache could not be cleared", { code: "inference_runtime_mismatch", exitCode: 12 });
|
|
124
|
+
}
|
|
125
|
+
input.signal?.throwIfAborted();
|
|
126
|
+
const pinnedHandle = handle;
|
|
127
|
+
let stopping;
|
|
128
|
+
return { service_handle: handle, base_url: baseUrl, wire_model: access.wireModel, engine_token: access.engineToken, admin_token: access.adminToken, observation,
|
|
129
|
+
checkHealth: async () => {
|
|
130
|
+
const current = this.status(await this.client.inspect(owner), owner, pinnedHandle);
|
|
131
|
+
if (current.state !== "ready" || current.resourcesReleased)
|
|
132
|
+
throw new HitchError("model-node process is no longer ready", { code: "inference_process_exited", exitCode: 12 });
|
|
133
|
+
await this.client.route(input.root, access.port);
|
|
134
|
+
const health = await this.request(`${baseUrl}/health`, { signal: AbortSignal.timeout(5_000) });
|
|
135
|
+
if (!health.ok)
|
|
136
|
+
throw new HitchError("model-node route is unavailable", { code: "inference_route_unavailable", exitCode: 12 });
|
|
137
|
+
},
|
|
138
|
+
stop: () => stopping ??= this.stopConfirmed(owner, pinnedHandle).finally(() => { stopping = undefined; }),
|
|
139
|
+
};
|
|
140
|
+
}
|
|
141
|
+
async stopConfirmed(owner, handle) {
|
|
142
|
+
const stopped = this.status(await this.client.stop(owner), owner, handle);
|
|
143
|
+
if (!stopped.resourcesReleased || !["stopped", "failed"].includes(stopped.state))
|
|
144
|
+
throw new HitchError("model-node process or GPU remains owned", { code: "inference_recovery_ambiguous", exitCode: 12 });
|
|
145
|
+
}
|
|
146
|
+
async stopOrphan(root, record) {
|
|
147
|
+
if (record.container_id || record.service_handle?.kind === "docker")
|
|
148
|
+
return "ambiguous";
|
|
149
|
+
const handle = record.service_handle;
|
|
150
|
+
if (handle && (handle.node_id !== this.client.node.nodeId || handle.generation !== this.client.node.generation))
|
|
151
|
+
return "ambiguous";
|
|
152
|
+
try {
|
|
153
|
+
await this.stopConfirmed({ serviceId: record.service_id, ownerId: hitchRootId(root), inferenceId: record.inference_id }, handle);
|
|
154
|
+
return "stopped";
|
|
155
|
+
}
|
|
156
|
+
catch {
|
|
157
|
+
return "ambiguous";
|
|
158
|
+
}
|
|
159
|
+
}
|
|
160
|
+
}
|
|
161
|
+
function validateInput(input) {
|
|
162
|
+
parseLocalModelManifest(input.model);
|
|
163
|
+
parseInferenceRuntimeManifest(input.runtime);
|
|
164
|
+
validateInferenceLockShape(input.lock);
|
|
165
|
+
if (input.runtime.schema_version !== "2" || input.runtime.package.kind !== "python-env"
|
|
166
|
+
|| input.lock.runtime_id !== input.runtime.runtime_id || input.lock.model_id !== input.model.model_id)
|
|
167
|
+
throw new TypeError("process launcher requires a locked v2 Python runtime/model");
|
|
168
|
+
}
|
|
169
|
+
//# sourceMappingURL=process-sglang.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"process-sglang.js","sourceRoot":"","sources":["../../../src/inference/process-sglang.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,UAAU,EAAE,KAAK,EAAE,WAAW,EAAE,UAAU,EAAE,MAAM,wBAAwB,CAAC;AACpF,OAAO,EAAE,6BAA6B,EAAE,2BAA2B,EAAE,2BAA2B,EAAE,uBAAuB,EAAE,MAAM,eAAe,CAAC;AACjJ,OAAO,EAAE,0BAA0B,EAAE,MAAM,WAAW,CAAC;AACvD,OAAO,EAAE,iCAAiC,EAAE,MAAM,kBAAkB,CAAC;AACrE,OAAO,EAAE,mBAAmB,EAAE,MAAM,qBAAqB,CAAC;AAkB1D,MAAM,OAAO,qBAAqB;IACX;IAA8C;IAAnE,YAAqB,MAA2B,EAAmB,UAAwB,KAAK;QAA3E,WAAM,GAAN,MAAM,CAAqB;QAAmB,YAAO,GAAP,OAAO,CAAsB;IAAG,CAAC;IAE5F,MAAM,CAAC,KAAc,EAAE,KAAyB,EAAE,QAAwB;QAChF,MAAM,MAAM,GAAG,KAA0B,CAAC;QAC1C,IAAI,CAAC,MAAM,IAAI,MAAM,CAAC,aAAa,KAAK,CAAC,IAAI,OAAO,MAAM,CAAC,iBAAiB,KAAK,SAAS;eACrF,CAAC,CAAC,WAAW,EAAE,UAAU,EAAE,OAAO,EAAE,UAAU,EAAE,SAAS,EAAE,QAAQ,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,KAAK,CAAC;YAAE,MAAM,IAAI,SAAS,CAAC,mCAAmC,CAAC,CAAC;QAC3J,IAAI,KAAK,CAAC,WAAW,IAAI,MAAM,CAAC,WAAW,KAAK,KAAK,CAAC,WAAW;YAAE,MAAM,IAAI,UAAU,CAAC,2CAA2C,EAAE,EAAE,IAAI,EAAE,4BAA4B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;QAC3L,IAAI,MAAM,CAAC,MAAM,EAAE,CAAC;YAClB,MAAM,MAAM,GAAG,2BAA2B,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;YAC1D,IAAI,MAAM,CAAC,IAAI,KAAK,SAAS,IAAI,MAAM,CAAC,UAAU,KAAK,KAAK,CAAC,SAAS,IAAI,MAAM,CAAC,OAAO,KAAK,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM;mBAC/G,MAAM,CAAC,UAAU,KAAK,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,UAAU,IAAI,CAAC,QAAQ,IAAI,UAAU,CAAC,MAAM,CAAC,KAAK,UAAU,CAAC,QAAQ,CAAC,CAAC,EAAE,CAAC;gBACpH,MAAM,IAAI,UAAU,CAAC,qCAAqC,EAAE,EAAE,IAAI,EAAE,4BAA4B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;YACpH,CAAC;QACH,CAAC;aAAM,IAAI,QAAQ,IAAI,MAAM,CAAC,KAAK,KAAK,OAAO;YAAE,MAAM,IAAI,SAAS,CAAC,sCAAsC,CAAC,CAAC;QAC7G,OAAO,MAAM,CAAC;IAChB,CAAC;IAED,KAAK,CAAC,KAAK,CAAC,KAAwB;QAClC,aAAa,CAAC,KAAK,CAAC,CAAC;QACrB,MAAM,KAAK,GAAG,EAAE,SAAS,EAAE,KAAK,CAAC,SAAS,EAAE,OAAO,EAAE,WAAW,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,WAAW,EAAE,KAAK,CAAC,IAAI,CAAC,YAAY,EAAE,CAAC;QACrH,MAAM,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC;QACjC,IAAI,MAAiC,CAAC;QACtC,IAAI,CAAC;YACH,IAAI,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,EAAE,KAAK,CAAC,CAAC;YAChE,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAAC;YACtE,OAAO,IAAI,EAAE,CAAC;gBACZ,IAAI,MAAM,CAAC,MAAM,IAAI,CAAC,MAAM,EAAE,CAAC;oBAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;oBAAC,MAAM,KAAK,CAAC,gBAAgB,EAAE,CAAC,MAAM,CAAC,CAAC;gBAAC,CAAC;gBACjG,IAAI,KAAK,CAAC,MAAM,EAAE,OAAO;oBAAE,MAAM,IAAI,UAAU,CAAC,0BAA0B,EAAE,EAAE,IAAI,EAAE,WAAW,EAAE,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC;gBAChH,IAAI,MAAM,CAAC,KAAK,KAAK,OAAO;oBAAE,MAAM;gBACpC,IAAI,MAAM,CAAC,KAAK,KAAK,QAAQ,IAAI,MAAM,CAAC,KAAK,KAAK,SAAS,IAAI,MAAM,CAAC,KAAK,KAAK,UAAU;oBAAE,MAAM,IAAI,UAAU,CAAC,yCAAyC,EAAE,EAAE,IAAI,EAAE,0BAA0B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;gBAChN,IAAI,IAAI,CAAC,GAAG,EAAE,IAAI,QAAQ;oBAAE,MAAM,IAAI,UAAU,CAAC,8BAA8B,EAAE,EAAE,IAAI,EAAE,yBAAyB,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;gBACpI,MAAM,KAAK,CAAC,GAAG,CAAC,CAAC;gBAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,KAAK,EAAE,MAAM,CAAC,CAAC;YAC1F,CAAC;YACD,IAAI,CAAC,MAAM;gBAAE,MAAM,IAAI,SAAS,CAAC,2CAA2C,CAAC,CAAC;YAC9E,OAAO,MAAM,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC;QAC1D,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,IAAI,CAAC;gBAAC,MAAM,IAAI,CAAC,aAAa,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;YAAC,CAAC;YAAC,MAAM,CAAC;gBACtD,MAAM,IAAI,UAAU,CAAC,sEAAsE,EAAE,EAAE,IAAI,EAAE,8BAA8B,EAAE,QAAQ,EAAE,EAAE,EAAE,KAAK,EAAE,KAAK,EAAE,CAAC,CAAC;YACrK,CAAC;YACD,MAAM,KAAK,CAAC;QACd,CAAC;IACH,CAAC;IAED,KAAK,CAAC,MAAM,CAAC,KAAwB;QACnC,MAAM,EAAE,MAAM,EAAE,GAAG,QAAQ,EAAE,GAAG,KAAK,CAAC;QACtC,KAAK,GAAG,EAAE,GAAG,eAAe,CAAC,QAAQ,CAAC,EAAE,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC;QACxE,aAAa,CAAC,KAAK,CAAC,CAAC;QACrB,MAAM,MAAM,GAAG,2BAA2B,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;QACzD,MAAM,MAAM,GAAG,MAAM,CAAC,cAAc,CAAC;QACrC,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,MAAM,IAAI,CAAC,KAAK,CAAC,WAAW,IAAI,KAAK,CAAC,WAAW,CAAC,cAAc,KAAK,GAAG;eACpF,MAAM,CAAC,KAAK,KAAK,OAAO,IAAI,MAAM,CAAC,YAAY,IAAI,MAAM,EAAE,IAAI,KAAK,SAAS;eAC7E,MAAM,CAAC,YAAY,KAAK,KAAK,CAAC,IAAI,CAAC,YAAY;eAC/C,UAAU,CAAC,MAAM,CAAC,UAAU,IAAI,IAAI,CAAC,KAAK,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,IAAI,IAAI,CAAC;eACnF,KAAK,CAAC,IAAI,CAAC,UAAU,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,OAAO,KAAK,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,IAAI,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,UAAU,KAAK,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,UAAU,CAAC;eACxJ,MAAM,CAAC,UAAU,KAAK,MAAM,CAAC,UAAU,IAAI,MAAM,CAAC,OAAO,KAAK,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,IAAI,MAAM,CAAC,UAAU,KAAK,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,UAAU,EAAE,CAAC;YAChJ,MAAM,IAAI,UAAU,CAAC,uEAAuE,EAAE,EAAE,IAAI,EAAE,8BAA8B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;QACxJ,CAAC;QACD,KAAK,CAAC,MAAM,EAAE,cAAc,EAAE,CAAC;QAC/B,MAAM,KAAK,GAAG,EAAE,SAAS,EAAE,MAAM,CAAC,UAAU,EAAE,OAAO,EAAE,WAAW,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,WAAW,EAAE,KAAK,CAAC,IAAI,CAAC,YAAY,EAAE,CAAC;QACvH,4EAA4E;QAC5E,8EAA8E;QAC9E,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,KAAK,EAAE,MAAM,CAAC,CAAC;QAC3E,IAAI,MAAM,CAAC,WAAW,KAAK,UAAU,CAAC,EAAE,SAAS,EAAE,KAAK,CAAC,SAAS,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO;YACxF,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO,EAAE,IAAI,EAAE,KAAK,CAAC,IAAI,EAAE,CAAC;YAAE,MAAM,IAAI,SAAS,CAAC,yCAAyC,CAAC,CAAC;QAClI,OAAO,IAAI,CAAC,OAAO,CAAC,EAAE,GAAG,KAAK,EAAE,SAAS,EAAE,MAAM,CAAC,UAAU,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,MAAM,EAAE,EAAE,WAAW,EAAE,KAAK,CAAC,WAAW,EAAE,CAAC,CAAC;IAC7H,CAAC;IAEO,KAAK,CAAC,OAAO,CAAC,KAAwB,EAAE,KAAyB,EAAE,MAAkB,EAAE,MAAqB,EAClH,QAAyD;QACzD,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAAC,MAAM,MAAM,GAAG,MAAM,CAAC,OAAO,CAAC;QAC5D,IAAI,MAAM,CAAC,KAAK,KAAK,OAAO,IAAI,MAAM,CAAC,iBAAiB,IAAI,CAAC,MAAM,IAAI,CAAC,MAAM;eACzE,CAAC,gBAAgB,CAAC,IAAI,CAAC,MAAM,CAAC,WAAW,CAAC,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC;eACvF,MAAM,CAAC,SAAS,KAAK,SAAS,KAAK,CAAC,KAAK,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,EAAE;YAAE,MAAM,IAAI,SAAS,CAAC,+DAA+D,CAAC,CAAC;QAC7J,MAAM,OAAO,GAAG,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,IAAI,EAAE,MAAM,CAAC,IAAI,CAAC,CAAC;QACjE,MAAM,OAAO,GAAG,EAAE,aAAa,EAAE,UAAU,MAAM,CAAC,WAAW,EAAE,EAAE,CAAC;QAClE,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,OAAO,YAAY,EAAE,EAAE,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,EAAE,WAAW,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;QAC9H,MAAM,SAAS,GAAG,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,MAAM,CAAC,IAAI,EAAwC,CAAC,CAAC,CAAC,IAAI,CAAC;QAC/F,IAAI,CAAC,SAAS,EAAE,IAAI,EAAE,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,EAAE,KAAK,MAAM,CAAC,SAAS,CAAC;YAAE,MAAM,IAAI,UAAU,CAAC,uCAAuC,EAAE,EAAE,IAAI,EAAE,4BAA4B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;QACxL,MAAM,QAAQ,GAA+B,EAAE,IAAI,EAAE,YAAY,EAAE,OAAO,EAAE,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,EAAE,UAAU,EAAE,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,UAAU;YAC1I,kBAAkB,EAAE,UAAU,CAAC,MAAM,CAAC,EAAE,cAAc,EAAE,MAAM,CAAC,aAAa,EAAE,eAAe,EAAE,MAAM,CAAC,cAAc,EAAE,kBAAkB,EAAE,MAAM,CAAC,gBAAgB,EAAE,CAAC;QACtK,MAAM,kBAAkB,GAAG,iCAAiC,CAAC,MAAM,CAAC,UAAU,EAAE,QAAQ,EAAE,MAAM,CAAC,OAAO,IAAI,IAAI,EAAE,KAAK,CAAC,IAAI,EAAE,KAAK,CAAC,OAAO,EACzI,EAAE,OAAO,EAAE,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,EAAE,UAAU,EAAE,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,UAAU,EAAE,CAAC,CAAC;QACjF,IAAI,WAAW,GAAG,kBAAkB,CAAC;QACrC,IAAI,QAAQ,EAAE,CAAC;YACb,MAAM,KAAK,GAAG,QAAQ,CAAC,WAAW,CAAC;YACnC,IAAI,CAAC,KAAK,IAAI,OAAO,KAAK,CAAC,WAAW,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC,IAAI,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,WAAW,CAAC,GAAG,IAAI,CAAC,GAAG,EAAE;mBAC/I,UAAU,CAAC,EAAE,GAAG,kBAAkB,EAAE,WAAW,EAAE,KAAK,CAAC,WAAW,EAAE,CAAC,KAAK,UAAU,CAAC,KAAK,CAAC,EAAE,CAAC;gBACjG,MAAM,IAAI,UAAU,CAAC,kEAAkE,EAAE,EAAE,IAAI,EAAE,4BAA4B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;YACjJ,CAAC;YACD,yEAAyE;YACzE,WAAW,GAAG,eAAe,CAAC,KAAK,CAAC,CAAC;QACvC,CAAC;aAAM,CAAC;YACN,MAAM,mBAAmB,CAAC,IAAI,CAAC,OAAO,EAAE,OAAO,EAAE,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,SAAS,EAAE,KAAK,CAAC,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,CAAC;YAChH,MAAM,OAAO,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,OAAO,cAAc,EAAE,EAAE,OAAO,EAAE,EAAE,aAAa,EAAE,UAAU,MAAM,CAAC,UAAU,EAAE,EAAE,EAAE,MAAM,EAAE,WAAW,CAAC,OAAO,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC;YACjK,IAAI,CAAC,OAAO,CAAC,EAAE;gBAAE,MAAM,IAAI,UAAU,CAAC,8CAA8C,EAAE,EAAE,IAAI,EAAE,4BAA4B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;QAC9I,CAAC;QACD,KAAK,CAAC,MAAM,EAAE,cAAc,EAAE,CAAC;QAC/B,MAAM,YAAY,GAAG,MAAM,CAAC;QAC5B,IAAI,QAAmC,CAAC;QACxC,OAAO,EAAE,cAAc,EAAE,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,UAAU,EAAE,MAAM,CAAC,SAAS,EAAE,YAAY,EAAE,MAAM,CAAC,WAAW,EAAE,WAAW,EAAE,MAAM,CAAC,UAAU,EAAE,WAAW;YAC7J,WAAW,EAAE,KAAK,IAAI,EAAE;gBACtB,MAAM,OAAO,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,KAAK,EAAE,YAAY,CAAC,CAAC;gBACnF,IAAI,OAAO,CAAC,KAAK,KAAK,OAAO,IAAI,OAAO,CAAC,iBAAiB;oBAAE,MAAM,IAAI,UAAU,CAAC,uCAAuC,EAAE,EAAE,IAAI,EAAE,0BAA0B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;gBAC9K,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,IAAI,EAAE,MAAM,CAAC,IAAI,CAAC,CAAC;gBACjD,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,OAAO,SAAS,EAAE,EAAE,MAAM,EAAE,WAAW,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;gBAC/F,IAAI,CAAC,MAAM,CAAC,EAAE;oBAAE,MAAM,IAAI,UAAU,CAAC,iCAAiC,EAAE,EAAE,IAAI,EAAE,6BAA6B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;YACjI,CAAC;YACD,IAAI,EAAE,GAAG,EAAE,CAAC,QAAQ,KAAK,IAAI,CAAC,aAAa,CAAC,KAAK,EAAE,YAAY,CAAC,CAAC,OAAO,CAAC,GAAG,EAAE,GAAG,QAAQ,GAAG,SAAS,CAAC,CAAC,CAAC,CAAC;SAC1G,CAAC;IACJ,CAAC;IAEO,KAAK,CAAC,aAAa,CAAC,KAAyB,EAAE,MAAsB;QAC3E,MAAM,OAAO,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,KAAK,EAAE,MAAM,CAAC,CAAC;QAC1E,IAAI,CAAC,OAAO,CAAC,iBAAiB,IAAI,CAAC,CAAC,SAAS,EAAE,QAAQ,CAAC,CAAC,QAAQ,CAAC,OAAO,CAAC,KAAK,CAAC;YAAE,MAAM,IAAI,UAAU,CAAC,yCAAyC,EAAE,EAAE,IAAI,EAAE,8BAA8B,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC;IAC5M,CAAC;IAED,KAAK,CAAC,UAAU,CAAC,IAAY,EAAE,MAAgC;QAC7D,IAAI,MAAM,CAAC,YAAY,IAAI,MAAM,CAAC,cAAc,EAAE,IAAI,KAAK,QAAQ;YAAE,OAAO,WAAW,CAAC;QACxF,MAAM,MAAM,GAAG,MAAM,CAAC,cAAc,CAAC;QACrC,IAAI,MAAM,IAAI,CAAC,MAAM,CAAC,OAAO,KAAK,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,IAAI,MAAM,CAAC,UAAU,KAAK,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,UAAU,CAAC;YAAE,OAAO,WAAW,CAAC;QACpI,IAAI,CAAC;YAAC,MAAM,IAAI,CAAC,aAAa,CAAC,EAAE,SAAS,EAAE,MAAM,CAAC,UAAU,EAAE,OAAO,EAAE,WAAW,CAAC,IAAI,CAAC,EAAE,WAAW,EAAE,MAAM,CAAC,YAAY,EAAE,EAAE,MAAM,CAAC,CAAC;YAAC,OAAO,SAAS,CAAC;QAAC,CAAC;QAC3J,MAAM,CAAC;YAAC,OAAO,WAAW,CAAC;QAAC,CAAC;IAC/B,CAAC;CACF;AAED,SAAS,aAAa,CAAC,KAA4D;IACjF,uBAAuB,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;IAAC,6BAA6B,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IAAC,0BAA0B,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;IAC3H,IAAI,KAAK,CAAC,OAAO,CAAC,cAAc,KAAK,GAAG,IAAI,KAAK,CAAC,OAAO,CAAC,OAAO,CAAC,IAAI,KAAK,YAAY;WAClF,KAAK,CAAC,IAAI,CAAC,UAAU,KAAK,KAAK,CAAC,OAAO,CAAC,UAAU,IAAI,KAAK,CAAC,IAAI,CAAC,QAAQ,KAAK,KAAK,CAAC,KAAK,CAAC,QAAQ;QAAE,MAAM,IAAI,SAAS,CAAC,4DAA4D,CAAC,CAAC;AAC7L,CAAC"}
|
|
@@ -0,0 +1,29 @@
|
|
|
1
|
+
import { HitchError } from "../foundation/index.js";
|
|
2
|
+
/** Both launchers prove the same streaming/non-streaming protocol before use. */
|
|
3
|
+
export async function probeSGLangProtocol(request, baseUrl, token, wireModel, api) {
|
|
4
|
+
const headers = { Authorization: `Bearer ${token}`, "content-type": "application/json" };
|
|
5
|
+
for (const stream of [false, true]) {
|
|
6
|
+
const body = api === "responses"
|
|
7
|
+
? { model: wireModel, input: "Reply with one word.", max_output_tokens: 8, temperature: 0, store: false, stream }
|
|
8
|
+
: { model: wireModel, messages: [{ role: "user", content: "Reply with one word." }], max_tokens: 8, temperature: 0, stream };
|
|
9
|
+
const response = await request(`${baseUrl}/v1/${api === "responses" ? "responses" : "chat/completions"}`, {
|
|
10
|
+
method: "POST", headers, body: JSON.stringify(body), signal: AbortSignal.timeout(30_000),
|
|
11
|
+
});
|
|
12
|
+
if (!response.ok)
|
|
13
|
+
throw failure(`SGLang ${api} probe returned HTTP ${response.status}`);
|
|
14
|
+
if (stream && !response.headers.get("content-type")?.includes("text/event-stream"))
|
|
15
|
+
throw failure("SGLang streaming probe did not return SSE");
|
|
16
|
+
const chunks = stream ? (await response.text()).split(/\r?\n/).filter(line => line.startsWith("data: ") && line !== "data: [DONE]")
|
|
17
|
+
.map(line => JSON.parse(line.slice(6))) : [await response.json()];
|
|
18
|
+
if (api === "responses") {
|
|
19
|
+
const body = stream ? chunks.findLast(event => event.type === "response.completed" || event.type === "response.incomplete")?.response : chunks[0];
|
|
20
|
+
if (body?.status !== "completed" && body?.status !== "incomplete")
|
|
21
|
+
throw failure("SGLang Responses probe lacks a successful terminal payload");
|
|
22
|
+
}
|
|
23
|
+
else if (!chunks.some(chunk => chunk.choices?.some(choice => ["stop", "length"].includes(String(choice.finish_reason))))) {
|
|
24
|
+
throw failure("SGLang Chat Completions probe lacks terminal output");
|
|
25
|
+
}
|
|
26
|
+
}
|
|
27
|
+
}
|
|
28
|
+
function failure(message) { return new HitchError(message, { code: "inference_protocol_unsupported", exitCode: 12 }); }
|
|
29
|
+
//# sourceMappingURL=protocol-probe.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"protocol-probe.js","sourceRoot":"","sources":["../../../src/inference/protocol-probe.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,UAAU,EAAE,MAAM,wBAAwB,CAAC;AAEpD,iFAAiF;AACjF,MAAM,CAAC,KAAK,UAAU,mBAAmB,CAAC,OAAqB,EAAE,OAAe,EAAE,KAAa,EAAE,SAAiB,EAAE,GAAqC;IACvJ,MAAM,OAAO,GAAG,EAAE,aAAa,EAAE,UAAU,KAAK,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC;IACzF,KAAK,MAAM,MAAM,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,EAAE,CAAC;QACnC,MAAM,IAAI,GAAG,GAAG,KAAK,WAAW;YAC9B,CAAC,CAAC,EAAE,KAAK,EAAE,SAAS,EAAE,KAAK,EAAE,sBAAsB,EAAE,iBAAiB,EAAE,CAAC,EAAE,WAAW,EAAE,CAAC,EAAE,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE;YACjH,CAAC,CAAC,EAAE,KAAK,EAAE,SAAS,EAAE,QAAQ,EAAE,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,sBAAsB,EAAE,CAAC,EAAE,UAAU,EAAE,CAAC,EAAE,WAAW,EAAE,CAAC,EAAE,MAAM,EAAE,CAAC;QAC/H,MAAM,QAAQ,GAAG,MAAM,OAAO,CAAC,GAAG,OAAO,OAAO,GAAG,KAAK,WAAW,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,kBAAkB,EAAE,EAAE;YACxG,MAAM,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,WAAW,CAAC,OAAO,CAAC,MAAM,CAAC;SACzF,CAAC,CAAC;QACH,IAAI,CAAC,QAAQ,CAAC,EAAE;YAAE,MAAM,OAAO,CAAC,UAAU,GAAG,wBAAwB,QAAQ,CAAC,MAAM,EAAE,CAAC,CAAC;QACxF,IAAI,MAAM,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,EAAE,QAAQ,CAAC,mBAAmB,CAAC;YAAE,MAAM,OAAO,CAAC,2CAA2C,CAAC,CAAC;QAC/I,MAAM,MAAM,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,UAAU,CAAC,QAAQ,CAAC,IAAI,IAAI,KAAK,cAAc,CAAC;aAChI,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAA4B,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,QAAQ,CAAC,IAAI,EAA6B,CAAC,CAAC;QAC1H,IAAI,GAAG,KAAK,WAAW,EAAE,CAAC;YACxB,MAAM,IAAI,GAAG,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,IAAI,KAAK,oBAAoB,IAAI,KAAK,CAAC,IAAI,KAAK,qBAAqB,CAAC,EAAE,QAA4C,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC;YACtL,IAAI,IAAI,EAAE,MAAM,KAAK,WAAW,IAAI,IAAI,EAAE,MAAM,KAAK,YAAY;gBAAE,MAAM,OAAO,CAAC,4DAA4D,CAAC,CAAC;QACjJ,CAAC;aAAM,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAE,KAAK,CAAC,OAA0D,EAAE,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,MAAM,CAAC,aAAa,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;YAC/K,MAAM,OAAO,CAAC,qDAAqD,CAAC,CAAC;QACvE,CAAC;IACH,CAAC;AACH,CAAC;AACD,SAAS,OAAO,CAAC,OAAe,IAAgB,OAAO,IAAI,UAAU,CAAC,OAAO,EAAE,EAAE,IAAI,EAAE,gCAAgC,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC"}
|
|
@@ -0,0 +1,24 @@
|
|
|
1
|
+
import { HitchError } from "../foundation/index.js";
|
|
2
|
+
import { addModelManifest, verifyLocalModel } from "./model-store.js";
|
|
3
|
+
import { parseLocalModelManifest } from "./manifest.js";
|
|
4
|
+
import { loadModelNodeClient } from "./node-registry.js";
|
|
5
|
+
import { modelNodeSnapshot, parseModelSnapshotRef, rememberModelNodeSnapshot } from "./model-location.js";
|
|
6
|
+
export async function addModelFromNode(root, name, snapshot, binding, force = false) {
|
|
7
|
+
const snapshotRef = parseModelSnapshotRef(snapshot);
|
|
8
|
+
const client = await loadModelNodeClient(root, binding);
|
|
9
|
+
const model = parseLocalModelManifest(await client.call("cas.hfManifest", { snapshotRef }));
|
|
10
|
+
await rememberModelNodeSnapshot(root, model, binding, snapshotRef);
|
|
11
|
+
return addModelManifest({ root, manifest: model, name, force });
|
|
12
|
+
}
|
|
13
|
+
export async function verifyModelOnNode(root, model, binding) {
|
|
14
|
+
const snapshotRef = await modelNodeSnapshot(root, model, binding);
|
|
15
|
+
if (!snapshotRef) {
|
|
16
|
+
await verifyLocalModel(root, model);
|
|
17
|
+
return;
|
|
18
|
+
}
|
|
19
|
+
const client = await loadModelNodeClient(root, binding);
|
|
20
|
+
const observed = parseLocalModelManifest(await client.call("cas.hfManifest", { snapshotRef }));
|
|
21
|
+
if (observed.model_id !== model.model_id)
|
|
22
|
+
throw new HitchError("model-node HF contents differ from the registered model", { code: "local_model_integrity_failed", exitCode: 5 });
|
|
23
|
+
}
|
|
24
|
+
//# sourceMappingURL=remote-model-store.js.map
|