@atomic-chat/core 0.5.1 → 0.13.6
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 +22 -0
- package/dist/app-daemon.js +20 -67
- package/dist/app-daemon.js.map +1 -1
- package/dist/backend/advisor/advisor.d.ts +132 -0
- package/dist/backend/advisor/advisor.js +537 -0
- package/dist/backend/advisor/advisor.js.map +1 -0
- package/dist/backend/advisor/index.d.ts +6 -0
- package/dist/backend/advisor/index.js +7 -0
- package/dist/backend/advisor/index.js.map +1 -0
- package/dist/backend/advisor/policy.d.ts +81 -0
- package/dist/backend/advisor/policy.js +154 -0
- package/dist/backend/advisor/policy.js.map +1 -0
- package/dist/backend/catalog/archive.d.ts +21 -7
- package/dist/backend/catalog/archive.js +57 -17
- package/dist/backend/catalog/archive.js.map +1 -1
- package/dist/backend/catalog/bundled-manifest-baseline.js +42 -26
- package/dist/backend/catalog/bundled-manifest-baseline.js.map +1 -1
- package/dist/backend/catalog/cuda-family.d.ts +12 -2
- package/dist/backend/catalog/cuda-family.js +30 -7
- package/dist/backend/catalog/cuda-family.js.map +1 -1
- package/dist/backend/catalog/index.d.ts +5 -0
- package/dist/backend/catalog/index.js +5 -0
- package/dist/backend/catalog/index.js.map +1 -1
- package/dist/backend/catalog/manifest.d.ts +5 -3
- package/dist/backend/catalog/manifest.js +12 -9
- package/dist/backend/catalog/manifest.js.map +1 -1
- package/dist/backend/catalog/migrate.d.ts +8 -6
- package/dist/backend/catalog/migrate.js +26 -15
- package/dist/backend/catalog/migrate.js.map +1 -1
- package/dist/backend/catalog/prism-catalog.d.ts +42 -0
- package/dist/backend/catalog/prism-catalog.js +101 -0
- package/dist/backend/catalog/prism-catalog.js.map +1 -0
- package/dist/backend/catalog/prism-manifest-baseline.d.ts +2 -0
- package/dist/backend/catalog/prism-manifest-baseline.js +11 -0
- package/dist/backend/catalog/prism-manifest-baseline.js.map +1 -0
- package/dist/backend/catalog/prism-manifest-baseline.json +134 -0
- package/dist/backend/catalog/prism-manifest.d.ts +90 -0
- package/dist/backend/catalog/prism-manifest.js +171 -0
- package/dist/backend/catalog/prism-manifest.js.map +1 -0
- package/dist/backend/catalog/turboquant-catalog.d.ts +60 -0
- package/dist/backend/catalog/turboquant-catalog.js +162 -0
- package/dist/backend/catalog/turboquant-catalog.js.map +1 -0
- package/dist/backend/catalog/turboquant-index.d.ts +104 -0
- package/dist/backend/catalog/turboquant-index.js +260 -0
- package/dist/backend/catalog/turboquant-index.js.map +1 -0
- package/dist/backend/index.d.ts +1 -0
- package/dist/backend/index.js +1 -0
- package/dist/backend/index.js.map +1 -1
- package/dist/backend/install/runtime-backend.d.ts +6 -6
- package/dist/backend/install/runtime-backend.js +23 -20
- package/dist/backend/install/runtime-backend.js.map +1 -1
- package/dist/backend/install/service.d.ts +106 -5
- package/dist/backend/install/service.js +247 -21
- package/dist/backend/install/service.js.map +1 -1
- package/dist/backend/installed/bundled.d.ts +19 -0
- package/dist/backend/installed/bundled.js +45 -0
- package/dist/backend/installed/bundled.js.map +1 -0
- package/dist/backend/installed/index.d.ts +1 -0
- package/dist/backend/installed/index.js +1 -0
- package/dist/backend/installed/index.js.map +1 -1
- package/dist/backend/installed/installed.js +3 -0
- package/dist/backend/installed/installed.js.map +1 -1
- package/dist/backend/optimal/optimal-cache.d.ts +27 -8
- package/dist/backend/optimal/optimal-cache.js +38 -19
- package/dist/backend/optimal/optimal-cache.js.map +1 -1
- package/dist/backend/optimal/optimal-store.js +1 -1
- package/dist/backend/optimal/optimal-store.js.map +1 -1
- package/dist/backend/select/categories.d.ts +2 -1
- package/dist/backend/select/categories.js +8 -1
- package/dist/backend/select/categories.js.map +1 -1
- package/dist/backend/select/features.d.ts +13 -5
- package/dist/backend/select/features.js +36 -6
- package/dist/backend/select/features.js.map +1 -1
- package/dist/backend/select/index.d.ts +2 -0
- package/dist/backend/select/index.js +2 -0
- package/dist/backend/select/index.js.map +1 -1
- package/dist/backend/select/prism.d.ts +54 -0
- package/dist/backend/select/prism.js +212 -0
- package/dist/backend/select/prism.js.map +1 -0
- package/dist/backend/select/recovery.d.ts +4 -2
- package/dist/backend/select/recovery.js +17 -5
- package/dist/backend/select/recovery.js.map +1 -1
- package/dist/backend/select/tiers.d.ts +20 -11
- package/dist/backend/select/tiers.js +49 -11
- package/dist/backend/select/tiers.js.map +1 -1
- package/dist/backend/select/turboquant-tiers.d.ts +45 -0
- package/dist/backend/select/turboquant-tiers.js +114 -0
- package/dist/backend/select/turboquant-tiers.js.map +1 -0
- package/dist/backend/types.d.ts +6 -69
- package/dist/cli/commands/backends.d.ts +10 -0
- package/dist/cli/commands/backends.js +142 -0
- package/dist/cli/commands/backends.js.map +1 -0
- package/dist/cli/commands/hardware.d.ts +13 -0
- package/dist/cli/commands/hardware.js +71 -0
- package/dist/cli/commands/hardware.js.map +1 -0
- package/dist/cli/commands/host-step.d.ts +28 -0
- package/dist/cli/commands/host-step.js +67 -0
- package/dist/cli/commands/host-step.js.map +1 -0
- package/dist/cli/commands/index.d.ts +4 -1
- package/dist/cli/commands/index.js +4 -1
- package/dist/cli/commands/index.js.map +1 -1
- package/dist/cli/commands/serve.js +4 -4
- package/dist/cli/commands/serve.js.map +1 -1
- package/dist/cli/commands/shared.d.ts +6 -1
- package/dist/cli/commands/shared.js +7 -0
- package/dist/cli/commands/shared.js.map +1 -1
- package/dist/cli/io.d.ts +11 -0
- package/dist/cli/io.js +28 -3
- package/dist/cli/io.js.map +1 -1
- package/dist/cli/launch.js +2 -2
- package/dist/cli/launch.js.map +1 -1
- package/dist/cli/main.d.ts +1 -1
- package/dist/cli/main.js +11 -1
- package/dist/cli/main.js.map +1 -1
- package/dist/client/control-client.d.ts +128 -7
- package/dist/client/control-client.js +211 -12
- package/dist/client/control-client.js.map +1 -1
- package/dist/cloud/registry.js +9 -1
- package/dist/cloud/registry.js.map +1 -1
- package/dist/config/paths.d.ts +137 -0
- package/dist/config/paths.js +163 -0
- package/dist/config/paths.js.map +1 -1
- package/dist/contracts/backend-advisor.d.ts +225 -0
- package/dist/contracts/backend-advisor.js +14 -0
- package/dist/contracts/backend-advisor.js.map +1 -0
- package/dist/contracts/control-api.d.ts +7 -1
- package/dist/contracts/control-api.js +7 -1
- package/dist/contracts/control-api.js.map +1 -1
- package/dist/contracts/decision.d.ts +394 -0
- package/dist/contracts/decision.js +69 -0
- package/dist/contracts/decision.js.map +1 -0
- package/dist/contracts/diffusion.d.ts +62 -9
- package/dist/contracts/embedding.d.ts +103 -0
- package/dist/contracts/embedding.js +24 -0
- package/dist/contracts/embedding.js.map +1 -0
- package/dist/contracts/engine-builds.d.ts +108 -0
- package/dist/contracts/engine-builds.js +11 -0
- package/dist/contracts/engine-builds.js.map +1 -0
- package/dist/contracts/engines.d.ts +157 -0
- package/dist/contracts/engines.js +50 -0
- package/dist/contracts/engines.js.map +1 -0
- package/dist/contracts/environment.d.ts +774 -0
- package/dist/contracts/environment.js +116 -0
- package/dist/contracts/environment.js.map +1 -0
- package/dist/contracts/errors.d.ts +43 -2
- package/dist/contracts/errors.js +41 -0
- package/dist/contracts/errors.js.map +1 -1
- package/dist/contracts/events.d.ts +82 -4
- package/dist/contracts/hardware.d.ts +79 -6
- package/dist/contracts/hardware.js +9 -5
- package/dist/contracts/hardware.js.map +1 -1
- package/dist/contracts/index.d.ts +8 -0
- package/dist/contracts/index.js +8 -0
- package/dist/contracts/index.js.map +1 -1
- package/dist/contracts/model-compatibility.d.ts +97 -0
- package/dist/contracts/model-compatibility.js +6 -0
- package/dist/contracts/model-compatibility.js.map +1 -0
- package/dist/contracts/model-setup.d.ts +115 -0
- package/dist/contracts/model-setup.js +33 -0
- package/dist/contracts/model-setup.js.map +1 -0
- package/dist/contracts/model-yml.d.ts +14 -0
- package/dist/contracts/model-yml.js +1 -0
- package/dist/contracts/model-yml.js.map +1 -1
- package/dist/contracts/session.d.ts +45 -2
- package/dist/contracts/session.js +14 -1
- package/dist/contracts/session.js.map +1 -1
- package/dist/contracts/settings.d.ts +44 -1
- package/dist/contracts/settings.js.map +1 -1
- package/dist/core/atomic-core.d.ts +46 -5
- package/dist/core/atomic-core.js +25 -0
- package/dist/core/atomic-core.js.map +1 -1
- package/dist/core/create.js +523 -23
- package/dist/core/create.js.map +1 -1
- package/dist/core/gpu/index.d.ts +6 -0
- package/dist/core/gpu/index.js +7 -0
- package/dist/core/gpu/index.js.map +1 -0
- package/dist/core/gpu/policy.d.ts +68 -0
- package/dist/core/gpu/policy.js +99 -0
- package/dist/core/gpu/policy.js.map +1 -0
- package/dist/core/gpu/residency.d.ts +44 -0
- package/dist/core/gpu/residency.js +101 -0
- package/dist/core/gpu/residency.js.map +1 -0
- package/dist/core/gpu-residency.d.ts +39 -0
- package/dist/core/gpu-residency.js +32 -0
- package/dist/core/gpu-residency.js.map +1 -0
- package/dist/core/index.d.ts +1 -1
- package/dist/core/managed-engines.d.ts +207 -0
- package/dist/core/managed-engines.js +579 -0
- package/dist/core/managed-engines.js.map +1 -0
- package/dist/core/managed-environment.d.ts +72 -0
- package/dist/core/managed-environment.js +218 -0
- package/dist/core/managed-environment.js.map +1 -0
- package/dist/core/model-setup/index.d.ts +4 -0
- package/dist/core/model-setup/index.js +5 -0
- package/dist/core/model-setup/index.js.map +1 -0
- package/dist/core/model-setup/plan.d.ts +56 -0
- package/dist/core/model-setup/plan.js +158 -0
- package/dist/core/model-setup/plan.js.map +1 -0
- package/dist/core/model-setup/service.d.ts +92 -0
- package/dist/core/model-setup/service.js +329 -0
- package/dist/core/model-setup/service.js.map +1 -0
- package/dist/core/model-setup/store.d.ts +29 -0
- package/dist/core/model-setup/store.js +116 -0
- package/dist/core/model-setup/store.js.map +1 -0
- package/dist/core/model-setup/wiring.d.ts +59 -0
- package/dist/core/model-setup/wiring.js +132 -0
- package/dist/core/model-setup/wiring.js.map +1 -0
- package/dist/core/prism.d.ts +38 -0
- package/dist/core/prism.js +43 -0
- package/dist/core/prism.js.map +1 -0
- package/dist/core/public-server.d.ts +7 -0
- package/dist/core/public-server.js +5 -2
- package/dist/core/public-server.js.map +1 -1
- package/dist/core/sessions.d.ts +6 -1
- package/dist/core/sessions.js +18 -14
- package/dist/core/sessions.js.map +1 -1
- package/dist/core/tensorrt-llm.d.ts +169 -0
- package/dist/core/tensorrt-llm.js +428 -0
- package/dist/core/tensorrt-llm.js.map +1 -0
- package/dist/core/types.d.ts +43 -1
- package/dist/core/types.js.map +1 -1
- package/dist/decision/args.d.ts +110 -0
- package/dist/decision/args.js +126 -0
- package/dist/decision/args.js.map +1 -0
- package/dist/decision/backoff.d.ts +25 -0
- package/dist/decision/backoff.js +33 -0
- package/dist/decision/backoff.js.map +1 -0
- package/dist/decision/engine-candidates.d.ts +52 -0
- package/dist/decision/engine-candidates.js +138 -0
- package/dist/decision/engine-candidates.js.map +1 -0
- package/dist/decision/engine-fallback.d.ts +41 -0
- package/dist/decision/engine-fallback.js +57 -0
- package/dist/decision/engine-fallback.js.map +1 -0
- package/dist/decision/engine-version.d.ts +27 -0
- package/dist/decision/engine-version.js +54 -0
- package/dist/decision/engine-version.js.map +1 -0
- package/dist/decision/engine.d.ts +114 -0
- package/dist/decision/engine.js +253 -0
- package/dist/decision/engine.js.map +1 -0
- package/dist/decision/http.d.ts +36 -0
- package/dist/decision/http.js +73 -0
- package/dist/decision/http.js.map +1 -0
- package/dist/decision/index.d.ts +20 -0
- package/dist/decision/index.js +21 -0
- package/dist/decision/index.js.map +1 -0
- package/dist/decision/model-facts.d.ts +32 -0
- package/dist/decision/model-facts.js +40 -0
- package/dist/decision/model-facts.js.map +1 -0
- package/dist/decision/outcome.d.ts +35 -0
- package/dist/decision/outcome.js +105 -0
- package/dist/decision/outcome.js.map +1 -0
- package/dist/decision/process.d.ts +71 -0
- package/dist/decision/process.js +142 -0
- package/dist/decision/process.js.map +1 -0
- package/dist/decision/readiness.d.ts +74 -0
- package/dist/decision/readiness.js +165 -0
- package/dist/decision/readiness.js.map +1 -0
- package/dist/decision/request.d.ts +11 -0
- package/dist/decision/request.js +56 -0
- package/dist/decision/request.js.map +1 -0
- package/dist/decision/service.d.ts +202 -0
- package/dist/decision/service.js +711 -0
- package/dist/decision/service.js.map +1 -0
- package/dist/decision/upstream-version.d.ts +23 -0
- package/dist/decision/upstream-version.js +34 -0
- package/dist/decision/upstream-version.js.map +1 -0
- package/dist/decision/wiring.d.ts +67 -0
- package/dist/decision/wiring.js +101 -0
- package/dist/decision/wiring.js.map +1 -0
- package/dist/diffusion/args.d.ts +7 -4
- package/dist/diffusion/args.js +31 -7
- package/dist/diffusion/args.js.map +1 -1
- package/dist/diffusion/compat.d.ts +1 -1
- package/dist/diffusion/compat.js +6 -2
- package/dist/diffusion/compat.js.map +1 -1
- package/dist/diffusion/image-job.js +1 -1
- package/dist/diffusion/image-job.js.map +1 -1
- package/dist/diffusion/install.d.ts +35 -11
- package/dist/diffusion/install.js +74 -49
- package/dist/diffusion/install.js.map +1 -1
- package/dist/diffusion/job-kind.d.ts +33 -7
- package/dist/diffusion/jobs.d.ts +6 -1
- package/dist/diffusion/jobs.js +64 -18
- package/dist/diffusion/jobs.js.map +1 -1
- package/dist/diffusion/parse.d.ts +1 -2
- package/dist/diffusion/parse.js +1 -13
- package/dist/diffusion/parse.js.map +1 -1
- package/dist/diffusion/progress.js +3 -0
- package/dist/diffusion/progress.js.map +1 -1
- package/dist/diffusion/server-process.d.ts +6 -0
- package/dist/diffusion/server-process.js +3 -2
- package/dist/diffusion/server-process.js.map +1 -1
- package/dist/diffusion/service.d.ts +52 -6
- package/dist/diffusion/service.js +215 -40
- package/dist/diffusion/service.js.map +1 -1
- package/dist/diffusion/session.d.ts +27 -3
- package/dist/diffusion/session.js +108 -17
- package/dist/diffusion/session.js.map +1 -1
- package/dist/diffusion/state.d.ts +19 -1
- package/dist/diffusion/state.js +10 -0
- package/dist/diffusion/state.js.map +1 -1
- package/dist/diffusion/tracker.d.ts +21 -1
- package/dist/diffusion/tracker.js +43 -6
- package/dist/diffusion/tracker.js.map +1 -1
- package/dist/diffusion/types.d.ts +7 -0
- package/dist/diffusion/types.js +11 -0
- package/dist/diffusion/types.js.map +1 -1
- package/dist/diffusion/validate.js +2 -2
- package/dist/diffusion/validate.js.map +1 -1
- package/dist/diffusion/video-estimate.d.ts +214 -0
- package/dist/diffusion/video-estimate.js +402 -0
- package/dist/diffusion/video-estimate.js.map +1 -0
- package/dist/diffusion/video-eta.d.ts +90 -0
- package/dist/diffusion/video-eta.js +248 -0
- package/dist/diffusion/video-eta.js.map +1 -0
- package/dist/diffusion/video-history.d.ts +45 -0
- package/dist/diffusion/video-history.js +121 -0
- package/dist/diffusion/video-history.js.map +1 -0
- package/dist/diffusion/video-job.d.ts +13 -5
- package/dist/diffusion/video-job.js +42 -15
- package/dist/diffusion/video-job.js.map +1 -1
- package/dist/diffusion/video-recipe.js +3 -0
- package/dist/diffusion/video-recipe.js.map +1 -1
- package/dist/diffusion/video-tiling.d.ts +54 -0
- package/dist/diffusion/video-tiling.js +70 -0
- package/dist/diffusion/video-tiling.js.map +1 -0
- package/dist/diffusion/wiring.d.ts +14 -1
- package/dist/diffusion/wiring.js +7 -0
- package/dist/diffusion/wiring.js.map +1 -1
- package/dist/diffusion/workflow.d.ts +8 -2
- package/dist/diffusion/workflow.js +13 -5
- package/dist/diffusion/workflow.js.map +1 -1
- package/dist/downloads/downloader.d.ts +22 -0
- package/dist/downloads/downloader.js +57 -2
- package/dist/downloads/downloader.js.map +1 -1
- package/dist/downloads/protocol.d.ts +3 -8
- package/dist/downloads/protocol.js.map +1 -1
- package/dist/downloads/segmented.d.ts +40 -0
- package/dist/downloads/segmented.js +170 -0
- package/dist/downloads/segmented.js.map +1 -0
- package/dist/embedding/args.d.ts +39 -0
- package/dist/embedding/args.js +46 -0
- package/dist/embedding/args.js.map +1 -0
- package/dist/embedding/engine.d.ts +50 -0
- package/dist/embedding/engine.js +100 -0
- package/dist/embedding/engine.js.map +1 -0
- package/dist/embedding/ffmpeg.d.ts +18 -0
- package/dist/embedding/ffmpeg.js +63 -0
- package/dist/embedding/ffmpeg.js.map +1 -0
- package/dist/embedding/index.d.ts +16 -0
- package/dist/embedding/index.js +17 -0
- package/dist/embedding/index.js.map +1 -0
- package/dist/embedding/model-facts.d.ts +51 -0
- package/dist/embedding/model-facts.js +84 -0
- package/dist/embedding/model-facts.js.map +1 -0
- package/dist/embedding/process.d.ts +70 -0
- package/dist/embedding/process.js +140 -0
- package/dist/embedding/process.js.map +1 -0
- package/dist/embedding/readiness.d.ts +56 -0
- package/dist/embedding/readiness.js +114 -0
- package/dist/embedding/readiness.js.map +1 -0
- package/dist/embedding/request.d.ts +43 -0
- package/dist/embedding/request.js +285 -0
- package/dist/embedding/request.js.map +1 -0
- package/dist/embedding/service.d.ts +152 -0
- package/dist/embedding/service.js +602 -0
- package/dist/embedding/service.js.map +1 -0
- package/dist/embedding/wiring.d.ts +51 -0
- package/dist/embedding/wiring.js +79 -0
- package/dist/embedding/wiring.js.map +1 -0
- package/dist/engine-builds/builds.d.ts +54 -0
- package/dist/engine-builds/builds.js +151 -0
- package/dist/engine-builds/builds.js.map +1 -0
- package/dist/engine-builds/index.d.ts +10 -0
- package/dist/engine-builds/index.js +11 -0
- package/dist/engine-builds/index.js.map +1 -0
- package/dist/engine-builds/install.d.ts +42 -0
- package/dist/engine-builds/install.js +100 -0
- package/dist/engine-builds/install.js.map +1 -0
- package/dist/engine-builds/manifest.d.ts +121 -0
- package/dist/engine-builds/manifest.js +254 -0
- package/dist/engine-builds/manifest.js.map +1 -0
- package/dist/engine-builds/order.d.ts +24 -0
- package/dist/engine-builds/order.js +44 -0
- package/dist/engine-builds/order.js.map +1 -0
- package/dist/engine-builds/select.d.ts +123 -0
- package/dist/engine-builds/select.js +223 -0
- package/dist/engine-builds/select.js.map +1 -0
- package/dist/engine-builds/service.d.ts +123 -0
- package/dist/engine-builds/service.js +392 -0
- package/dist/engine-builds/service.js.map +1 -0
- package/dist/engines/descriptor-order.d.ts +28 -0
- package/dist/engines/descriptor-order.js +60 -0
- package/dist/engines/descriptor-order.js.map +1 -0
- package/dist/engines/engine-build.d.ts +23 -0
- package/dist/engines/engine-build.js +58 -0
- package/dist/engines/engine-build.js.map +1 -0
- package/dist/engines/index.d.ts +10 -0
- package/dist/engines/index.js +11 -0
- package/dist/engines/index.js.map +1 -0
- package/dist/engines/llamacpp.d.ts +71 -0
- package/dist/engines/llamacpp.js +176 -0
- package/dist/engines/llamacpp.js.map +1 -0
- package/dist/engines/managed.d.ts +49 -0
- package/dist/engines/managed.js +88 -0
- package/dist/engines/managed.js.map +1 -0
- package/dist/engines/service.d.ts +46 -0
- package/dist/engines/service.js +61 -0
- package/dist/engines/service.js.map +1 -0
- package/dist/engines/versions.d.ts +65 -0
- package/dist/engines/versions.js +259 -0
- package/dist/engines/versions.js.map +1 -0
- package/dist/hardware/cpu-flags.d.ts +44 -0
- package/dist/hardware/cpu-flags.js +206 -0
- package/dist/hardware/cpu-flags.js.map +1 -0
- package/dist/hardware/drm-sysfs.d.ts +51 -0
- package/dist/hardware/drm-sysfs.js +67 -0
- package/dist/hardware/drm-sysfs.js.map +1 -0
- package/dist/hardware/facts.d.ts +42 -0
- package/dist/hardware/facts.js +92 -0
- package/dist/hardware/facts.js.map +1 -0
- package/dist/hardware/index.d.ts +19 -3
- package/dist/hardware/index.js +19 -3
- package/dist/hardware/index.js.map +1 -1
- package/dist/hardware/merge.d.ts +55 -0
- package/dist/hardware/merge.js +141 -0
- package/dist/hardware/merge.js.map +1 -0
- package/dist/hardware/nvidia-smi.d.ts +43 -0
- package/dist/hardware/nvidia-smi.js +118 -0
- package/dist/hardware/nvidia-smi.js.map +1 -0
- package/dist/hardware/probe-common.d.ts +72 -0
- package/dist/hardware/probe-common.js +106 -0
- package/dist/hardware/probe-common.js.map +1 -0
- package/dist/hardware/probe-darwin.d.ts +20 -0
- package/dist/hardware/probe-darwin.js +76 -0
- package/dist/hardware/probe-darwin.js.map +1 -0
- package/dist/hardware/probe-linux.d.ts +9 -0
- package/dist/hardware/probe-linux.js +133 -0
- package/dist/hardware/probe-linux.js.map +1 -0
- package/dist/hardware/probe-windows.d.ts +7 -0
- package/dist/hardware/probe-windows.js +50 -0
- package/dist/hardware/probe-windows.js.map +1 -0
- package/dist/hardware/probe.d.ts +16 -0
- package/dist/hardware/probe.js +65 -0
- package/dist/hardware/probe.js.map +1 -0
- package/dist/hardware/service.d.ts +60 -0
- package/dist/hardware/service.js +114 -0
- package/dist/hardware/service.js.map +1 -0
- package/dist/hardware/vulkan.d.ts +47 -0
- package/dist/hardware/vulkan.js +148 -0
- package/dist/hardware/vulkan.js.map +1 -0
- package/dist/hardware/windows-video.d.ts +97 -0
- package/dist/hardware/windows-video.js +214 -0
- package/dist/hardware/windows-video.js.map +1 -0
- package/dist/host/app-daemon.d.ts +47 -0
- package/dist/host/app-daemon.js +146 -0
- package/dist/host/app-daemon.js.map +1 -0
- package/dist/host/index.d.ts +5 -3
- package/dist/host/index.js +5 -3
- package/dist/host/index.js.map +1 -1
- package/dist/host/log-file.d.ts +74 -0
- package/dist/host/log-file.js +214 -0
- package/dist/host/log-file.js.map +1 -0
- package/dist/host/recipes/enable-wsl.d.ts +63 -0
- package/dist/host/recipes/enable-wsl.js +73 -0
- package/dist/host/recipes/enable-wsl.js.map +1 -0
- package/dist/host/recipes/executor-io-windows.d.ts +69 -0
- package/dist/host/recipes/executor-io-windows.js +235 -0
- package/dist/host/recipes/executor-io-windows.js.map +1 -0
- package/dist/host/recipes/executor-io.d.ts +80 -0
- package/dist/host/recipes/executor-io.js +211 -0
- package/dist/host/recipes/executor-io.js.map +1 -0
- package/dist/host/recipes/executor.d.ts +81 -0
- package/dist/host/recipes/executor.js +678 -0
- package/dist/host/recipes/executor.js.map +1 -0
- package/dist/host/recipes/guest-executor.d.ts +29 -0
- package/dist/host/recipes/guest-executor.js +76 -0
- package/dist/host/recipes/guest-executor.js.map +1 -0
- package/dist/host/recipes/index.d.ts +16 -0
- package/dist/host/recipes/index.js +17 -0
- package/dist/host/recipes/index.js.map +1 -0
- package/dist/host/recipes/install-container-runtime.d.ts +389 -0
- package/dist/host/recipes/install-container-runtime.js +796 -0
- package/dist/host/recipes/install-container-runtime.js.map +1 -0
- package/dist/host/recipes/openpgp.d.ts +23 -0
- package/dist/host/recipes/openpgp.js +153 -0
- package/dist/host/recipes/openpgp.js.map +1 -0
- package/dist/host/recipes/request-file.d.ts +96 -0
- package/dist/host/recipes/request-file.js +103 -0
- package/dist/host/recipes/request-file.js.map +1 -0
- package/dist/lock/process-journal.d.ts +5 -2
- package/dist/lock/process-journal.js.map +1 -1
- package/dist/models/capabilities.d.ts +7 -0
- package/dist/models/capabilities.js +17 -1
- package/dist/models/capabilities.js.map +1 -1
- package/dist/models/compatibility/index.d.ts +10 -0
- package/dist/models/compatibility/index.js +11 -0
- package/dist/models/compatibility/index.js.map +1 -0
- package/dist/models/compatibility/inspect.d.ts +28 -0
- package/dist/models/compatibility/inspect.js +90 -0
- package/dist/models/compatibility/inspect.js.map +1 -0
- package/dist/models/compatibility/resolve.d.ts +59 -0
- package/dist/models/compatibility/resolve.js +219 -0
- package/dist/models/compatibility/resolve.js.map +1 -0
- package/dist/models/compatibility/rules-baseline.d.ts +2 -0
- package/dist/models/compatibility/rules-baseline.js +10 -0
- package/dist/models/compatibility/rules-baseline.js.map +1 -0
- package/dist/models/compatibility/rules-baseline.json +292 -0
- package/dist/models/compatibility/rules-service.d.ts +36 -0
- package/dist/models/compatibility/rules-service.js +109 -0
- package/dist/models/compatibility/rules-service.js.map +1 -0
- package/dist/models/compatibility/rules.d.ts +81 -0
- package/dist/models/compatibility/rules.js +164 -0
- package/dist/models/compatibility/rules.js.map +1 -0
- package/dist/models/compatibility/service.d.ts +62 -0
- package/dist/models/compatibility/service.js +138 -0
- package/dist/models/compatibility/service.js.map +1 -0
- package/dist/models/decision-checkpoint.d.ts +22 -0
- package/dist/models/decision-checkpoint.js +33 -0
- package/dist/models/decision-checkpoint.js.map +1 -0
- package/dist/models/gguf/classify.d.ts +27 -1
- package/dist/models/gguf/classify.js +78 -2
- package/dist/models/gguf/classify.js.map +1 -1
- package/dist/models/gguf/index.d.ts +1 -0
- package/dist/models/gguf/index.js +1 -0
- package/dist/models/gguf/index.js.map +1 -1
- package/dist/models/gguf/tensors.d.ts +57 -0
- package/dist/models/gguf/tensors.js +239 -0
- package/dist/models/gguf/tensors.js.map +1 -0
- package/dist/models/index.d.ts +2 -0
- package/dist/models/index.js +2 -0
- package/dist/models/index.js.map +1 -1
- package/dist/router/resolve.d.ts +7 -1
- package/dist/router/resolve.js +14 -1
- package/dist/router/resolve.js.map +1 -1
- package/dist/runtime/container/argv.d.ts +88 -0
- package/dist/runtime/container/argv.js +406 -0
- package/dist/runtime/container/argv.js.map +1 -0
- package/dist/runtime/container/docker-binary.d.ts +4 -0
- package/dist/runtime/container/docker-binary.js +34 -0
- package/dist/runtime/container/docker-binary.js.map +1 -0
- package/dist/runtime/container/env.d.ts +17 -0
- package/dist/runtime/container/env.js +52 -0
- package/dist/runtime/container/env.js.map +1 -0
- package/dist/runtime/container/exec.d.ts +27 -0
- package/dist/runtime/container/exec.js +79 -0
- package/dist/runtime/container/exec.js.map +1 -0
- package/dist/runtime/container/execution-journal.d.ts +55 -0
- package/dist/runtime/container/execution-journal.js +167 -0
- package/dist/runtime/container/execution-journal.js.map +1 -0
- package/dist/runtime/container/heartbeat.d.ts +47 -0
- package/dist/runtime/container/heartbeat.js +78 -0
- package/dist/runtime/container/heartbeat.js.map +1 -0
- package/dist/runtime/container/index.d.ts +30 -0
- package/dist/runtime/container/index.js +31 -0
- package/dist/runtime/container/index.js.map +1 -0
- package/dist/runtime/container/operations.d.ts +40 -0
- package/dist/runtime/container/operations.js +233 -0
- package/dist/runtime/container/operations.js.map +1 -0
- package/dist/runtime/container/pull.d.ts +48 -0
- package/dist/runtime/container/pull.js +280 -0
- package/dist/runtime/container/pull.js.map +1 -0
- package/dist/runtime/container/reconcile.d.ts +34 -0
- package/dist/runtime/container/reconcile.js +79 -0
- package/dist/runtime/container/reconcile.js.map +1 -0
- package/dist/runtime/container/types.d.ts +124 -0
- package/dist/runtime/container/types.js +2 -0
- package/dist/runtime/container/types.js.map +1 -0
- package/dist/runtime/container/watchdog.d.ts +114 -0
- package/dist/runtime/container/watchdog.js +418 -0
- package/dist/runtime/container/watchdog.js.map +1 -0
- package/dist/runtime/container/wiring.d.ts +78 -0
- package/dist/runtime/container/wiring.js +92 -0
- package/dist/runtime/container/wiring.js.map +1 -0
- package/dist/runtime/container/wsl-exec.d.ts +13 -0
- package/dist/runtime/container/wsl-exec.js +18 -0
- package/dist/runtime/container/wsl-exec.js.map +1 -0
- package/dist/runtime/environment/cached-document.d.ts +128 -0
- package/dist/runtime/environment/cached-document.js +218 -0
- package/dist/runtime/environment/cached-document.js.map +1 -0
- package/dist/runtime/environment/canonical-json.d.ts +78 -0
- package/dist/runtime/environment/canonical-json.js +86 -0
- package/dist/runtime/environment/canonical-json.js.map +1 -0
- package/dist/runtime/environment/descriptor-provider.d.ts +121 -0
- package/dist/runtime/environment/descriptor-provider.js +159 -0
- package/dist/runtime/environment/descriptor-provider.js.map +1 -0
- package/dist/runtime/environment/descriptor.d.ts +35 -0
- package/dist/runtime/environment/descriptor.js +239 -0
- package/dist/runtime/environment/descriptor.js.map +1 -0
- package/dist/runtime/environment/diagnostics.d.ts +50 -0
- package/dist/runtime/environment/diagnostics.js +114 -0
- package/dist/runtime/environment/diagnostics.js.map +1 -0
- package/dist/runtime/environment/document-fields.d.ts +39 -0
- package/dist/runtime/environment/document-fields.js +96 -0
- package/dist/runtime/environment/document-fields.js.map +1 -0
- package/dist/runtime/environment/environment-manifest-provider.d.ts +65 -0
- package/dist/runtime/environment/environment-manifest-provider.js +89 -0
- package/dist/runtime/environment/environment-manifest-provider.js.map +1 -0
- package/dist/runtime/environment/environment-manifest.d.ts +35 -0
- package/dist/runtime/environment/environment-manifest.js +131 -0
- package/dist/runtime/environment/environment-manifest.js.map +1 -0
- package/dist/runtime/environment/guest-host.d.ts +43 -0
- package/dist/runtime/environment/guest-host.js +88 -0
- package/dist/runtime/environment/guest-host.js.map +1 -0
- package/dist/runtime/environment/host-exec.d.ts +38 -0
- package/dist/runtime/environment/host-exec.js +145 -0
- package/dist/runtime/environment/host-exec.js.map +1 -0
- package/dist/runtime/environment/host-step-failure.d.ts +33 -0
- package/dist/runtime/environment/host-step-failure.js +34 -0
- package/dist/runtime/environment/host-step-failure.js.map +1 -0
- package/dist/runtime/environment/index.d.ts +56 -0
- package/dist/runtime/environment/index.js +57 -0
- package/dist/runtime/environment/index.js.map +1 -0
- package/dist/runtime/environment/installations.d.ts +25 -0
- package/dist/runtime/environment/installations.js +70 -0
- package/dist/runtime/environment/installations.js.map +1 -0
- package/dist/runtime/environment/inventory.d.ts +42 -0
- package/dist/runtime/environment/inventory.js +57 -0
- package/dist/runtime/environment/inventory.js.map +1 -0
- package/dist/runtime/environment/linux-blockers.d.ts +130 -0
- package/dist/runtime/environment/linux-blockers.js +251 -0
- package/dist/runtime/environment/linux-blockers.js.map +1 -0
- package/dist/runtime/environment/linux-docker-facts.d.ts +200 -0
- package/dist/runtime/environment/linux-docker-facts.js +342 -0
- package/dist/runtime/environment/linux-docker-facts.js.map +1 -0
- package/dist/runtime/environment/linux-docker-network.d.ts +80 -0
- package/dist/runtime/environment/linux-docker-network.js +195 -0
- package/dist/runtime/environment/linux-docker-network.js.map +1 -0
- package/dist/runtime/environment/linux-host.d.ts +22 -0
- package/dist/runtime/environment/linux-host.js +98 -0
- package/dist/runtime/environment/linux-host.js.map +1 -0
- package/dist/runtime/environment/linux-plan.d.ts +126 -0
- package/dist/runtime/environment/linux-plan.js +392 -0
- package/dist/runtime/environment/linux-plan.js.map +1 -0
- package/dist/runtime/environment/linux-probe.d.ts +245 -0
- package/dist/runtime/environment/linux-probe.js +315 -0
- package/dist/runtime/environment/linux-probe.js.map +1 -0
- package/dist/runtime/environment/linux-provisioner.d.ts +102 -0
- package/dist/runtime/environment/linux-provisioner.js +734 -0
- package/dist/runtime/environment/linux-provisioner.js.map +1 -0
- package/dist/runtime/environment/provisioner-descriptors.d.ts +17 -0
- package/dist/runtime/environment/provisioner-descriptors.js +62 -0
- package/dist/runtime/environment/provisioner-descriptors.js.map +1 -0
- package/dist/runtime/environment/recovery.d.ts +74 -0
- package/dist/runtime/environment/recovery.js +121 -0
- package/dist/runtime/environment/recovery.js.map +1 -0
- package/dist/runtime/environment/rootfs-download.d.ts +2 -0
- package/dist/runtime/environment/rootfs-download.js +47 -0
- package/dist/runtime/environment/rootfs-download.js.map +1 -0
- package/dist/runtime/environment/service.d.ts +280 -0
- package/dist/runtime/environment/service.js +776 -0
- package/dist/runtime/environment/service.js.map +1 -0
- package/dist/runtime/environment/state.d.ts +163 -0
- package/dist/runtime/environment/state.js +479 -0
- package/dist/runtime/environment/state.js.map +1 -0
- package/dist/runtime/environment/store-models.d.ts +19 -0
- package/dist/runtime/environment/store-models.js +30 -0
- package/dist/runtime/environment/store-models.js.map +1 -0
- package/dist/runtime/environment/store.d.ts +210 -0
- package/dist/runtime/environment/store.js +438 -0
- package/dist/runtime/environment/store.js.map +1 -0
- package/dist/runtime/environment/windows-environment-record.d.ts +48 -0
- package/dist/runtime/environment/windows-environment-record.js +89 -0
- package/dist/runtime/environment/windows-environment-record.js.map +1 -0
- package/dist/runtime/environment/windows-forwarding.d.ts +27 -0
- package/dist/runtime/environment/windows-forwarding.js +84 -0
- package/dist/runtime/environment/windows-forwarding.js.map +1 -0
- package/dist/runtime/environment/windows-guest-setup.d.ts +48 -0
- package/dist/runtime/environment/windows-guest-setup.js +157 -0
- package/dist/runtime/environment/windows-guest-setup.js.map +1 -0
- package/dist/runtime/environment/windows-host.d.ts +46 -0
- package/dist/runtime/environment/windows-host.js +163 -0
- package/dist/runtime/environment/windows-host.js.map +1 -0
- package/dist/runtime/environment/windows-plan.d.ts +78 -0
- package/dist/runtime/environment/windows-plan.js +216 -0
- package/dist/runtime/environment/windows-plan.js.map +1 -0
- package/dist/runtime/environment/windows-probe.d.ts +135 -0
- package/dist/runtime/environment/windows-probe.js +261 -0
- package/dist/runtime/environment/windows-probe.js.map +1 -0
- package/dist/runtime/environment/windows-provisioner.d.ts +113 -0
- package/dist/runtime/environment/windows-provisioner.js +782 -0
- package/dist/runtime/environment/windows-provisioner.js.map +1 -0
- package/dist/runtime/environment/wiring.d.ts +117 -0
- package/dist/runtime/environment/wiring.js +354 -0
- package/dist/runtime/environment/wiring.js.map +1 -0
- package/dist/runtime/foundation-models/runtime.d.ts +8 -0
- package/dist/runtime/foundation-models/runtime.js +5 -2
- package/dist/runtime/foundation-models/runtime.js.map +1 -1
- package/dist/runtime/llamacpp/args.d.ts +10 -1
- package/dist/runtime/llamacpp/args.js +43 -7
- package/dist/runtime/llamacpp/args.js.map +1 -1
- package/dist/runtime/llamacpp/fit-margin.d.ts +46 -0
- package/dist/runtime/llamacpp/fit-margin.js +47 -0
- package/dist/runtime/llamacpp/fit-margin.js.map +1 -0
- package/dist/runtime/llamacpp/index.d.ts +1 -0
- package/dist/runtime/llamacpp/index.js +1 -0
- package/dist/runtime/llamacpp/index.js.map +1 -1
- package/dist/runtime/llamacpp/load-plan.d.ts +23 -1
- package/dist/runtime/llamacpp/load-plan.js +82 -1
- package/dist/runtime/llamacpp/load-plan.js.map +1 -1
- package/dist/runtime/llamacpp/policy.d.ts +16 -0
- package/dist/runtime/llamacpp/policy.js +20 -0
- package/dist/runtime/llamacpp/policy.js.map +1 -1
- package/dist/runtime/llamacpp/probe.d.ts +21 -1
- package/dist/runtime/llamacpp/probe.js +32 -6
- package/dist/runtime/llamacpp/probe.js.map +1 -1
- package/dist/runtime/llamacpp/runtime.d.ts +56 -1
- package/dist/runtime/llamacpp/runtime.js +142 -15
- package/dist/runtime/llamacpp/runtime.js.map +1 -1
- package/dist/runtime/managed-engines/index.d.ts +8 -0
- package/dist/runtime/managed-engines/index.js +9 -0
- package/dist/runtime/managed-engines/index.js.map +1 -0
- package/dist/runtime/managed-engines/runtime.d.ts +140 -0
- package/dist/runtime/managed-engines/runtime.js +402 -0
- package/dist/runtime/managed-engines/runtime.js.map +1 -0
- package/dist/runtime/managed-engines/spec.d.ts +70 -0
- package/dist/runtime/managed-engines/spec.js +39 -0
- package/dist/runtime/managed-engines/spec.js.map +1 -0
- package/dist/runtime/managed-models/check.d.ts +50 -0
- package/dist/runtime/managed-models/check.js +157 -0
- package/dist/runtime/managed-models/check.js.map +1 -0
- package/dist/runtime/managed-models/compatibility.d.ts +256 -0
- package/dist/runtime/managed-models/compatibility.js +594 -0
- package/dist/runtime/managed-models/compatibility.js.map +1 -0
- package/dist/runtime/managed-models/delete.d.ts +15 -0
- package/dist/runtime/managed-models/delete.js +60 -0
- package/dist/runtime/managed-models/delete.js.map +1 -0
- package/dist/runtime/managed-models/guest-files.d.ts +8 -0
- package/dist/runtime/managed-models/guest-files.js +49 -0
- package/dist/runtime/managed-models/guest-files.js.map +1 -0
- package/dist/runtime/managed-models/index.d.ts +17 -0
- package/dist/runtime/managed-models/index.js +18 -0
- package/dist/runtime/managed-models/index.js.map +1 -0
- package/dist/runtime/managed-models/location.d.ts +22 -0
- package/dist/runtime/managed-models/location.js +53 -0
- package/dist/runtime/managed-models/location.js.map +1 -0
- package/dist/runtime/managed-models/migrate.d.ts +30 -0
- package/dist/runtime/managed-models/migrate.js +119 -0
- package/dist/runtime/managed-models/migrate.js.map +1 -0
- package/dist/runtime/managed-models/model-dir.d.ts +36 -0
- package/dist/runtime/managed-models/model-dir.js +136 -0
- package/dist/runtime/managed-models/model-dir.js.map +1 -0
- package/dist/runtime/managed-models/quant-format.d.ts +71 -0
- package/dist/runtime/managed-models/quant-format.js +299 -0
- package/dist/runtime/managed-models/quant-format.js.map +1 -0
- package/dist/runtime/managed-models/registry.d.ts +26 -0
- package/dist/runtime/managed-models/registry.js +66 -0
- package/dist/runtime/managed-models/registry.js.map +1 -0
- package/dist/runtime/managed-text/adapter.d.ts +283 -0
- package/dist/runtime/managed-text/adapter.js +96 -0
- package/dist/runtime/managed-text/adapter.js.map +1 -0
- package/dist/runtime/managed-text/backend-target.d.ts +3 -0
- package/dist/runtime/managed-text/backend-target.js +52 -0
- package/dist/runtime/managed-text/backend-target.js.map +1 -0
- package/dist/runtime/managed-text/deployment.d.ts +14 -0
- package/dist/runtime/managed-text/deployment.js +32 -0
- package/dist/runtime/managed-text/deployment.js.map +1 -0
- package/dist/runtime/managed-text/engine-cache.d.ts +15 -0
- package/dist/runtime/managed-text/engine-cache.js +64 -0
- package/dist/runtime/managed-text/engine-cache.js.map +1 -0
- package/dist/runtime/managed-text/gateway.d.ts +158 -0
- package/dist/runtime/managed-text/gateway.js +535 -0
- package/dist/runtime/managed-text/gateway.js.map +1 -0
- package/dist/runtime/managed-text/generation-defaults.d.ts +23 -0
- package/dist/runtime/managed-text/generation-defaults.js +51 -0
- package/dist/runtime/managed-text/generation-defaults.js.map +1 -0
- package/dist/runtime/managed-text/index.d.ts +26 -0
- package/dist/runtime/managed-text/index.js +27 -0
- package/dist/runtime/managed-text/index.js.map +1 -0
- package/dist/runtime/managed-text/lifecycle.d.ts +284 -0
- package/dist/runtime/managed-text/lifecycle.js +896 -0
- package/dist/runtime/managed-text/lifecycle.js.map +1 -0
- package/dist/runtime/managed-text/load-policy.d.ts +47 -0
- package/dist/runtime/managed-text/load-policy.js +111 -0
- package/dist/runtime/managed-text/load-policy.js.map +1 -0
- package/dist/runtime/managed-text/mount-source.d.ts +14 -0
- package/dist/runtime/managed-text/mount-source.js +18 -0
- package/dist/runtime/managed-text/mount-source.js.map +1 -0
- package/dist/runtime/managed-text/readiness.d.ts +33 -0
- package/dist/runtime/managed-text/readiness.js +88 -0
- package/dist/runtime/managed-text/readiness.js.map +1 -0
- package/dist/runtime/managed-text/request-rules.d.ts +18 -0
- package/dist/runtime/managed-text/request-rules.js +62 -0
- package/dist/runtime/managed-text/request-rules.js.map +1 -0
- package/dist/runtime/managed-text/types.d.ts +88 -0
- package/dist/runtime/managed-text/types.js +12 -0
- package/dist/runtime/managed-text/types.js.map +1 -0
- package/dist/runtime/managed-text/wsl-deployment.d.ts +24 -0
- package/dist/runtime/managed-text/wsl-deployment.js +70 -0
- package/dist/runtime/managed-text/wsl-deployment.js.map +1 -0
- package/dist/runtime/mlx/errors.d.ts +6 -1
- package/dist/runtime/mlx/errors.js +7 -2
- package/dist/runtime/mlx/errors.js.map +1 -1
- package/dist/runtime/mlx/runtime.d.ts +36 -1
- package/dist/runtime/mlx/runtime.js +85 -4
- package/dist/runtime/mlx/runtime.js.map +1 -1
- package/dist/runtime/shared/backend-output.d.ts +23 -0
- package/dist/runtime/shared/backend-output.js +32 -0
- package/dist/runtime/shared/backend-output.js.map +1 -0
- package/dist/runtime/shared/bounded-output.d.ts +24 -0
- package/dist/runtime/shared/bounded-output.js +63 -0
- package/dist/runtime/shared/bounded-output.js.map +1 -0
- package/dist/runtime/shared/external-sessions.js +9 -1
- package/dist/runtime/shared/external-sessions.js.map +1 -1
- package/dist/runtime/shared/gpu-occupancy.d.ts +60 -0
- package/dist/runtime/shared/gpu-occupancy.js +10 -0
- package/dist/runtime/shared/gpu-occupancy.js.map +1 -0
- package/dist/runtime/shared/index.d.ts +4 -0
- package/dist/runtime/shared/index.js +4 -0
- package/dist/runtime/shared/index.js.map +1 -1
- package/dist/runtime/shared/local-runtime.d.ts +35 -0
- package/dist/runtime/shared/ports.d.ts +16 -0
- package/dist/runtime/shared/ports.js +42 -1
- package/dist/runtime/shared/ports.js.map +1 -1
- package/dist/runtime/shared/process.d.ts +11 -0
- package/dist/runtime/shared/process.js +20 -0
- package/dist/runtime/shared/process.js.map +1 -1
- package/dist/runtime/shared/redact-args.d.ts +7 -0
- package/dist/runtime/shared/redact-args.js +26 -0
- package/dist/runtime/shared/redact-args.js.map +1 -0
- package/dist/runtime/shared/sidecar.d.ts +15 -0
- package/dist/runtime/shared/sidecar.js +43 -8
- package/dist/runtime/shared/sidecar.js.map +1 -1
- package/dist/runtime/tensorrt-llm/adapter.d.ts +247 -0
- package/dist/runtime/tensorrt-llm/adapter.js +1028 -0
- package/dist/runtime/tensorrt-llm/adapter.js.map +1 -0
- package/dist/runtime/tensorrt-llm/check.d.ts +12 -0
- package/dist/runtime/tensorrt-llm/check.js +22 -0
- package/dist/runtime/tensorrt-llm/check.js.map +1 -0
- package/dist/runtime/tensorrt-llm/compatibility.d.ts +124 -0
- package/dist/runtime/tensorrt-llm/compatibility.js +213 -0
- package/dist/runtime/tensorrt-llm/compatibility.js.map +1 -0
- package/dist/runtime/tensorrt-llm/cuda-graphs.d.ts +32 -0
- package/dist/runtime/tensorrt-llm/cuda-graphs.js +33 -0
- package/dist/runtime/tensorrt-llm/cuda-graphs.js.map +1 -0
- package/dist/runtime/tensorrt-llm/delete.d.ts +15 -0
- package/dist/runtime/tensorrt-llm/delete.js +60 -0
- package/dist/runtime/tensorrt-llm/delete.js.map +1 -0
- package/dist/runtime/tensorrt-llm/engine.d.ts +3 -0
- package/dist/runtime/tensorrt-llm/engine.js +30 -0
- package/dist/runtime/tensorrt-llm/engine.js.map +1 -0
- package/dist/runtime/tensorrt-llm/guest-files.d.ts +8 -0
- package/dist/runtime/tensorrt-llm/guest-files.js +49 -0
- package/dist/runtime/tensorrt-llm/guest-files.js.map +1 -0
- package/dist/runtime/tensorrt-llm/host-facts.d.ts +55 -0
- package/dist/runtime/tensorrt-llm/host-facts.js +87 -0
- package/dist/runtime/tensorrt-llm/host-facts.js.map +1 -0
- package/dist/runtime/tensorrt-llm/index.d.ts +30 -0
- package/dist/runtime/tensorrt-llm/index.js +31 -0
- package/dist/runtime/tensorrt-llm/index.js.map +1 -0
- package/dist/runtime/tensorrt-llm/installation.d.ts +23 -0
- package/dist/runtime/tensorrt-llm/installation.js +53 -0
- package/dist/runtime/tensorrt-llm/installation.js.map +1 -0
- package/dist/runtime/tensorrt-llm/kv-cache.d.ts +41 -0
- package/dist/runtime/tensorrt-llm/kv-cache.js +48 -0
- package/dist/runtime/tensorrt-llm/kv-cache.js.map +1 -0
- package/dist/runtime/tensorrt-llm/location.d.ts +20 -0
- package/dist/runtime/tensorrt-llm/location.js +48 -0
- package/dist/runtime/tensorrt-llm/location.js.map +1 -0
- package/dist/runtime/tensorrt-llm/model-dir.d.ts +38 -0
- package/dist/runtime/tensorrt-llm/model-dir.js +137 -0
- package/dist/runtime/tensorrt-llm/model-dir.js.map +1 -0
- package/dist/runtime/tensorrt-llm/prelaunch.d.ts +20 -0
- package/dist/runtime/tensorrt-llm/prelaunch.js +168 -0
- package/dist/runtime/tensorrt-llm/prelaunch.js.map +1 -0
- package/dist/runtime/tensorrt-llm/quant-format.d.ts +65 -0
- package/dist/runtime/tensorrt-llm/quant-format.js +189 -0
- package/dist/runtime/tensorrt-llm/quant-format.js.map +1 -0
- package/dist/runtime/tensorrt-llm/registry.d.ts +26 -0
- package/dist/runtime/tensorrt-llm/registry.js +68 -0
- package/dist/runtime/tensorrt-llm/registry.js.map +1 -0
- package/dist/runtime/tensorrt-llm/route-policy.d.ts +19 -0
- package/dist/runtime/tensorrt-llm/route-policy.js +17 -0
- package/dist/runtime/tensorrt-llm/route-policy.js.map +1 -0
- package/dist/runtime/tensorrt-llm/runtime.d.ts +130 -0
- package/dist/runtime/tensorrt-llm/runtime.js +395 -0
- package/dist/runtime/tensorrt-llm/runtime.js.map +1 -0
- package/dist/runtime/tensorrt-llm/settings.d.ts +7 -0
- package/dist/runtime/tensorrt-llm/settings.js +48 -0
- package/dist/runtime/tensorrt-llm/settings.js.map +1 -0
- package/dist/runtime/vllm/adapter.d.ts +46 -0
- package/dist/runtime/vllm/adapter.js +425 -0
- package/dist/runtime/vllm/adapter.js.map +1 -0
- package/dist/runtime/vllm/engine.d.ts +16 -0
- package/dist/runtime/vllm/engine.js +37 -0
- package/dist/runtime/vllm/engine.js.map +1 -0
- package/dist/runtime/vllm/index.d.ts +11 -0
- package/dist/runtime/vllm/index.js +12 -0
- package/dist/runtime/vllm/index.js.map +1 -0
- package/dist/runtime/vllm/memory.d.ts +69 -0
- package/dist/runtime/vllm/memory.js +123 -0
- package/dist/runtime/vllm/memory.js.map +1 -0
- package/dist/runtime/vllm/settings.d.ts +52 -0
- package/dist/runtime/vllm/settings.js +151 -0
- package/dist/runtime/vllm/settings.js.map +1 -0
- package/dist/runtime/wsl/guest-paths.d.ts +33 -0
- package/dist/runtime/wsl/guest-paths.js +83 -0
- package/dist/runtime/wsl/guest-paths.js.map +1 -0
- package/dist/runtime/wsl/guest-scope.d.ts +14 -0
- package/dist/runtime/wsl/guest-scope.js +98 -0
- package/dist/runtime/wsl/guest-scope.js.map +1 -0
- package/dist/runtime/wsl/index.d.ts +12 -0
- package/dist/runtime/wsl/index.js +13 -0
- package/dist/runtime/wsl/index.js.map +1 -0
- package/dist/runtime/wsl/keeper.d.ts +29 -0
- package/dist/runtime/wsl/keeper.js +48 -0
- package/dist/runtime/wsl/keeper.js.map +1 -0
- package/dist/runtime/wsl/transport.d.ts +106 -0
- package/dist/runtime/wsl/transport.js +181 -0
- package/dist/runtime/wsl/transport.js.map +1 -0
- package/dist/server/control/index.d.ts +1 -1
- package/dist/server/control/router.js +16 -1
- package/dist/server/control/router.js.map +1 -1
- package/dist/server/control/routes/backends.js +26 -0
- package/dist/server/control/routes/backends.js.map +1 -1
- package/dist/server/control/routes/decision.d.ts +10 -0
- package/dist/server/control/routes/decision.js +43 -0
- package/dist/server/control/routes/decision.js.map +1 -0
- package/dist/server/control/routes/diffusion-video.js +4 -0
- package/dist/server/control/routes/diffusion-video.js.map +1 -1
- package/dist/server/control/routes/diffusion.d.ts +1 -1
- package/dist/server/control/routes/diffusion.js +4 -9
- package/dist/server/control/routes/diffusion.js.map +1 -1
- package/dist/server/control/routes/embedding.d.ts +9 -0
- package/dist/server/control/routes/embedding.js +43 -0
- package/dist/server/control/routes/embedding.js.map +1 -0
- package/dist/server/control/routes/engine-builds.d.ts +14 -0
- package/dist/server/control/routes/engine-builds.js +82 -0
- package/dist/server/control/routes/engine-builds.js.map +1 -0
- package/dist/server/control/routes/engines.d.ts +17 -0
- package/dist/server/control/routes/engines.js +152 -0
- package/dist/server/control/routes/engines.js.map +1 -0
- package/dist/server/control/routes/environments.d.ts +16 -0
- package/dist/server/control/routes/environments.js +207 -0
- package/dist/server/control/routes/environments.js.map +1 -0
- package/dist/server/control/routes/hardware.d.ts +1 -1
- package/dist/server/control/routes/hardware.js +11 -6
- package/dist/server/control/routes/hardware.js.map +1 -1
- package/dist/server/control/routes/model-setups.d.ts +11 -0
- package/dist/server/control/routes/model-setups.js +55 -0
- package/dist/server/control/routes/model-setups.js.map +1 -0
- package/dist/server/control/routes/models.d.ts +3 -2
- package/dist/server/control/routes/models.js +45 -2
- package/dist/server/control/routes/models.js.map +1 -1
- package/dist/server/control/types.d.ts +172 -10
- package/dist/server/control/types.js.map +1 -1
- package/dist/server/http.d.ts +9 -1
- package/dist/server/http.js +51 -1
- package/dist/server/http.js.map +1 -1
- package/dist/server/public/decision-preview.d.ts +17 -0
- package/dist/server/public/decision-preview.js +119 -0
- package/dist/server/public/decision-preview.js.map +1 -0
- package/dist/server/public/decision.d.ts +32 -0
- package/dist/server/public/decision.js +135 -0
- package/dist/server/public/decision.js.map +1 -0
- package/dist/server/public/embedding.d.ts +31 -0
- package/dist/server/public/embedding.js +126 -0
- package/dist/server/public/embedding.js.map +1 -0
- package/dist/server/public/errors.d.ts +2 -1
- package/dist/server/public/errors.js +5 -2
- package/dist/server/public/errors.js.map +1 -1
- package/dist/server/public/forward.js +75 -3
- package/dist/server/public/forward.js.map +1 -1
- package/dist/server/public/gates.d.ts +18 -0
- package/dist/server/public/gates.js +41 -14
- package/dist/server/public/gates.js.map +1 -1
- package/dist/server/public/index.d.ts +6 -2
- package/dist/server/public/index.js +23 -7
- package/dist/server/public/index.js.map +1 -1
- package/dist/server/public/listing.d.ts +13 -2
- package/dist/server/public/listing.js +33 -9
- package/dist/server/public/listing.js.map +1 -1
- package/dist/server/public/policy.d.ts +9 -0
- package/dist/server/public/policy.js +68 -0
- package/dist/server/public/policy.js.map +1 -0
- package/dist/server/public/responses.js +10 -0
- package/dist/server/public/responses.js.map +1 -1
- package/dist/server/public/trace.d.ts +2 -2
- package/dist/server/public/trace.js +6 -2
- package/dist/server/public/trace.js.map +1 -1
- package/dist/server/public/types.d.ts +91 -1
- package/dist/settings/decision.d.ts +15 -0
- package/dist/settings/decision.js +88 -0
- package/dist/settings/decision.js.map +1 -0
- package/dist/settings/embedding.d.ts +12 -0
- package/dist/settings/embedding.js +82 -0
- package/dist/settings/embedding.js.map +1 -0
- package/dist/settings/index.d.ts +2 -0
- package/dist/settings/index.js +2 -0
- package/dist/settings/index.js.map +1 -1
- package/dist/settings/schema/atomic-prism.json +433 -0
- package/dist/settings/schema/tensorrt-llm.json +253 -0
- package/dist/settings/schema/vllm.json +273 -0
- package/dist/settings/schema.js +58 -0
- package/dist/settings/schema.js.map +1 -1
- package/dist/settings/store.d.ts +13 -2
- package/dist/settings/store.js +44 -0
- package/dist/settings/store.js.map +1 -1
- package/dist/util/index.d.ts +1 -0
- package/dist/util/index.js +1 -0
- package/dist/util/index.js.map +1 -1
- package/dist/util/result.d.ts +19 -0
- package/dist/util/result.js +15 -0
- package/dist/util/result.js.map +1 -0
- package/dist/version.d.ts +1 -1
- package/dist/version.js +1 -1
- package/dist/version.js.map +1 -1
- package/package.json +2 -2
- package/src/settings/schema/atomic-prism.json +433 -0
- package/src/settings/schema/tensorrt-llm.json +253 -0
- package/src/settings/schema/vllm.json +273 -0
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"runtime.js","sourceRoot":"","sources":["../../../src/runtime/tensorrt-llm/runtime.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAiCG;AACH,OAAO,EAAE,eAAe,EAAE,MAAM,0BAA0B,CAAA;AAa1D,OAAO,EAAE,oBAAoB,EAAE,MAAM,oBAAoB,CAAA;AAUzD,OAAO,EAAE,kBAAkB,EAAE,MAAM,cAAc,CAAA;AACjD,OAAO,EAAE,gBAAgB,EAAE,eAAe,EAAE,MAAM,oBAAoB,CAAA;AAKtE,OAAO,EAAE,gCAAgC,EAAE,MAAM,gBAAgB,CAAA;AACjE,OAAO,EAAE,sBAAsB,EAAE,MAAM,mBAAmB,CAAA;AAC1D,OAAO,EAAE,mBAAmB,EAAE,MAAM,eAAe,CAAA;AAiDnD,MAAM,IAAI,GAA4B;IACpC,KAAK,EAAE,KAAK;IACZ,SAAS,EAAE,KAAK;IAChB,iBAAiB,EAAE,KAAK;IACxB,MAAM,EAAE,KAAK;IACb,UAAU,EAAE,KAAK;IACjB,SAAS,EAAE,KAAK;CACjB,CAAA;AAED,uGAAuG;AACvG,SAAS,QAAQ,CAAC,KAAwB,EAAE,KAAuB;IACjE,IAAI,KAAK,CAAC,YAAY,KAAK,IAAI;QAAE,OAAO,IAAI,CAAA;IAC5C,OAAO,KAAK,CAAC,UAAU,CAAC,cAAc,CAAC,KAAK,CAAC,YAAY,CAAC,IAAI,IAAI,CAAA;AACpE,CAAC;AAED;;;;;;GAMG;AACH,SAAS,kBAAkB,CACzB,KAAwB,EACxB,QAA4B;IAE5B,MAAM,YAAY,GAAG,QAAQ,CAAC,aAAa,CAAC,CAAC,CAAC,CAAA;IAC9C,OAAO,YAAY,KAAK,SAAS,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,UAAU,CAAC,cAAc,CAAC,YAAY,CAAC,IAAI,IAAI,CAAC,CAAA;AACpG,CAAC;AAED,MAAM,OAAO,kBAAkB;IAmBA;IAlBrB,OAAO,GAAgC,IAAI,CAAA;IAC3C,MAAM,GAAG,KAAK,CAAA;IACtB,kGAAkG;IAC1F,SAAS,GAAG,CAAC,CAAA;IACrB,sGAAsG;IACrF,UAAU,GAAG,IAAI,GAAG,EAAkB,CAAA;IACvD,sFAAsF;IACrE,OAAO,GAAG,IAAI,GAAG,EAAU,CAAA;IAC5C,wGAAwG;IACvF,mBAAmB,GAAG,IAAI,GAAG,EAO3C,CAAA;IAEH,YAA6B,IAA4B;QAA5B,SAAI,GAAJ,IAAI,CAAwB;IAAG,CAAC;IAE7D,IAAI;QACF,OAAO,IAAI,CAAC,OAAO,EAAE,IAAI,EAAE,IAAI,EAAE,CAAA;IACnC,CAAC;IAED,WAAW,CAAC,OAAe;QACzB,OAAO,IAAI,CAAC,OAAO,EAAE,WAAW,CAAC,OAAO,CAAC,CAAA;IAC3C,CAAC;IAED,eAAe;QACb,OAAO,IAAI,CAAC,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAA;IACvD,CAAC;IAED,SAAS,CAAC,OAAe;QACvB,OAAO,IAAI,CAAC,OAAO,EAAE,SAAS,CAAC,OAAO,CAAC,IAAI,KAAK,CAAA;IAClD,CAAC;IAED,KAAK,CAAC,IAAI,CAAC,OAAe,EAAE,OAAyB,EAAE;QACrD,MAAM,EAAE,MAAM,EAAE,GAAG,IAAI,CAAA;QACvB,IAAI,CAAC,UAAU,EAAE,CAAA;QACjB,IAAI,CAAC,gBAAgB,CAAC,OAAO,CAAC,CAAA;QAC9B,oBAAoB,CAAC,MAAM,CAAC,CAAA;QAC5B,IAAI,IAAI,CAAC,WAAW,EAAE,CAAC;YACrB,MAAM,IAAI,eAAe,CAAC,kBAAkB,EAAE,8CAA8C,EAAE,OAAO,CAAC,CAAA;QACxG,CAAC;QACD,mGAAmG;QACnG,MAAM,QAAQ,GAAG,mBAAmB,CAAC,IAAI,CAAC,IAAI,CAAC,QAAQ,EAAE,EAAE,IAAI,CAAC,SAAS,CAAC,CAAA;QAC1E,MAAM,SAAS,GAAG,MAAM,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,CAAC,KAAK,CAAC,CAAC,KAAc,EAAE,EAAE;YACrE,MAAM,IAAI,eAAe,CACvB,6BAA6B,EAC7B,wFAAwF,EACxF,KAAK,YAAY,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,CACvD,CAAA;QACH,CAAC,CAAC,CAAA;QACF,IAAI,SAAS,KAAK,IAAI,EAAE,CAAC;YACvB,MAAM,IAAI,eAAe,CACvB,6BAA6B,EAC7B,6EAA6E,CAC9E,CAAA;QACH,CAAC;QACD,8FAA8F;QAC9F,oEAAoE;QACpE,IAAI,CAAC,OAAO,GAAG,SAAS,CAAA;QACxB,MAAM,KAAK,GAAG,MAAM,IAAI,CAAC,IAAI,CAAC,iBAAiB,EAAE,CAAA;QACjD,oBAAoB,CAAC,MAAM,CAAC,CAAA;QAC5B,MAAM,KAAK,GAAG,MAAM,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,CAAA;QAC5C,MAAM,KAAK,GAAG,MAAM,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,CAAA;QACzC,oBAAoB,CAAC,MAAM,CAAC,CAAA;QAC5B,IAAI,CAAC,UAAU,EAAE,CAAA;QAEjB,MAAM,GAAG,GAAG,eAAe,CAAC,KAAK,CAAC,IAAI,EAAE,KAAK,CAAC,MAAM,EAAE,QAAQ,CAAC,MAAM,IAAI,SAAS,CAAC,CAAA;QACnF,IAAI,GAAG,KAAK,IAAI,EAAE,CAAC;YACjB,MAAM,IAAI,eAAe,CACvB,8BAA8B,EAC9B,+EAA+E,CAChF,CAAA;QACH,CAAC;QACD,MAAM,WAAW,GACf,QAAQ,CAAC,MAAM,KAAK,IAAI,IAAI,GAAG,CAAC,MAAM,KAAK,QAAQ,CAAC,MAAM;YACxD,CAAC,CAAC,EAAE,gBAAgB,EAAE,QAAQ,CAAC,MAAM,EAAE,MAAM,EAAE,GAAG,CAAC,MAAM,EAAE;YAC3D,CAAC,CAAC,SAAS,CAAA;QACf,MAAM,MAAM,GAAuB;YACjC,aAAa,EAAE,QAAQ,CAAC,cAAc;YACtC,4BAA4B,EAAE,QAAQ,CAAC,iCAAiC;SACzE,CAAA;QAED,gGAAgG;QAChG,6FAA6F;QAC7F,gGAAgG;QAChG,iGAAiG;QACjG,iGAAiG;QACjG,0BAA0B;QAC1B,MAAM,QAAQ,GAAG,MAAM,gCAAgC,CACrD,KAAK,EACL,KAAK,CAAC,UAAU,EAChB,KAAK,CAAC,IAAI,EACV,KAAK,CAAC,MAAM,EACZ,EAAE,KAAK,EAAE,GAAG,CAAC,MAAM,EAAE,MAAM,EAAE,CAC9B,CAAA;QACD,oBAAoB,CAAC,MAAM,CAAC,CAAA;QAC5B,IAAI,CAAC,UAAU,EAAE,CAAA;QAEjB,0FAA0F;QAC1F,MAAM,MAAM,GAAG,kBAAkB,CAAC,KAAK,EAAE,QAAQ,CAAC,CAAA;QAElD,MAAM,EAAE,UAAU,EAAE,GAAG,KAAK,CAAA;QAC5B,iGAAiG;QACjG,6FAA6F;QAC7F,IAAI,CAAC,gBAAgB,CAAC,OAAO,CAAC,CAAA;QAC9B,MAAM,OAAO,GAAG,MAAM,SAAS,CAAC,IAAI,CAAC;YACnC,OAAO;YACP,SAAS,EAAE,KAAK,CAAC,GAAG;YACpB,WAAW,EAAE,KAAK,CAAC,WAAW;YAC9B,YAAY,EAAE;gBACZ,aAAa,EAAE,UAAU,CAAC,aAAa;gBACvC,SAAS,EAAE,UAAU,CAAC,SAAS;gBAC/B,UAAU,EAAE,UAAU,CAAC,UAAU;gBACjC,wBAAwB,EAAE,UAAU,CAAC,wBAAwB;gBAC7D,KAAK,EAAE,KAAK,CAAC,KAAK;aACnB;YACD,MAAM;YACN,OAAO,EAAE,GAAG,CAAC,MAAM;YACnB,2FAA2F;YAC3F,yDAAyD;YACzD,aAAa,EAAE,GAAG,CAAC,gBAAgB,KAAK,IAAI;YAC5C,iBAAiB,EAAE,GAAG,CAAC,gBAAgB;YACvC,oBAAoB,EAAE,GAAG,CAAC,kBAAkB;YAC5C,OAAO,EAAE,KAAK,CAAC,OAAO;YACtB,QAAQ;YACR,6FAA6F;YAC7F,gFAAgF;YAChF,YAAY,EAAE,CAAC,WAAW,EAAE,UAAU,EAAE,EAAE,CACxC,IAAI,CAAC,UAAU,CAAC,SAAS,EAAE,EAAE,OAAO,EAAE,UAAU,EAAE,KAAK,EAAE,GAAG,CAAC,MAAM,EAAE,EAAE,WAAW,CAAC;YACrF,qFAAqF;YACrF,4FAA4F;YAC5F,2FAA2F;YAC3F,YAAY,EAAE,GAAG,EAAE,CAAC,IAAI,CAAC,uBAAuB,CAAC,GAAG,CAAC,MAAM,EAAE,UAAU,EAAE,QAAQ,EAAE,MAAM,CAAC;YAC1F,GAAG,CAAC,IAAI,CAAC,WAAW,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,IAAI,CAAC,WAAW,GAAG,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YACjF,GAAG,CAAC,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YAC3C,GAAG,CAAC,WAAW,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,cAAc,EAAE,WAAW,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;SACtE,CAAC,CAAA;QACF,IAAI,CAAC,mBAAmB,CAAC,GAAG,CAAC,OAAO,EAAE;YACpC,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,EAAE;YACpC,YAAY,EAAE,kBAAkB,CAAC,YAAY,CAAC,EAAE,QAAQ,EAAE,MAAM,EAAE,CAAC;YACnE,MAAM,EAAE,EAAE,aAAa,EAAE,QAAQ,CAAC,cAAc,EAAE,eAAe,EAAE,QAAQ,CAAC,iBAAiB,EAAE;SAChG,CAAC,CAAA;QACF,OAAO,OAAO,CAAA;IAChB,CAAC;IAED;;;OAGG;IACH,KAAK,CAAC,MAAM,CAAC,OAAe;QAC1B,wFAAwF;QACxF,4FAA4F;QAC5F,MAAM,SAAS,GAAG,IAAI,CAAC,cAAc,EAAE,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAA;QACzD,MAAM,IAAI,CAAC,OAAO,EAAE,MAAM,CAAC,OAAO,CAAC,CAAA;QACnC,IAAI,CAAC,mBAAmB,CAAC,MAAM,CAAC,OAAO,CAAC,CAAA;QACxC,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,UAAU,EAAE,SAAS,EAAE,CAAA;IACjD,CAAC;IAED;;;;OAIG;IACH,cAAc;QACZ,OAAO,CAAC,GAAG,IAAI,GAAG,CAAC,CAAC,IAAI,CAAC,OAAO,EAAE,YAAY,EAAE,IAAI,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAA;IAClF,CAAC;IAED;;;;OAIG;IACH,YAAY;QACV,IAAI,CAAC,SAAS,IAAI,CAAC,CAAA;QACnB,IAAI,QAAQ,GAAG,KAAK,CAAA;QACpB,OAAO,GAAG,EAAE;YACV,IAAI,QAAQ;gBAAE,OAAM;YACpB,QAAQ,GAAG,IAAI,CAAA;YACf,IAAI,CAAC,SAAS,IAAI,CAAC,CAAA;QACrB,CAAC,CAAA;IACH,CAAC;IAED;;;;OAIG;IACH,YAAY,CAAC,OAAe;QAC1B,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,OAAO,EAAE,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAA;QACrE,IAAI,QAAQ,GAAG,KAAK,CAAA;QACpB,OAAO,GAAG,EAAE;YACV,IAAI,QAAQ;gBAAE,OAAM;YACpB,QAAQ,GAAG,IAAI,CAAA;YACf,MAAM,IAAI,GAAG,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAA;YACpD,IAAI,IAAI,GAAG,CAAC;gBAAE,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,OAAO,EAAE,IAAI,CAAC,CAAA;;gBAC3C,IAAI,CAAC,UAAU,CAAC,MAAM,CAAC,OAAO,CAAC,CAAA;QACtC,CAAC,CAAA;IACH,CAAC;IAED,gGAAgG;IAChG,eAAe,CAAC,QAAgB;QAC9B,OAAO,OAAO,CAAC,OAAO,CAAC,EAAE,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,aAAa,EAAE,CAAC,CAAA;IAC9D,CAAC;IAED,eAAe,CAAC,OAAe;QAC7B,OAAO,OAAO,CAAC,MAAM,CACnB,IAAI,eAAe,CACjB,kBAAkB,EAClB,uFAAuF,EACvF,OAAO,CACR,CACF,CAAA;IACH,CAAC;IAED,KAAK,CAAC,QAAQ;QACZ,IAAI,CAAC,MAAM,GAAG,IAAI,CAAA;QAClB,MAAM,IAAI,CAAC,OAAO,EAAE,QAAQ,EAAE,CAAA;QAC9B,IAAI,CAAC,mBAAmB,CAAC,KAAK,EAAE,CAAA;IAClC,CAAC;IAED,WAAW,CAAC,OAAe;QACzB,MAAM,OAAO,GAAG,IAAI,CAAC,WAAW,CAAC,OAAO,CAAC,CAAA;QACzC,IAAI,OAAO,KAAK,SAAS;YAAE,OAAO,SAAS,CAAA;QAC3C,MAAM,KAAK,GAAG,IAAI,CAAC,mBAAmB,CAAC,GAAG,CAAC,OAAO,CAAC,CAAA;QACnD,wFAAwF;QACxF,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,CAAC,UAAU,KAAK,OAAO,CAAC,UAAU;YAAE,OAAO,sBAAsB,CAAC,IAAI,CAAC,CAAA;QACvG,OAAO,sBAAsB,CAAC,KAAK,CAAC,YAAY,EAAE,KAAK,CAAC,MAAM,CAAC,CAAA;IACjE,CAAC;IAED,wGAAwG;IACxG,KAAK,CAAC,YAAY,CAAC,OAAe;QAChC,MAAM,IAAI,GAAiC;YACzC,OAAO;YACP,YAAY,EAAE,KAAK;YACnB,WAAW,EAAE,KAAK;YAClB,UAAU,EAAE,KAAK;YACjB,KAAK,EAAE,KAAK;YACZ,QAAQ,EAAE,KAAK;YACf,MAAM,EAAE,KAAK;YACb,YAAY,EAAE,EAAE;YAChB,GAAG,IAAI;YACP,YAAY,EAAE,IAAI;SACnB,CAAA;QACD,IAAI,CAAC;YACH,MAAM,CAAC,KAAK,EAAE,KAAK,CAAC,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,iBAAiB,EAAE,EAAE,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,CAAA;YACnG,MAAM,QAAQ,GAAG,mBAAmB,CAAC,EAAE,CAAC,CAAA;YACxC,OAAO;gBACL,GAAG,IAAI;gBACP,GAAG,kBAAkB,CAAC,YAAY,CAAC,EAAE,QAAQ,EAAE,MAAM,EAAE,QAAQ,CAAC,KAAK,EAAE,KAAK,CAAC,EAAE,CAAC;gBAChF,YAAY,EAAE,KAAK,CAAC,YAAY;aACjC,CAAA;QACH,CAAC;QAAC,MAAM,CAAC;YACP,OAAO,IAAI,CAAA;QACb,CAAC;IACH,CAAC;IAED,KAAK,CAAC,IAAI,CAAC,OAAe;QACxB,MAAM,SAAS,GAAG,IAAI,CAAC,OAAO,CAAA;QAC9B,MAAM,OAAO,GAAG,SAAS,EAAE,WAAW,CAAC,OAAO,CAAC,CAAA;QAC/C,IAAI,SAAS,IAAI,OAAO,EAAE,CAAC;YACzB,OAAO;gBACL,QAAQ,EAAE,OAAO;gBACjB,MAAM,EAAE,SAAS;gBACjB,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,EAAE;gBACpC,QAAQ,EAAE,CAAC,MAAM,SAAS,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,IAAI,EAAE;aAChD,CAAA;QACH,CAAC;QACD,MAAM,OAAO,GAAG,SAAS,EAAE,WAAW,CAAC,OAAO,CAAC,CAAA;QAC/C,IAAI,OAAO,KAAK,SAAS;YAAE,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAA;QACnF,OAAO;YACL,QAAQ,EAAE,OAAO;YACjB,MAAM,EAAE,cAAc;YACtB,UAAU,EAAE,OAAO,CAAC,UAAU;YAC9B,QAAQ,EAAE,OAAO,CAAC,QAAQ;YAC1B,KAAK,EAAE,OAAO,CAAC,KAAK;YACpB,EAAE,EAAE,OAAO,CAAC,EAAE;SACf,CAAA;IACH,CAAC;IAEO,UAAU;QAChB,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;YAChB,MAAM,IAAI,eAAe,CAAC,kBAAkB,EAAE,6CAA6C,CAAC,CAAA;QAC9F,CAAC;IACH,CAAC;IAEO,gBAAgB,CAAC,OAAe;QACtC,IAAI,IAAI,CAAC,SAAS,GAAG,CAAC,EAAE,CAAC;YACvB,MAAM,IAAI,eAAe,CACvB,4BAA4B,EAC5B,yFAAyF,EACzF,OAAO,CACR,CAAA;QACH,CAAC;QACD,IAAI,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC;YACjC,MAAM,IAAI,eAAe,CAAC,4BAA4B,EAAE,6BAA6B,EAAE,OAAO,CAAC,CAAA;QACjG,CAAC;IACH,CAAC;IAED;;;OAGG;IACH,YAAY;QACV,OAAO,CAAC,IAAI,CAAC,OAAO,EAAE,YAAY,EAAE,IAAI,EAAE,CAAC;aACxC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,KAAK,SAAS,IAAI,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC;aACtE,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;YACX,QAAQ,EAAE,CAAC,CAAC,QAAQ;YACpB,KAAK,EAAE,CAAC,CAAC,CAAC,QAAQ,CAAC;YACnB,SAAS,EAAE,KAAK;YAChB,KAAK,EAAE,CAAC,CAAC,KAAK;YACd,GAAG,CAAC,CAAC,CAAC,KAAK,KAAK,kBAAkB,IAAI,CAAC,CAAC,YAAY,KAAK,IAAI;gBAC3D,CAAC,CAAC;oBACE,MAAM,EACJ,oFAAoF;wBACpF,aAAa,CAAC,CAAC,YAAY,kBAAkB,CAAC,CAAC,YAAY,IAAI;iBAClE;gBACH,CAAC,CAAC,EAAE,CAAC;SACR,CAAC,CAAC,CAAA;IACP,CAAC;IAED;;;;;;OAMG;IACK,KAAK,CAAC,UAAU,CACtB,SAA+B,EAC/B,IAA4D,EAC5D,MAAmB;QAEnB,0FAA0F;QAC1F,MAAM,IAAI,GAAG,IAAI,GAAG,CAAC,SAAS,CAAC,YAAY,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC,CAAA;QACvE,KAAK,MAAM,UAAU,IAAI,IAAI,CAAC,OAAO;YAAE,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,UAAU,CAAC;gBAAE,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,UAAU,CAAC,CAAA;QACjG,MAAM,IAAI,GAAG,GAAG,EAAE;YAChB,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,IAAI,CAAC,UAAU,CAAC,CAAA;QACnC,CAAC,CAAA;QACD,IAAI,IAAI,CAAC,IAAI,CAAC,QAAQ,EAAE,CAAC;YACvB,MAAM,KAAK,GAAG;gBACZ,QAAQ,EAAE,IAAI,CAAC,OAAO;gBACtB,KAAK,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC;gBACnB,SAAS,EAAE,KAAK;gBAChB,qBAAqB,EAAE,IAAI;aAC5B,CAAA;YACD,MAAM,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,KAAK,EAAE,MAAM,EAAE,IAAI,CAAC,CAAA;QAC/C,CAAC;aAAM,CAAC;YACN,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC,SAAS,CAAC,YAAY,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAA;YACvE,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,CAAA;YAC3B,KAAK,MAAM,KAAK,IAAI,MAAM;gBAAE,MAAM,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,CAAA;QACtD,CAAC;QACD,IAAI,EAAE,CAAA;IACR,CAAC;IAED;;;;;;;OAOG;IACK,KAAK,CAAC,uBAAuB,CACnC,KAAa,EACb,UAA6B,EAC7B,QAA4B,EAC5B,MAA0B;QAE1B,MAAM,KAAK,GAAG,MAAM,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,CAAA;QACzC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,GAAG,CAAC,MAAM,KAAK,KAAK,CAAC,EAAE,CAAC;YACpD,MAAM,IAAI,eAAe,CACvB,8BAA8B,EAC9B,gEAAgE,EAChE,KAAK,CACN,CAAA;QACH,CAAC;QACD,MAAM,OAAO,GAAG,gBAAgB,CAAC,QAAQ,EAAE,UAAU,EAAE,KAAK,CAAC,IAAI,EAAE,KAAK,CAAC,MAAM,EAAE,MAAM,CAAC,CAAA;QACxF,IAAI,CAAC,OAAO,CAAC,OAAO,CAAC,EAAE,EAAE,CAAC;YACxB,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,OAAO,EAAE,GAAG,OAAO,CAAC,OAAO,CAAC,KAAK,CAAA;YACxD,MAAM,IAAI,eAAe,CAAC,IAAI,EAAE,OAAO,EAAE,OAAO,CAAC,CAAA;QACnD,CAAC;IACH,CAAC;CACF"}
|
|
@@ -0,0 +1,7 @@
|
|
|
1
|
+
import type { TensorrtLlmSettings } from './adapter.js';
|
|
2
|
+
/**
|
|
3
|
+
* Stored values, overlaid by a load's own overrides (the control API's `overrides`), reduced to this
|
|
4
|
+
* provider's keys and validated. Anything else in either object — another provider's key, a stray
|
|
5
|
+
* field a client sent — is ignored rather than passed to the engine.
|
|
6
|
+
*/
|
|
7
|
+
export declare function tensorrtLlmSettings(stored: Record<string, unknown>, overrides?: Record<string, unknown>): TensorrtLlmSettings;
|
|
@@ -0,0 +1,48 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* The `tensorrt-llm` provider settings as the settings store holds them (`src/settings/schema/
|
|
3
|
+
* tensorrt-llm.json`, task 2.14) turned into what the adapter validates (`TensorrtLlmSettings`). The
|
|
4
|
+
* store keeps "not set" as a value a settings form can show — `gpu_id: ''`, `load_timeout_seconds:
|
|
5
|
+
* 0` — while the adapter spells it `null`; this is the one place that translates between the two.
|
|
6
|
+
* Validation itself is the adapter's (`validateTensorrtLlmSettings`), so a load is refused with
|
|
7
|
+
* `INVALID_ARGUMENT` before any container exists (spec "Настройки MUST валидироваться схемой до
|
|
8
|
+
* запуска контейнера").
|
|
9
|
+
*/
|
|
10
|
+
import { validateTensorrtLlmSettings } from './adapter.js';
|
|
11
|
+
import { GENERATION_DEFAULT_KEYS } from '../managed-text/generation-defaults.js';
|
|
12
|
+
const KEYS = [
|
|
13
|
+
'gpu_id',
|
|
14
|
+
'context_length',
|
|
15
|
+
'max_output_tokens',
|
|
16
|
+
'kv_cache_free_gpu_memory_fraction',
|
|
17
|
+
'max_batch_size',
|
|
18
|
+
'kv_cache_max_tokens',
|
|
19
|
+
'cuda_graphs',
|
|
20
|
+
'kv_cache_dtype',
|
|
21
|
+
'load_timeout_seconds',
|
|
22
|
+
'enable_prefix_caching',
|
|
23
|
+
'overlap_scheduler',
|
|
24
|
+
'capacity_scheduler_policy',
|
|
25
|
+
'dtype',
|
|
26
|
+
...Object.values(GENERATION_DEFAULT_KEYS),
|
|
27
|
+
];
|
|
28
|
+
/**
|
|
29
|
+
* Stored values, overlaid by a load's own overrides (the control API's `overrides`), reduced to this
|
|
30
|
+
* provider's keys and validated. Anything else in either object — another provider's key, a stray
|
|
31
|
+
* field a client sent — is ignored rather than passed to the engine.
|
|
32
|
+
*/
|
|
33
|
+
export function tensorrtLlmSettings(stored, overrides = {}) {
|
|
34
|
+
const merged = {};
|
|
35
|
+
for (const key of KEYS) {
|
|
36
|
+
const value = key in overrides ? overrides[key] : stored[key];
|
|
37
|
+
if (value !== undefined)
|
|
38
|
+
merged[key] = value;
|
|
39
|
+
}
|
|
40
|
+
if (typeof merged['gpu_id'] === 'string' && merged['gpu_id'].trim() === '')
|
|
41
|
+
merged['gpu_id'] = null;
|
|
42
|
+
if (merged['load_timeout_seconds'] === 0)
|
|
43
|
+
merged['load_timeout_seconds'] = null;
|
|
44
|
+
if (merged['kv_cache_max_tokens'] === 0)
|
|
45
|
+
merged['kv_cache_max_tokens'] = null;
|
|
46
|
+
return validateTensorrtLlmSettings(merged);
|
|
47
|
+
}
|
|
48
|
+
//# sourceMappingURL=settings.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"settings.js","sourceRoot":"","sources":["../../../src/runtime/tensorrt-llm/settings.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,EAAE,2BAA2B,EAAE,MAAM,cAAc,CAAA;AAE1D,OAAO,EAAE,uBAAuB,EAAE,MAAM,wCAAwC,CAAA;AAEhF,MAAM,IAAI,GAAG;IACX,QAAQ;IACR,gBAAgB;IAChB,mBAAmB;IACnB,mCAAmC;IACnC,gBAAgB;IAChB,qBAAqB;IACrB,aAAa;IACb,gBAAgB;IAChB,sBAAsB;IACtB,uBAAuB;IACvB,mBAAmB;IACnB,2BAA2B;IAC3B,OAAO;IACP,GAAG,MAAM,CAAC,MAAM,CAAC,uBAAuB,CAAC;CACjC,CAAA;AAEV;;;;GAIG;AACH,MAAM,UAAU,mBAAmB,CACjC,MAA+B,EAC/B,YAAqC,EAAE;IAEvC,MAAM,MAAM,GAA4B,EAAE,CAAA;IAC1C,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;QACvB,MAAM,KAAK,GAAG,GAAG,IAAI,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,CAAA;QAC7D,IAAI,KAAK,KAAK,SAAS;YAAE,MAAM,CAAC,GAAG,CAAC,GAAG,KAAK,CAAA;IAC9C,CAAC;IACD,IAAI,OAAO,MAAM,CAAC,QAAQ,CAAC,KAAK,QAAQ,IAAI,MAAM,CAAC,QAAQ,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE;QAAE,MAAM,CAAC,QAAQ,CAAC,GAAG,IAAI,CAAA;IACnG,IAAI,MAAM,CAAC,sBAAsB,CAAC,KAAK,CAAC;QAAE,MAAM,CAAC,sBAAsB,CAAC,GAAG,IAAI,CAAA;IAC/E,IAAI,MAAM,CAAC,qBAAqB,CAAC,KAAK,CAAC;QAAE,MAAM,CAAC,qBAAqB,CAAC,GAAG,IAAI,CAAA;IAC7E,OAAO,2BAA2B,CAAC,MAAM,CAAC,CAAA;AAC5C,CAAC"}
|
|
@@ -0,0 +1,46 @@
|
|
|
1
|
+
import type { ModelFamilySupport } from '../../contracts/index.js';
|
|
2
|
+
import { type ManagedExitClassification, type ManagedLaunchContext, type ManagedEngineLaunch, type ManagedRoute, type ManagedTextAdapter, type ManagedTextCapabilities } from '../managed-text/index.js';
|
|
3
|
+
import { type VllmSettings } from './settings.js';
|
|
4
|
+
/** The port `vllm serve` listens on inside its container. */
|
|
5
|
+
export declare const VLLM_CONTAINER_PORT = 8000;
|
|
6
|
+
/** Below this much card memory, `cuda_graphs: auto` runs eager: graphs cost memory a small card needs. */
|
|
7
|
+
export declare const VLLM_CUDA_GRAPHS_MIN_VRAM_BYTES: number;
|
|
8
|
+
/** What core decides for a launch from the card as it stands right before the container (design D9). */
|
|
9
|
+
export interface VllmLaunchPlan {
|
|
10
|
+
/** `--kv-cache-memory-bytes` from the KV cache size setting; `null` lets vLLM size it within its share. */
|
|
11
|
+
kvCacheMemoryBytes: number | null;
|
|
12
|
+
/** `--gpu-memory-utilization`: what vLLM's start-up check compares the card's free memory against. */
|
|
13
|
+
gpuMemoryUtilization: number;
|
|
14
|
+
/**
|
|
15
|
+
* The checkpoint has a vision or audio part: `--limit-mm-per-prompt` then reserves nothing for it.
|
|
16
|
+
* A text-only model gets no multimodal flag at all.
|
|
17
|
+
*/
|
|
18
|
+
multimodal: boolean;
|
|
19
|
+
}
|
|
20
|
+
export declare function buildVllmLaunch(context: ManagedLaunchContext<VllmSettings>): ManagedEngineLaunch;
|
|
21
|
+
/** A base for engine start-up, imports and the first graph compilation; more per GiB of weights. */
|
|
22
|
+
export declare const VLLM_READINESS_BASE_MS: number;
|
|
23
|
+
export declare const VLLM_READINESS_PER_GIB_MS = 20000;
|
|
24
|
+
export declare function vllmReadinessTimeoutMs(weightBytes: number, settings: VllmSettings): number;
|
|
25
|
+
export declare function classifyVllmExit(log: string, exitCode: number | null): ManagedExitClassification;
|
|
26
|
+
export declare function vllmCapabilities(context: {
|
|
27
|
+
settings: VllmSettings;
|
|
28
|
+
family: ModelFamilySupport | null;
|
|
29
|
+
}): ManagedTextCapabilities;
|
|
30
|
+
export declare const VLLM_ROUTES: readonly ManagedRoute[];
|
|
31
|
+
export declare const VLLM_REWRITABLE_ROUTES: readonly ManagedRoute[];
|
|
32
|
+
/**
|
|
33
|
+
* Caps the output a request asks for at `max_output_tokens` (one key, as the client sent it); a
|
|
34
|
+
* request that asks for none is capped by vLLM's launch default instead. Refuses what the session cannot do with `:1337`'s own wording
|
|
35
|
+
* and code: tools without a parser, structured output the family does not declare, images.
|
|
36
|
+
*/
|
|
37
|
+
export declare function vllmRewriteRequestBody(route: string, body: unknown, settings: VllmSettings, capabilities?: ManagedTextCapabilities): unknown;
|
|
38
|
+
/** vLLM's context overflow as OpenAI's `context_length_exceeded`; anything else relayed as it is. */
|
|
39
|
+
export declare function mapVllmContextLengthError(status: number, body: string): object | null;
|
|
40
|
+
/**
|
|
41
|
+
* Reasoning parsers that may file a no-thinking reply under `reasoning_content` (design D8: the same
|
|
42
|
+
* hook as TensorRT-LLM's; the live run of task 6.1 confirms which of vLLM 0.31's do).
|
|
43
|
+
*/
|
|
44
|
+
export declare const VLLM_REASONING_AT_START_PARSERS: ReadonlySet<string>;
|
|
45
|
+
export declare function vllmRewriteResponseFor(route: string, requestBody: unknown, family: ModelFamilySupport | null): ((json: Record<string, unknown>) => Record<string, unknown>) | null;
|
|
46
|
+
export declare const vllmAdapter: ManagedTextAdapter<VllmSettings>;
|
|
@@ -0,0 +1,425 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* The `vllm` engine's `ManagedTextAdapter` (change `add-vllm-runtime`, task 3.2, design D8; spec
|
|
3
|
+
* `vllm-runtime`): how core runs `vllm serve` from the `vllm/vllm-openai` image on the shared
|
|
4
|
+
* managed-text lifecycle.
|
|
5
|
+
*
|
|
6
|
+
* - **argv**: `vllm serve <model> --served-model-name <id> --host 0.0.0.0 --port 8000
|
|
7
|
+
* --max-model-len <ctx> --max-num-seqs <n> --gpu-memory-utilization <U>
|
|
8
|
+
* --override-generation-config {"max_new_tokens":<cap>, …sampling defaults}`,
|
|
9
|
+
* `--limit-mm-per-prompt {"image":0,"video":0}` for a multimodal checkpoint, the further options the
|
|
10
|
+
* settings set (`--max-num-batched-tokens`, `--no-enable-prefix-caching`, `--cpu-offload-gb`,
|
|
11
|
+
* `--dtype`, `--seed`, `--async-scheduling`, `--kv-cache-memory-bytes` when the KV cache size is
|
|
12
|
+
* set), plus `--enforce-eager` when CUDA graphs are off,
|
|
13
|
+
* `--kv-cache-dtype fp8` on compute capability 8.9 and newer, and the family's tool-call and
|
|
14
|
+
* reasoning parsers. `U` is core's (`plan`, from the card as it stands right before the container
|
|
15
|
+
* is created, design D9): vLLM's default share of the card's *total* memory would refuse to start on
|
|
16
|
+
* any desktop whose card is partly taken. Within `U`, vLLM sizes the KV cache itself — it knows every
|
|
17
|
+
* model's KV shape, hybrid ones included — unless the KV cache size setting fixes it. Never `--trust-remote-code` (no code from a model repository runs), `--enable-log-requests`
|
|
18
|
+
* (prompts would land in the logs core serves) or `--api-key` (the session gateway checks the key).
|
|
19
|
+
* - **env**: usage stats off (`VLLM_NO_USAGE_STATS`, `DO_NOT_TRACK`), Hugging Face offline (every file
|
|
20
|
+
* is already in the store), and every compile cache — vLLM's own, Triton's, inductor's, FlashInfer's
|
|
21
|
+
* — under the engine cache, so a second start of the same model reuses the first one's work.
|
|
22
|
+
* - **readiness** `GET /health`; **routes** chat completions, completions and the model list;
|
|
23
|
+
* requests capped at the output setting, refused for tools without a parser, structured output the
|
|
24
|
+
* family does not declare, and images (a cap the client did not ask for comes from the launch's
|
|
25
|
+
* default, so a long prompt is not refused for the room a written cap would take); vLLM's context overflow answered as OpenAI's
|
|
26
|
+
* `context_length_exceeded` with both numbers.
|
|
27
|
+
*
|
|
28
|
+
* Log lines, exit wording and error bodies follow vLLM 0.31's source (`test/helpers/vllm-log-fixtures.ts`);
|
|
29
|
+
* the live run of task 6.1 confirms or replaces them.
|
|
30
|
+
*/
|
|
31
|
+
import { AtomicCoreError } from '../../contracts/index.js';
|
|
32
|
+
import { MANAGED_TEXT_ADAPTER_CONTRACT_VERSION, ManagedRequestRefusal, asksForStructuredOutput, asksForTools, reasoningIntoContent, thinkingRequested, } from '../managed-text/index.js';
|
|
33
|
+
import { validateVllmSettings } from './settings.js';
|
|
34
|
+
const GiB = 1024 ** 3;
|
|
35
|
+
/** The port `vllm serve` listens on inside its container. */
|
|
36
|
+
export const VLLM_CONTAINER_PORT = 8000;
|
|
37
|
+
/** Below this much card memory, `cuda_graphs: auto` runs eager: graphs cost memory a small card needs. */
|
|
38
|
+
export const VLLM_CUDA_GRAPHS_MIN_VRAM_BYTES = 12 * GiB;
|
|
39
|
+
function isPlan(value) {
|
|
40
|
+
if (value === null || typeof value !== 'object')
|
|
41
|
+
return false;
|
|
42
|
+
const plan = value;
|
|
43
|
+
return ((plan['kvCacheMemoryBytes'] === null ||
|
|
44
|
+
(typeof plan['kvCacheMemoryBytes'] === 'number' &&
|
|
45
|
+
Number.isInteger(plan['kvCacheMemoryBytes']) &&
|
|
46
|
+
plan['kvCacheMemoryBytes'] > 0)) &&
|
|
47
|
+
typeof plan['gpuMemoryUtilization'] === 'number' &&
|
|
48
|
+
plan['gpuMemoryUtilization'] > 0 &&
|
|
49
|
+
plan['gpuMemoryUtilization'] <= 1 &&
|
|
50
|
+
typeof plan['multimodal'] === 'boolean');
|
|
51
|
+
}
|
|
52
|
+
/** Compute capability 8.9 (Ada) and newer: FP8 KV cache. Unknown or older: the model's own precision. */
|
|
53
|
+
function supportsFp8Kv(computeCapability) {
|
|
54
|
+
if (computeCapability === null || computeCapability === undefined)
|
|
55
|
+
return false;
|
|
56
|
+
const [major, minor] = computeCapability.split('.').map((part) => Number.parseInt(part, 10));
|
|
57
|
+
if (major === undefined || Number.isNaN(major))
|
|
58
|
+
return false;
|
|
59
|
+
return major > 8 || (major === 8 && (minor ?? 0) >= 9);
|
|
60
|
+
}
|
|
61
|
+
function eager(context) {
|
|
62
|
+
const mode = context.settings.cuda_graphs;
|
|
63
|
+
if (mode === 'on')
|
|
64
|
+
return false;
|
|
65
|
+
if (mode === 'off')
|
|
66
|
+
return true;
|
|
67
|
+
// A unified-memory card has the host's memory: never the small-card case.
|
|
68
|
+
if (context.unifiedMemory)
|
|
69
|
+
return false;
|
|
70
|
+
const total = context.gpuTotalVramBytes ?? null;
|
|
71
|
+
return total === null || total < VLLM_CUDA_GRAPHS_MIN_VRAM_BYTES;
|
|
72
|
+
}
|
|
73
|
+
/** Every compile cache of the engine, under the engine cache (`ManagedLaunchContext.engineCachePath`). */
|
|
74
|
+
function cacheEnv(cache) {
|
|
75
|
+
return {
|
|
76
|
+
VLLM_CACHE_ROOT: `${cache}/vllm`,
|
|
77
|
+
TRITON_CACHE_DIR: `${cache}/triton`,
|
|
78
|
+
TORCHINDUCTOR_CACHE_DIR: `${cache}/inductor`,
|
|
79
|
+
FLASHINFER_WORKSPACE_BASE: `${cache}/flashinfer`,
|
|
80
|
+
};
|
|
81
|
+
}
|
|
82
|
+
export function buildVllmLaunch(context) {
|
|
83
|
+
if (!isPlan(context.plan)) {
|
|
84
|
+
throw new AtomicCoreError('INVALID_ARGUMENT', 'A vLLM launch needs core’s memory plan (KV cache bytes and memory share), computed right before the container is created.');
|
|
85
|
+
}
|
|
86
|
+
const { settings, family } = context;
|
|
87
|
+
const argv = [
|
|
88
|
+
'vllm',
|
|
89
|
+
'serve',
|
|
90
|
+
context.modelPath,
|
|
91
|
+
'--served-model-name',
|
|
92
|
+
context.modelId,
|
|
93
|
+
'--host',
|
|
94
|
+
'0.0.0.0',
|
|
95
|
+
'--port',
|
|
96
|
+
String(VLLM_CONTAINER_PORT),
|
|
97
|
+
'--max-model-len',
|
|
98
|
+
String(settings.context_length),
|
|
99
|
+
'--max-num-seqs',
|
|
100
|
+
String(settings.max_num_seqs),
|
|
101
|
+
'--gpu-memory-utilization',
|
|
102
|
+
String(Number(context.plan.gpuMemoryUtilization.toFixed(4))),
|
|
103
|
+
// The output cap for a request that names none: vLLM's own default, so a long prompt keeps the
|
|
104
|
+
// room it has (a cap written into every request would make vLLM refuse input + cap > context).
|
|
105
|
+
'--override-generation-config',
|
|
106
|
+
JSON.stringify({ max_new_tokens: settings.max_output_tokens, ...settings.generation }),
|
|
107
|
+
];
|
|
108
|
+
if (context.plan.kvCacheMemoryBytes !== null) {
|
|
109
|
+
argv.push('--kv-cache-memory-bytes', String(context.plan.kvCacheMemoryBytes));
|
|
110
|
+
}
|
|
111
|
+
if (settings.max_num_batched_tokens !== null) {
|
|
112
|
+
argv.push('--max-num-batched-tokens', String(settings.max_num_batched_tokens));
|
|
113
|
+
}
|
|
114
|
+
if (!settings.enable_prefix_caching)
|
|
115
|
+
argv.push('--no-enable-prefix-caching');
|
|
116
|
+
if (settings.cpu_offload_gb > 0)
|
|
117
|
+
argv.push('--cpu-offload-gb', String(settings.cpu_offload_gb));
|
|
118
|
+
if (settings.dtype !== 'auto')
|
|
119
|
+
argv.push('--dtype', settings.dtype);
|
|
120
|
+
if (settings.seed !== null)
|
|
121
|
+
argv.push('--seed', String(settings.seed));
|
|
122
|
+
if (settings.async_scheduling)
|
|
123
|
+
argv.push('--async-scheduling');
|
|
124
|
+
if (context.plan.multimodal)
|
|
125
|
+
argv.push('--limit-mm-per-prompt', JSON.stringify({ image: 0, video: 0 }));
|
|
126
|
+
if (eager(context))
|
|
127
|
+
argv.push('--enforce-eager');
|
|
128
|
+
if (settings.kv_cache_dtype === 'fp8' && supportsFp8Kv(context.gpuComputeCapability)) {
|
|
129
|
+
argv.push('--kv-cache-dtype', 'fp8');
|
|
130
|
+
}
|
|
131
|
+
if (family?.tool_parser)
|
|
132
|
+
argv.push('--enable-auto-tool-choice', '--tool-call-parser', family.tool_parser);
|
|
133
|
+
if (family?.reasoning_parser)
|
|
134
|
+
argv.push('--reasoning-parser', family.reasoning_parser);
|
|
135
|
+
return {
|
|
136
|
+
engine: { container_port: VLLM_CONTAINER_PORT },
|
|
137
|
+
argv,
|
|
138
|
+
env: {
|
|
139
|
+
VLLM_NO_USAGE_STATS: '1',
|
|
140
|
+
DO_NOT_TRACK: '1',
|
|
141
|
+
HF_HUB_OFFLINE: '1',
|
|
142
|
+
TRANSFORMERS_OFFLINE: '1',
|
|
143
|
+
...cacheEnv(context.engineCachePath),
|
|
144
|
+
},
|
|
145
|
+
};
|
|
146
|
+
}
|
|
147
|
+
/** A base for engine start-up, imports and the first graph compilation; more per GiB of weights. */
|
|
148
|
+
export const VLLM_READINESS_BASE_MS = 10 * 60_000;
|
|
149
|
+
export const VLLM_READINESS_PER_GIB_MS = 20_000;
|
|
150
|
+
export function vllmReadinessTimeoutMs(weightBytes, settings) {
|
|
151
|
+
if (settings.load_timeout_seconds !== null)
|
|
152
|
+
return settings.load_timeout_seconds * 1000;
|
|
153
|
+
return Math.ceil(VLLM_READINESS_BASE_MS + (weightBytes / GiB) * VLLM_READINESS_PER_GIB_MS);
|
|
154
|
+
}
|
|
155
|
+
// ---------------------------------------------------------------------------------------------
|
|
156
|
+
// Exit classification
|
|
157
|
+
// ---------------------------------------------------------------------------------------------
|
|
158
|
+
const OOM = /CUDA out of memory\. Tried to allocate ([\d.]+) ?(GiB|MiB|KiB).*?of which ([\d.]+) ?(GiB|MiB|KiB|bytes) is free/;
|
|
159
|
+
// vLLM 0.31 names the device (`on device cuda:0 (6.89/8.0 GiB)`); earlier releases did not.
|
|
160
|
+
const FREE_BELOW_SHARE = /Free memory on device(?: [^\s(]+)? \(([\d.]+)\/([\d.]+) GiB\) on startup is less than desired GPU memory utilization \(([\d.]+), ([\d.]+) GiB\)/;
|
|
161
|
+
const KV_TOO_SMALL = /max seq len \((\d+)\).*?\(([\d.]+) GiB KV cache is needed, which is larger than the available KV cache memory \(([\d.]+) GiB\)\.(?: Based on the available memory, the estimated maximum model length is (\d+)\.)?/;
|
|
162
|
+
const UNSUPPORTED_ARCHITECTURE = /Model architectures \[([^\]]*)\] are not supported for now/;
|
|
163
|
+
const ANY_OOM = /out of memory/i;
|
|
164
|
+
const toGiB = (amount, unit) => unit === 'GiB'
|
|
165
|
+
? amount
|
|
166
|
+
: unit === 'MiB'
|
|
167
|
+
? amount / 1024
|
|
168
|
+
: unit === 'KiB'
|
|
169
|
+
? amount / 1024 ** 2
|
|
170
|
+
: amount / GiB;
|
|
171
|
+
const round = (value) => Math.round(value * 100) / 100;
|
|
172
|
+
export function classifyVllmExit(log, exitCode) {
|
|
173
|
+
const kv = KV_TOO_SMALL.exec(log);
|
|
174
|
+
if (kv !== null) {
|
|
175
|
+
const maxModelLen = Number(kv[1]);
|
|
176
|
+
const fits = kv[4] === undefined ? null : Number(kv[4]);
|
|
177
|
+
return {
|
|
178
|
+
kind: 'out-of-memory',
|
|
179
|
+
message: `The context length ${maxModelLen} does not fit the KV cache: it needs ${kv[2]} GiB, ${kv[3]} GiB is available` +
|
|
180
|
+
(fits === null ? '.' : `, which holds ${fits} tokens. Lower the context length or free GPU memory.`),
|
|
181
|
+
numbers: {
|
|
182
|
+
max_model_len: maxModelLen,
|
|
183
|
+
kv_needed_gib: Number(kv[2]),
|
|
184
|
+
kv_available_gib: Number(kv[3]),
|
|
185
|
+
...(fits === null ? {} : { kv_cache_tokens: fits }),
|
|
186
|
+
},
|
|
187
|
+
excerpt: kv[0],
|
|
188
|
+
};
|
|
189
|
+
}
|
|
190
|
+
const free = FREE_BELOW_SHARE.exec(log);
|
|
191
|
+
if (free !== null) {
|
|
192
|
+
return {
|
|
193
|
+
kind: 'out-of-memory',
|
|
194
|
+
message: `The GPU had ${free[1]} GiB of ${free[2]} GiB free once vLLM had started, less than the ${free[4]} GiB ` +
|
|
195
|
+
`the model's weights and KV cache need. Free GPU memory, or lower the context length or concurrent requests.`,
|
|
196
|
+
numbers: {
|
|
197
|
+
free_gib: Number(free[1]),
|
|
198
|
+
total_gib: Number(free[2]),
|
|
199
|
+
utilization: Number(free[3]),
|
|
200
|
+
requested_gib: Number(free[4]),
|
|
201
|
+
},
|
|
202
|
+
excerpt: free[0],
|
|
203
|
+
};
|
|
204
|
+
}
|
|
205
|
+
const oom = OOM.exec(log);
|
|
206
|
+
if (oom !== null) {
|
|
207
|
+
const tried = round(toGiB(Number(oom[1]), oom[2]));
|
|
208
|
+
const left = round(toGiB(Number(oom[3]), oom[4]));
|
|
209
|
+
return {
|
|
210
|
+
kind: 'out-of-memory',
|
|
211
|
+
message: `The GPU ran out of memory: vLLM tried to allocate ${oom[1]} ${oom[2]} with ${oom[3]} ${oom[4]} free.`,
|
|
212
|
+
numbers: { requested_gib: tried, free_gib: left },
|
|
213
|
+
excerpt: oom[0],
|
|
214
|
+
};
|
|
215
|
+
}
|
|
216
|
+
const arch = UNSUPPORTED_ARCHITECTURE.exec(log);
|
|
217
|
+
if (arch !== null) {
|
|
218
|
+
return {
|
|
219
|
+
kind: 'unsupported-model',
|
|
220
|
+
message: `This vLLM release cannot run the model's architecture: ${arch[1]?.replace(/'/g, '')}.`,
|
|
221
|
+
excerpt: arch[0],
|
|
222
|
+
};
|
|
223
|
+
}
|
|
224
|
+
if (ANY_OOM.test(log)) {
|
|
225
|
+
return { kind: 'out-of-memory', message: 'The GPU ran out of memory while vLLM started.' };
|
|
226
|
+
}
|
|
227
|
+
return {
|
|
228
|
+
kind: 'other',
|
|
229
|
+
message: `vLLM exited${exitCode === null ? '' : ` with code ${exitCode}`} before it was ready.`,
|
|
230
|
+
};
|
|
231
|
+
}
|
|
232
|
+
// ---------------------------------------------------------------------------------------------
|
|
233
|
+
// Capabilities, routes, request and response rewrites
|
|
234
|
+
// ---------------------------------------------------------------------------------------------
|
|
235
|
+
export function vllmCapabilities(context) {
|
|
236
|
+
const family = context.family;
|
|
237
|
+
return {
|
|
238
|
+
tools: Boolean(family?.tool_parser),
|
|
239
|
+
reasoning: Boolean(family?.reasoning_parser),
|
|
240
|
+
structured_output: family?.structured_output === true,
|
|
241
|
+
vision: false,
|
|
242
|
+
embeddings: false,
|
|
243
|
+
responses: false,
|
|
244
|
+
};
|
|
245
|
+
}
|
|
246
|
+
export const VLLM_ROUTES = [
|
|
247
|
+
{ method: 'POST', path: '/v1/chat/completions' },
|
|
248
|
+
{ method: 'POST', path: '/v1/completions' },
|
|
249
|
+
{ method: 'GET', path: '/v1/models' },
|
|
250
|
+
];
|
|
251
|
+
export const VLLM_REWRITABLE_ROUTES = [
|
|
252
|
+
{ method: 'POST', path: '/v1/chat/completions' },
|
|
253
|
+
{ method: 'POST', path: '/v1/completions' },
|
|
254
|
+
];
|
|
255
|
+
const REWRITABLE = new Set(VLLM_REWRITABLE_ROUTES.map((route) => route.path));
|
|
256
|
+
function positive(body, key) {
|
|
257
|
+
if (!(key in body) || body[key] === null)
|
|
258
|
+
return { present: false, value: 0 };
|
|
259
|
+
const value = body[key];
|
|
260
|
+
if (typeof value !== 'number' || !Number.isInteger(value) || value <= 0) {
|
|
261
|
+
throw new AtomicCoreError('INVALID_ARGUMENT', `${key} must be a positive integer.`);
|
|
262
|
+
}
|
|
263
|
+
return { present: true, value };
|
|
264
|
+
}
|
|
265
|
+
/** Any message part that is not text: an image, a video, audio — none of which this engine accepts here. */
|
|
266
|
+
const MEDIA_PART = new Set([
|
|
267
|
+
'image_url',
|
|
268
|
+
'input_image',
|
|
269
|
+
'image',
|
|
270
|
+
'video_url',
|
|
271
|
+
'video',
|
|
272
|
+
'input_audio',
|
|
273
|
+
'audio_url',
|
|
274
|
+
]);
|
|
275
|
+
function carriesMedia(body) {
|
|
276
|
+
const messages = body['messages'];
|
|
277
|
+
if (!Array.isArray(messages))
|
|
278
|
+
return false;
|
|
279
|
+
return messages.some((message) => {
|
|
280
|
+
const content = message !== null && typeof message === 'object' ? message.content : null;
|
|
281
|
+
return (Array.isArray(content) &&
|
|
282
|
+
content.some((part) => part !== null &&
|
|
283
|
+
typeof part === 'object' &&
|
|
284
|
+
MEDIA_PART.has(String(part.type))));
|
|
285
|
+
});
|
|
286
|
+
}
|
|
287
|
+
/**
|
|
288
|
+
* Caps the output a request asks for at `max_output_tokens` (one key, as the client sent it); a
|
|
289
|
+
* request that asks for none is capped by vLLM's launch default instead. Refuses what the session cannot do with `:1337`'s own wording
|
|
290
|
+
* and code: tools without a parser, structured output the family does not declare, images.
|
|
291
|
+
*/
|
|
292
|
+
export function vllmRewriteRequestBody(route, body, settings, capabilities) {
|
|
293
|
+
if (!REWRITABLE.has(route))
|
|
294
|
+
return body;
|
|
295
|
+
if (typeof body !== 'object' || body === null || Array.isArray(body))
|
|
296
|
+
return body;
|
|
297
|
+
const client = body;
|
|
298
|
+
const model = typeof client['model'] === 'string' ? `The model '${client['model']}'` : 'This model';
|
|
299
|
+
if (carriesMedia(client)) {
|
|
300
|
+
throw new ManagedRequestRefusal(`${model} does not support images.`, 'unsupported_capability');
|
|
301
|
+
}
|
|
302
|
+
if (capabilities !== undefined) {
|
|
303
|
+
if (!capabilities.tools && asksForTools(client)) {
|
|
304
|
+
throw new ManagedRequestRefusal(`${model} does not support tool calling.`, 'unsupported_capability');
|
|
305
|
+
}
|
|
306
|
+
if (!capabilities.structured_output && asksForStructuredOutput(client)) {
|
|
307
|
+
throw new ManagedRequestRefusal(`${model} does not support structured output.`, 'unsupported_capability');
|
|
308
|
+
}
|
|
309
|
+
}
|
|
310
|
+
const cap = settings.max_output_tokens;
|
|
311
|
+
if (route === '/v1/completions') {
|
|
312
|
+
const field = positive(client, 'max_tokens');
|
|
313
|
+
return field.present ? { ...client, max_tokens: Math.min(field.value, cap) } : client;
|
|
314
|
+
}
|
|
315
|
+
const legacy = positive(client, 'max_tokens');
|
|
316
|
+
const modern = positive(client, 'max_completion_tokens');
|
|
317
|
+
const asked = [legacy, modern].filter((field) => field.present).map((field) => field.value);
|
|
318
|
+
// Nothing asked: vLLM's launch default (`--override-generation-config`) caps it, within the room left.
|
|
319
|
+
if (asked.length === 0)
|
|
320
|
+
return client;
|
|
321
|
+
const rest = { ...client };
|
|
322
|
+
delete rest['max_tokens'];
|
|
323
|
+
delete rest['max_completion_tokens'];
|
|
324
|
+
rest[modern.present ? 'max_completion_tokens' : 'max_tokens'] = Math.min(cap, ...asked);
|
|
325
|
+
return rest;
|
|
326
|
+
}
|
|
327
|
+
/** vLLM's 400 wording for a request over the context, current and older, with the limit and the request. */
|
|
328
|
+
const OVERFLOW = [
|
|
329
|
+
/maximum context length is (\d+) tokens\. However, your request has (\d+) input tokens/,
|
|
330
|
+
/maximum context length is (\d+) tokens\. However, you requested (\d+) tokens/,
|
|
331
|
+
/maximum context length is (\d+) tokens and your request has (\d+) input tokens/,
|
|
332
|
+
];
|
|
333
|
+
function errorMessage(body) {
|
|
334
|
+
try {
|
|
335
|
+
const parsed = JSON.parse(body);
|
|
336
|
+
const message = parsed.error?.message ?? parsed.message;
|
|
337
|
+
return typeof message === 'string' ? message : '';
|
|
338
|
+
}
|
|
339
|
+
catch {
|
|
340
|
+
return '';
|
|
341
|
+
}
|
|
342
|
+
}
|
|
343
|
+
/** vLLM's context overflow as OpenAI's `context_length_exceeded`; anything else relayed as it is. */
|
|
344
|
+
export function mapVllmContextLengthError(status, body) {
|
|
345
|
+
if (status !== 400)
|
|
346
|
+
return null;
|
|
347
|
+
const message = errorMessage(body);
|
|
348
|
+
for (const pattern of OVERFLOW) {
|
|
349
|
+
const match = pattern.exec(message);
|
|
350
|
+
if (match === null)
|
|
351
|
+
continue;
|
|
352
|
+
const limit = Number(match[1]);
|
|
353
|
+
const requested = Number(match[2]);
|
|
354
|
+
return {
|
|
355
|
+
error: {
|
|
356
|
+
message: `This model's maximum context length is ${limit} tokens, but the request has ${requested} tokens. Shorten the conversation or the requested output.`,
|
|
357
|
+
type: 'invalid_request_error',
|
|
358
|
+
param: 'messages',
|
|
359
|
+
code: 'context_length_exceeded',
|
|
360
|
+
},
|
|
361
|
+
};
|
|
362
|
+
}
|
|
363
|
+
return null;
|
|
364
|
+
}
|
|
365
|
+
/**
|
|
366
|
+
* Reasoning parsers that may file a no-thinking reply under `reasoning_content` (design D8: the same
|
|
367
|
+
* hook as TensorRT-LLM's; the live run of task 6.1 confirms which of vLLM 0.31's do).
|
|
368
|
+
*/
|
|
369
|
+
export const VLLM_REASONING_AT_START_PARSERS = new Set(['qwen3']);
|
|
370
|
+
export function vllmRewriteResponseFor(route, requestBody, family) {
|
|
371
|
+
if (route !== '/v1/chat/completions')
|
|
372
|
+
return null;
|
|
373
|
+
const parser = family?.reasoning_parser ?? null;
|
|
374
|
+
if (parser === null || !VLLM_REASONING_AT_START_PARSERS.has(parser))
|
|
375
|
+
return null;
|
|
376
|
+
return thinkingRequested(requestBody) ? null : reasoningIntoContent;
|
|
377
|
+
}
|
|
378
|
+
/**
|
|
379
|
+
* `initializing-engine` from the first line that shows vLLM working on the model: reading weights,
|
|
380
|
+
* profiling memory, compiling, capturing CUDA graphs. Before them the load stays in
|
|
381
|
+
* `starting-container` (Python, CUDA and the API server coming up).
|
|
382
|
+
*/
|
|
383
|
+
const STAGE_MARKERS = [
|
|
384
|
+
{ stage: 'initializing-engine', pattern: /Starting to load model/ },
|
|
385
|
+
{ stage: 'initializing-engine', pattern: /Loading safetensors checkpoint shards/ },
|
|
386
|
+
{ stage: 'initializing-engine', pattern: /Model loading took/ },
|
|
387
|
+
{ stage: 'initializing-engine', pattern: /Compiling a graph for dynamic shape/ },
|
|
388
|
+
{ stage: 'initializing-engine', pattern: /Available KV cache memory/ },
|
|
389
|
+
{ stage: 'initializing-engine', pattern: /Capturing CUDA graphs/ },
|
|
390
|
+
];
|
|
391
|
+
export const vllmAdapter = {
|
|
392
|
+
id: 'vllm',
|
|
393
|
+
contractVersion: MANAGED_TEXT_ADAPTER_CONTRACT_VERSION,
|
|
394
|
+
readiness: { path: '/health', expectedStatus: 200 },
|
|
395
|
+
routes: VLLM_ROUTES,
|
|
396
|
+
rewritableRoutes: VLLM_REWRITABLE_ROUTES,
|
|
397
|
+
stageMarkers: STAGE_MARKERS,
|
|
398
|
+
validateSettings: (raw) => validateVllmSettings((raw ?? {})),
|
|
399
|
+
buildLaunch: buildVllmLaunch,
|
|
400
|
+
readinessTimeoutMs: vllmReadinessTimeoutMs,
|
|
401
|
+
classifyExit: classifyVllmExit,
|
|
402
|
+
capabilities: vllmCapabilities,
|
|
403
|
+
rewriteRequestBody: vllmRewriteRequestBody,
|
|
404
|
+
rewriteResponseFor: vllmRewriteResponseFor,
|
|
405
|
+
mapErrorResponse: (_route, status, body) => mapVllmContextLengthError(status, body),
|
|
406
|
+
// Everything vLLM is started with — the output cap too, its default for a request that names none;
|
|
407
|
+
// only the load timeout is a load's alone.
|
|
408
|
+
restartKey: (settings) => [
|
|
409
|
+
settings.context_length,
|
|
410
|
+
settings.max_output_tokens,
|
|
411
|
+
settings.max_num_seqs,
|
|
412
|
+
settings.kv_cache_memory_gib,
|
|
413
|
+
settings.gpu_memory_utilization,
|
|
414
|
+
settings.cuda_graphs,
|
|
415
|
+
settings.kv_cache_dtype,
|
|
416
|
+
settings.max_num_batched_tokens,
|
|
417
|
+
settings.enable_prefix_caching,
|
|
418
|
+
settings.cpu_offload_gb,
|
|
419
|
+
settings.dtype,
|
|
420
|
+
settings.seed,
|
|
421
|
+
settings.async_scheduling,
|
|
422
|
+
JSON.stringify(settings.generation),
|
|
423
|
+
],
|
|
424
|
+
};
|
|
425
|
+
//# sourceMappingURL=adapter.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"adapter.js","sourceRoot":"","sources":["../../../src/runtime/vllm/adapter.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA6BG;AACH,OAAO,EAAE,eAAe,EAAE,MAAM,0BAA0B,CAAA;AAE1D,OAAO,EACL,qCAAqC,EACrC,qBAAqB,EACrB,uBAAuB,EACvB,YAAY,EACZ,oBAAoB,EACpB,iBAAiB,GAQlB,MAAM,0BAA0B,CAAA;AACjC,OAAO,EAAE,oBAAoB,EAAqB,MAAM,eAAe,CAAA;AAEvE,MAAM,GAAG,GAAG,IAAI,IAAI,CAAC,CAAA;AAErB,6DAA6D;AAC7D,MAAM,CAAC,MAAM,mBAAmB,GAAG,IAAI,CAAA;AAEvC,0GAA0G;AAC1G,MAAM,CAAC,MAAM,+BAA+B,GAAG,EAAE,GAAG,GAAG,CAAA;AAevD,SAAS,MAAM,CAAC,KAAc;IAC5B,IAAI,KAAK,KAAK,IAAI,IAAI,OAAO,KAAK,KAAK,QAAQ;QAAE,OAAO,KAAK,CAAA;IAC7D,MAAM,IAAI,GAAG,KAAgC,CAAA;IAC7C,OAAO,CACL,CAAC,IAAI,CAAC,oBAAoB,CAAC,KAAK,IAAI;QAClC,CAAC,OAAO,IAAI,CAAC,oBAAoB,CAAC,KAAK,QAAQ;YAC7C,MAAM,CAAC,SAAS,CAAC,IAAI,CAAC,oBAAoB,CAAC,CAAC;YAC5C,IAAI,CAAC,oBAAoB,CAAC,GAAG,CAAC,CAAC,CAAC;QACpC,OAAO,IAAI,CAAC,sBAAsB,CAAC,KAAK,QAAQ;QAChD,IAAI,CAAC,sBAAsB,CAAC,GAAG,CAAC;QAChC,IAAI,CAAC,sBAAsB,CAAC,IAAI,CAAC;QACjC,OAAO,IAAI,CAAC,YAAY,CAAC,KAAK,SAAS,CACxC,CAAA;AACH,CAAC;AAED,yGAAyG;AACzG,SAAS,aAAa,CAAC,iBAA4C;IACjE,IAAI,iBAAiB,KAAK,IAAI,IAAI,iBAAiB,KAAK,SAAS;QAAE,OAAO,KAAK,CAAA;IAC/E,MAAM,CAAC,KAAK,EAAE,KAAK,CAAC,GAAG,iBAAiB,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,CAAA;IAC5F,IAAI,KAAK,KAAK,SAAS,IAAI,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC;QAAE,OAAO,KAAK,CAAA;IAC5D,OAAO,KAAK,GAAG,CAAC,IAAI,CAAC,KAAK,KAAK,CAAC,IAAI,CAAC,KAAK,IAAI,CAAC,CAAC,IAAI,CAAC,CAAC,CAAA;AACxD,CAAC;AAED,SAAS,KAAK,CAAC,OAA2C;IACxD,MAAM,IAAI,GAAG,OAAO,CAAC,QAAQ,CAAC,WAAW,CAAA;IACzC,IAAI,IAAI,KAAK,IAAI;QAAE,OAAO,KAAK,CAAA;IAC/B,IAAI,IAAI,KAAK,KAAK;QAAE,OAAO,IAAI,CAAA;IAC/B,0EAA0E;IAC1E,IAAI,OAAO,CAAC,aAAa;QAAE,OAAO,KAAK,CAAA;IACvC,MAAM,KAAK,GAAG,OAAO,CAAC,iBAAiB,IAAI,IAAI,CAAA;IAC/C,OAAO,KAAK,KAAK,IAAI,IAAI,KAAK,GAAG,+BAA+B,CAAA;AAClE,CAAC;AAED,0GAA0G;AAC1G,SAAS,QAAQ,CAAC,KAAa;IAC7B,OAAO;QACL,eAAe,EAAE,GAAG,KAAK,OAAO;QAChC,gBAAgB,EAAE,GAAG,KAAK,SAAS;QACnC,uBAAuB,EAAE,GAAG,KAAK,WAAW;QAC5C,yBAAyB,EAAE,GAAG,KAAK,aAAa;KACjD,CAAA;AACH,CAAC;AAED,MAAM,UAAU,eAAe,CAAC,OAA2C;IACzE,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,IAAI,CAAC,EAAE,CAAC;QAC1B,MAAM,IAAI,eAAe,CACvB,kBAAkB,EAClB,2HAA2H,CAC5H,CAAA;IACH,CAAC;IACD,MAAM,EAAE,QAAQ,EAAE,MAAM,EAAE,GAAG,OAAO,CAAA;IACpC,MAAM,IAAI,GAAG;QACX,MAAM;QACN,OAAO;QACP,OAAO,CAAC,SAAS;QACjB,qBAAqB;QACrB,OAAO,CAAC,OAAO;QACf,QAAQ;QACR,SAAS;QACT,QAAQ;QACR,MAAM,CAAC,mBAAmB,CAAC;QAC3B,iBAAiB;QACjB,MAAM,CAAC,QAAQ,CAAC,cAAc,CAAC;QAC/B,gBAAgB;QAChB,MAAM,CAAC,QAAQ,CAAC,YAAY,CAAC;QAC7B,0BAA0B;QAC1B,MAAM,CAAC,MAAM,CAAC,OAAO,CAAC,IAAI,CAAC,oBAAoB,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC;QAC5D,+FAA+F;QAC/F,+FAA+F;QAC/F,8BAA8B;QAC9B,IAAI,CAAC,SAAS,CAAC,EAAE,cAAc,EAAE,QAAQ,CAAC,iBAAiB,EAAE,GAAG,QAAQ,CAAC,UAAU,EAAE,CAAC;KACvF,CAAA;IACD,IAAI,OAAO,CAAC,IAAI,CAAC,kBAAkB,KAAK,IAAI,EAAE,CAAC;QAC7C,IAAI,CAAC,IAAI,CAAC,yBAAyB,EAAE,MAAM,CAAC,OAAO,CAAC,IAAI,CAAC,kBAAkB,CAAC,CAAC,CAAA;IAC/E,CAAC;IACD,IAAI,QAAQ,CAAC,sBAAsB,KAAK,IAAI,EAAE,CAAC;QAC7C,IAAI,CAAC,IAAI,CAAC,0BAA0B,EAAE,MAAM,CAAC,QAAQ,CAAC,sBAAsB,CAAC,CAAC,CAAA;IAChF,CAAC;IACD,IAAI,CAAC,QAAQ,CAAC,qBAAqB;QAAE,IAAI,CAAC,IAAI,CAAC,4BAA4B,CAAC,CAAA;IAC5E,IAAI,QAAQ,CAAC,cAAc,GAAG,CAAC;QAAE,IAAI,CAAC,IAAI,CAAC,kBAAkB,EAAE,MAAM,CAAC,QAAQ,CAAC,cAAc,CAAC,CAAC,CAAA;IAC/F,IAAI,QAAQ,CAAC,KAAK,KAAK,MAAM;QAAE,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,QAAQ,CAAC,KAAK,CAAC,CAAA;IACnE,IAAI,QAAQ,CAAC,IAAI,KAAK,IAAI;QAAE,IAAI,CAAC,IAAI,CAAC,QAAQ,EAAE,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC,CAAA;IACtE,IAAI,QAAQ,CAAC,gBAAgB;QAAE,IAAI,CAAC,IAAI,CAAC,oBAAoB,CAAC,CAAA;IAC9D,IAAI,OAAO,CAAC,IAAI,CAAC,UAAU;QAAE,IAAI,CAAC,IAAI,CAAC,uBAAuB,EAAE,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,CAAC,CAAC,CAAA;IACvG,IAAI,KAAK,CAAC,OAAO,CAAC;QAAE,IAAI,CAAC,IAAI,CAAC,iBAAiB,CAAC,CAAA;IAChD,IAAI,QAAQ,CAAC,cAAc,KAAK,KAAK,IAAI,aAAa,CAAC,OAAO,CAAC,oBAAoB,CAAC,EAAE,CAAC;QACrF,IAAI,CAAC,IAAI,CAAC,kBAAkB,EAAE,KAAK,CAAC,CAAA;IACtC,CAAC;IACD,IAAI,MAAM,EAAE,WAAW;QAAE,IAAI,CAAC,IAAI,CAAC,2BAA2B,EAAE,oBAAoB,EAAE,MAAM,CAAC,WAAW,CAAC,CAAA;IACzG,IAAI,MAAM,EAAE,gBAAgB;QAAE,IAAI,CAAC,IAAI,CAAC,oBAAoB,EAAE,MAAM,CAAC,gBAAgB,CAAC,CAAA;IACtF,OAAO;QACL,MAAM,EAAE,EAAE,cAAc,EAAE,mBAAmB,EAAE;QAC/C,IAAI;QACJ,GAAG,EAAE;YACH,mBAAmB,EAAE,GAAG;YACxB,YAAY,EAAE,GAAG;YACjB,cAAc,EAAE,GAAG;YACnB,oBAAoB,EAAE,GAAG;YACzB,GAAG,QAAQ,CAAC,OAAO,CAAC,eAAe,CAAC;SACrC;KACF,CAAA;AACH,CAAC;AAED,oGAAoG;AACpG,MAAM,CAAC,MAAM,sBAAsB,GAAG,EAAE,GAAG,MAAM,CAAA;AACjD,MAAM,CAAC,MAAM,yBAAyB,GAAG,MAAM,CAAA;AAE/C,MAAM,UAAU,sBAAsB,CAAC,WAAmB,EAAE,QAAsB;IAChF,IAAI,QAAQ,CAAC,oBAAoB,KAAK,IAAI;QAAE,OAAO,QAAQ,CAAC,oBAAoB,GAAG,IAAI,CAAA;IACvF,OAAO,IAAI,CAAC,IAAI,CAAC,sBAAsB,GAAG,CAAC,WAAW,GAAG,GAAG,CAAC,GAAG,yBAAyB,CAAC,CAAA;AAC5F,CAAC;AAED,gGAAgG;AAChG,sBAAsB;AACtB,gGAAgG;AAEhG,MAAM,GAAG,GACP,iHAAiH,CAAA;AACnH,4FAA4F;AAC5F,MAAM,gBAAgB,GACpB,iJAAiJ,CAAA;AACnJ,MAAM,YAAY,GAChB,oNAAoN,CAAA;AACtN,MAAM,wBAAwB,GAAG,4DAA4D,CAAA;AAC7F,MAAM,OAAO,GAAG,gBAAgB,CAAA;AAEhC,MAAM,KAAK,GAAG,CAAC,MAAc,EAAE,IAAY,EAAU,EAAE,CACrD,IAAI,KAAK,KAAK;IACZ,CAAC,CAAC,MAAM;IACR,CAAC,CAAC,IAAI,KAAK,KAAK;QACd,CAAC,CAAC,MAAM,GAAG,IAAI;QACf,CAAC,CAAC,IAAI,KAAK,KAAK;YACd,CAAC,CAAC,MAAM,GAAG,IAAI,IAAI,CAAC;YACpB,CAAC,CAAC,MAAM,GAAG,GAAG,CAAA;AACtB,MAAM,KAAK,GAAG,CAAC,KAAa,EAAE,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,GAAG,GAAG,CAAC,GAAG,GAAG,CAAA;AAE9D,MAAM,UAAU,gBAAgB,CAAC,GAAW,EAAE,QAAuB;IACnE,MAAM,EAAE,GAAG,YAAY,CAAC,IAAI,CAAC,GAAG,CAAC,CAAA;IACjC,IAAI,EAAE,KAAK,IAAI,EAAE,CAAC;QAChB,MAAM,WAAW,GAAG,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAA;QACjC,MAAM,IAAI,GAAG,EAAE,CAAC,CAAC,CAAC,KAAK,SAAS,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAA;QACvD,OAAO;YACL,IAAI,EAAE,eAAe;YACrB,OAAO,EACL,sBAAsB,WAAW,wCAAwC,EAAE,CAAC,CAAC,CAAC,SAAS,EAAE,CAAC,CAAC,CAAC,mBAAmB;gBAC/G,CAAC,IAAI,KAAK,IAAI,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,iBAAiB,IAAI,uDAAuD,CAAC;YACtG,OAAO,EAAE;gBACP,aAAa,EAAE,WAAW;gBAC1B,aAAa,EAAE,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;gBAC5B,gBAAgB,EAAE,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;gBAC/B,GAAG,CAAC,IAAI,KAAK,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,eAAe,EAAE,IAAI,EAAE,CAAC;aACpD;YACD,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC;SACf,CAAA;IACH,CAAC;IACD,MAAM,IAAI,GAAG,gBAAgB,CAAC,IAAI,CAAC,GAAG,CAAC,CAAA;IACvC,IAAI,IAAI,KAAK,IAAI,EAAE,CAAC;QAClB,OAAO;YACL,IAAI,EAAE,eAAe;YACrB,OAAO,EACL,eAAe,IAAI,CAAC,CAAC,CAAC,WAAW,IAAI,CAAC,CAAC,CAAC,kDAAkD,IAAI,CAAC,CAAC,CAAC,OAAO;gBACxG,6GAA6G;YAC/G,OAAO,EAAE;gBACP,QAAQ,EAAE,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBACzB,SAAS,EAAE,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBAC1B,WAAW,EAAE,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBAC5B,aAAa,EAAE,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;aAC/B;YACD,OAAO,EAAE,IAAI,CAAC,CAAC,CAAC;SACjB,CAAA;IACH,CAAC;IACD,MAAM,GAAG,GAAG,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAA;IACzB,IAAI,GAAG,KAAK,IAAI,EAAE,CAAC;QACjB,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,CAAW,CAAC,CAAC,CAAA;QAC5D,MAAM,IAAI,GAAG,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,CAAW,CAAC,CAAC,CAAA;QAC3D,OAAO;YACL,IAAI,EAAE,eAAe;YACrB,OAAO,EAAE,qDAAqD,GAAG,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,QAAQ;YAC/G,OAAO,EAAE,EAAE,aAAa,EAAE,KAAK,EAAE,QAAQ,EAAE,IAAI,EAAE;YACjD,OAAO,EAAE,GAAG,CAAC,CAAC,CAAC;SAChB,CAAA;IACH,CAAC;IACD,MAAM,IAAI,GAAG,wBAAwB,CAAC,IAAI,CAAC,GAAG,CAAC,CAAA;IAC/C,IAAI,IAAI,KAAK,IAAI,EAAE,CAAC;QAClB,OAAO;YACL,IAAI,EAAE,mBAAmB;YACzB,OAAO,EAAE,0DAA0D,IAAI,CAAC,CAAC,CAAC,EAAE,OAAO,CAAC,IAAI,EAAE,EAAE,CAAC,GAAG;YAChG,OAAO,EAAE,IAAI,CAAC,CAAC,CAAC;SACjB,CAAA;IACH,CAAC;IACD,IAAI,OAAO,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC;QACtB,OAAO,EAAE,IAAI,EAAE,eAAe,EAAE,OAAO,EAAE,+CAA+C,EAAE,CAAA;IAC5F,CAAC;IACD,OAAO;QACL,IAAI,EAAE,OAAO;QACb,OAAO,EAAE,cAAc,QAAQ,KAAK,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,cAAc,QAAQ,EAAE,uBAAuB;KAChG,CAAA;AACH,CAAC;AAED,gGAAgG;AAChG,sDAAsD;AACtD,gGAAgG;AAEhG,MAAM,UAAU,gBAAgB,CAAC,OAGhC;IACC,MAAM,MAAM,GAAG,OAAO,CAAC,MAAM,CAAA;IAC7B,OAAO;QACL,KAAK,EAAE,OAAO,CAAC,MAAM,EAAE,WAAW,CAAC;QACnC,SAAS,EAAE,OAAO,CAAC,MAAM,EAAE,gBAAgB,CAAC;QAC5C,iBAAiB,EAAE,MAAM,EAAE,iBAAiB,KAAK,IAAI;QACrD,MAAM,EAAE,KAAK;QACb,UAAU,EAAE,KAAK;QACjB,SAAS,EAAE,KAAK;KACjB,CAAA;AACH,CAAC;AAED,MAAM,CAAC,MAAM,WAAW,GAA4B;IAClD,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,sBAAsB,EAAE;IAChD,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,iBAAiB,EAAE;IAC3C,EAAE,MAAM,EAAE,KAAK,EAAE,IAAI,EAAE,YAAY,EAAE;CACtC,CAAA;AAED,MAAM,CAAC,MAAM,sBAAsB,GAA4B;IAC7D,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,sBAAsB,EAAE;IAChD,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,iBAAiB,EAAE;CAC5C,CAAA;AAED,MAAM,UAAU,GAAG,IAAI,GAAG,CAAC,sBAAsB,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAA;AAE7E,SAAS,QAAQ,CAAC,IAA6B,EAAE,GAAW;IAC1D,IAAI,CAAC,CAAC,GAAG,IAAI,IAAI,CAAC,IAAI,IAAI,CAAC,GAAG,CAAC,KAAK,IAAI;QAAE,OAAO,EAAE,OAAO,EAAE,KAAK,EAAE,KAAK,EAAE,CAAC,EAAE,CAAA;IAC7E,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,CAAA;IACvB,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,KAAK,CAAC,IAAI,KAAK,IAAI,CAAC,EAAE,CAAC;QACxE,MAAM,IAAI,eAAe,CAAC,kBAAkB,EAAE,GAAG,GAAG,8BAA8B,CAAC,CAAA;IACrF,CAAC;IACD,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,KAAK,EAAE,CAAA;AACjC,CAAC;AAED,4GAA4G;AAC5G,MAAM,UAAU,GAAG,IAAI,GAAG,CAAC;IACzB,WAAW;IACX,aAAa;IACb,OAAO;IACP,WAAW;IACX,OAAO;IACP,aAAa;IACb,WAAW;CACZ,CAAC,CAAA;AAEF,SAAS,YAAY,CAAC,IAA6B;IACjD,MAAM,QAAQ,GAAG,IAAI,CAAC,UAAU,CAAC,CAAA;IACjC,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,QAAQ,CAAC;QAAE,OAAO,KAAK,CAAA;IAC1C,OAAO,QAAQ,CAAC,IAAI,CAAC,CAAC,OAAO,EAAE,EAAE;QAC/B,MAAM,OAAO,GACX,OAAO,KAAK,IAAI,IAAI,OAAO,OAAO,KAAK,QAAQ,CAAC,CAAC,CAAE,OAAiC,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CAAA;QACrG,OAAO,CACL,KAAK,CAAC,OAAO,CAAC,OAAO,CAAC;YACtB,OAAO,CAAC,IAAI,CACV,CAAC,IAAI,EAAE,EAAE,CACP,IAAI,KAAK,IAAI;gBACb,OAAO,IAAI,KAAK,QAAQ;gBACxB,UAAU,CAAC,GAAG,CAAC,MAAM,CAAE,IAA2B,CAAC,IAAI,CAAC,CAAC,CAC5D,CACF,CAAA;IACH,CAAC,CAAC,CAAA;AACJ,CAAC;AAED;;;;GAIG;AACH,MAAM,UAAU,sBAAsB,CACpC,KAAa,EACb,IAAa,EACb,QAAsB,EACtB,YAAsC;IAEtC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC;QAAE,OAAO,IAAI,CAAA;IACvC,IAAI,OAAO,IAAI,KAAK,QAAQ,IAAI,IAAI,KAAK,IAAI,IAAI,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC;QAAE,OAAO,IAAI,CAAA;IACjF,MAAM,MAAM,GAAG,IAA+B,CAAA;IAC9C,MAAM,KAAK,GAAG,OAAO,MAAM,CAAC,OAAO,CAAC,KAAK,QAAQ,CAAC,CAAC,CAAC,cAAc,MAAM,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,YAAY,CAAA;IACnG,IAAI,YAAY,CAAC,MAAM,CAAC,EAAE,CAAC;QACzB,MAAM,IAAI,qBAAqB,CAAC,GAAG,KAAK,2BAA2B,EAAE,wBAAwB,CAAC,CAAA;IAChG,CAAC;IACD,IAAI,YAAY,KAAK,SAAS,EAAE,CAAC;QAC/B,IAAI,CAAC,YAAY,CAAC,KAAK,IAAI,YAAY,CAAC,MAAM,CAAC,EAAE,CAAC;YAChD,MAAM,IAAI,qBAAqB,CAAC,GAAG,KAAK,iCAAiC,EAAE,wBAAwB,CAAC,CAAA;QACtG,CAAC;QACD,IAAI,CAAC,YAAY,CAAC,iBAAiB,IAAI,uBAAuB,CAAC,MAAM,CAAC,EAAE,CAAC;YACvE,MAAM,IAAI,qBAAqB,CAC7B,GAAG,KAAK,sCAAsC,EAC9C,wBAAwB,CACzB,CAAA;QACH,CAAC;IACH,CAAC;IACD,MAAM,GAAG,GAAG,QAAQ,CAAC,iBAAiB,CAAA;IACtC,IAAI,KAAK,KAAK,iBAAiB,EAAE,CAAC;QAChC,MAAM,KAAK,GAAG,QAAQ,CAAC,MAAM,EAAE,YAAY,CAAC,CAAA;QAC5C,OAAO,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,GAAG,MAAM,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAA;IACvF,CAAC;IACD,MAAM,MAAM,GAAG,QAAQ,CAAC,MAAM,EAAE,YAAY,CAAC,CAAA;IAC7C,MAAM,MAAM,GAAG,QAAQ,CAAC,MAAM,EAAE,uBAAuB,CAAC,CAAA;IACxD,MAAM,KAAK,GAAG,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,KAAK,CAAC,CAAA;IAC3F,uGAAuG;IACvG,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,MAAM,CAAA;IACrC,MAAM,IAAI,GAA4B,EAAE,GAAG,MAAM,EAAE,CAAA;IACnD,OAAO,IAAI,CAAC,YAAY,CAAC,CAAA;IACzB,OAAO,IAAI,CAAC,uBAAuB,CAAC,CAAA;IACpC,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC,uBAAuB,CAAC,CAAC,CAAC,YAAY,CAAC,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,GAAG,KAAK,CAAC,CAAA;IACvF,OAAO,IAAI,CAAA;AACb,CAAC;AAED,4GAA4G;AAC5G,MAAM,QAAQ,GAAG;IACf,uFAAuF;IACvF,8EAA8E;IAC9E,gFAAgF;CACjF,CAAA;AAED,SAAS,YAAY,CAAC,IAAY;IAChC,IAAI,CAAC;QACH,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAyD,CAAA;QACvF,MAAM,OAAO,GAAG,MAAM,CAAC,KAAK,EAAE,OAAO,IAAI,MAAM,CAAC,OAAO,CAAA;QACvD,OAAO,OAAO,OAAO,KAAK,QAAQ,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAA;IACnD,CAAC;IAAC,MAAM,CAAC;QACP,OAAO,EAAE,CAAA;IACX,CAAC;AACH,CAAC;AAED,qGAAqG;AACrG,MAAM,UAAU,yBAAyB,CAAC,MAAc,EAAE,IAAY;IACpE,IAAI,MAAM,KAAK,GAAG;QAAE,OAAO,IAAI,CAAA;IAC/B,MAAM,OAAO,GAAG,YAAY,CAAC,IAAI,CAAC,CAAA;IAClC,KAAK,MAAM,OAAO,IAAI,QAAQ,EAAE,CAAC;QAC/B,MAAM,KAAK,GAAG,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,CAAA;QACnC,IAAI,KAAK,KAAK,IAAI;YAAE,SAAQ;QAC5B,MAAM,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAA;QAC9B,MAAM,SAAS,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAA;QAClC,OAAO;YACL,KAAK,EAAE;gBACL,OAAO,EAAE,0CAA0C,KAAK,gCAAgC,SAAS,4DAA4D;gBAC7J,IAAI,EAAE,uBAAuB;gBAC7B,KAAK,EAAE,UAAU;gBACjB,IAAI,EAAE,yBAAyB;aAChC;SACF,CAAA;IACH,CAAC;IACD,OAAO,IAAI,CAAA;AACb,CAAC;AAED;;;GAGG;AACH,MAAM,CAAC,MAAM,+BAA+B,GAAwB,IAAI,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAA;AAEtF,MAAM,UAAU,sBAAsB,CACpC,KAAa,EACb,WAAoB,EACpB,MAAiC;IAEjC,IAAI,KAAK,KAAK,sBAAsB;QAAE,OAAO,IAAI,CAAA;IACjD,MAAM,MAAM,GAAG,MAAM,EAAE,gBAAgB,IAAI,IAAI,CAAA;IAC/C,IAAI,MAAM,KAAK,IAAI,IAAI,CAAC,+BAA+B,CAAC,GAAG,CAAC,MAAM,CAAC;QAAE,OAAO,IAAI,CAAA;IAChF,OAAO,iBAAiB,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,oBAAoB,CAAA;AACrE,CAAC;AAED;;;;GAIG;AACH,MAAM,aAAa,GAAkC;IACnD,EAAE,KAAK,EAAE,qBAAqB,EAAE,OAAO,EAAE,wBAAwB,EAAE;IACnE,EAAE,KAAK,EAAE,qBAAqB,EAAE,OAAO,EAAE,uCAAuC,EAAE;IAClF,EAAE,KAAK,EAAE,qBAAqB,EAAE,OAAO,EAAE,oBAAoB,EAAE;IAC/D,EAAE,KAAK,EAAE,qBAAqB,EAAE,OAAO,EAAE,qCAAqC,EAAE;IAChF,EAAE,KAAK,EAAE,qBAAqB,EAAE,OAAO,EAAE,2BAA2B,EAAE;IACtE,EAAE,KAAK,EAAE,qBAAqB,EAAE,OAAO,EAAE,uBAAuB,EAAE;CACnE,CAAA;AAED,MAAM,CAAC,MAAM,WAAW,GAAqC;IAC3D,EAAE,EAAE,MAAM;IACV,eAAe,EAAE,qCAAqC;IACtD,SAAS,EAAE,EAAE,IAAI,EAAE,SAAS,EAAE,cAAc,EAAE,GAAG,EAAE;IACnD,MAAM,EAAE,WAAW;IACnB,gBAAgB,EAAE,sBAAsB;IACxC,YAAY,EAAE,aAAa;IAC3B,gBAAgB,EAAE,CAAC,GAAG,EAAE,EAAE,CAAC,oBAAoB,CAAC,CAAC,GAAG,IAAI,EAAE,CAA4B,CAAC;IACvF,WAAW,EAAE,eAAe;IAC5B,kBAAkB,EAAE,sBAAsB;IAC1C,YAAY,EAAE,gBAAgB;IAC9B,YAAY,EAAE,gBAAgB;IAC9B,kBAAkB,EAAE,sBAAsB;IAC1C,kBAAkB,EAAE,sBAAsB;IAC1C,gBAAgB,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,EAAE,CAAC,yBAAyB,CAAC,MAAM,EAAE,IAAI,CAAC;IACnF,mGAAmG;IACnG,2CAA2C;IAC3C,UAAU,EAAE,CAAC,QAAQ,EAAE,EAAE,CAAC;QACxB,QAAQ,CAAC,cAAc;QACvB,QAAQ,CAAC,iBAAiB;QAC1B,QAAQ,CAAC,YAAY;QACrB,QAAQ,CAAC,mBAAmB;QAC5B,QAAQ,CAAC,sBAAsB;QAC/B,QAAQ,CAAC,WAAW;QACpB,QAAQ,CAAC,cAAc;QACvB,QAAQ,CAAC,sBAAsB;QAC/B,QAAQ,CAAC,qBAAqB;QAC9B,QAAQ,CAAC,cAAc;QACvB,QAAQ,CAAC,KAAK;QACd,QAAQ,CAAC,IAAI;QACb,QAAQ,CAAC,gBAAgB;QACzB,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC,UAAU,CAAC;KACpC;CACF,CAAA"}
|
|
@@ -0,0 +1,16 @@
|
|
|
1
|
+
import type { ManagedEngineSpec } from '../managed-engines/spec.js';
|
|
2
|
+
import type { ManagedTextCapabilities } from '../managed-text/index.js';
|
|
3
|
+
import type { SessionRoutePolicy } from '../shared/index.js';
|
|
4
|
+
import { type VllmSettings } from './settings.js';
|
|
5
|
+
export declare const VLLM_ENGINE_ID = "vllm";
|
|
6
|
+
/**
|
|
7
|
+
* What the public server (`:1337`) must know about a `vllm` session to route it honestly: only the
|
|
8
|
+
* routes the adapter declares, tool calls and JSON output only as the session can, and vLLM's context
|
|
9
|
+
* overflow answered as `context_length_exceeded`. `null` capabilities (a session the core did not
|
|
10
|
+
* start) keep the routes and the mapping and leave tools and JSON output to the engine.
|
|
11
|
+
*/
|
|
12
|
+
export declare function vllmRoutePolicy(capabilities: ManagedTextCapabilities | null, limits?: {
|
|
13
|
+
contextLength: number;
|
|
14
|
+
maxOutputTokens: number;
|
|
15
|
+
} | null): SessionRoutePolicy;
|
|
16
|
+
export declare const VLLM_ENGINE: ManagedEngineSpec<VllmSettings>;
|
|
@@ -0,0 +1,37 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* vLLM as a managed engine (change `add-vllm-runtime`, task 3.4, design D3): its adapter, settings,
|
|
3
|
+
* memory rule and launch plan, route policy, and its descriptor in conf (`runtimes/vllm.json`, read
|
|
4
|
+
* only once conf main publishes it — until then the provider is unsupported, design D15).
|
|
5
|
+
*/
|
|
6
|
+
import { defaultDescriptorUrl } from '../environment/index.js';
|
|
7
|
+
import { VLLM_ROUTES, mapVllmContextLengthError, vllmAdapter } from './adapter.js';
|
|
8
|
+
import { vllmCheckEngine, vllmLaunchPlan } from './memory.js';
|
|
9
|
+
import { vllmSettings } from './settings.js';
|
|
10
|
+
export const VLLM_ENGINE_ID = 'vllm';
|
|
11
|
+
/**
|
|
12
|
+
* What the public server (`:1337`) must know about a `vllm` session to route it honestly: only the
|
|
13
|
+
* routes the adapter declares, tool calls and JSON output only as the session can, and vLLM's context
|
|
14
|
+
* overflow answered as `context_length_exceeded`. `null` capabilities (a session the core did not
|
|
15
|
+
* start) keep the routes and the mapping and leave tools and JSON output to the engine.
|
|
16
|
+
*/
|
|
17
|
+
export function vllmRoutePolicy(capabilities, limits = null) {
|
|
18
|
+
return {
|
|
19
|
+
routes: VLLM_ROUTES,
|
|
20
|
+
tools: capabilities?.tools ?? true,
|
|
21
|
+
structuredOutput: capabilities?.structured_output ?? true,
|
|
22
|
+
mapError: mapVllmContextLengthError,
|
|
23
|
+
...(limits === null ? {} : limits),
|
|
24
|
+
};
|
|
25
|
+
}
|
|
26
|
+
export const VLLM_ENGINE = {
|
|
27
|
+
engine_id: VLLM_ENGINE_ID,
|
|
28
|
+
provider: VLLM_ENGINE_ID,
|
|
29
|
+
label: 'vLLM',
|
|
30
|
+
descriptor: { engine_id: VLLM_ENGINE_ID, label: 'vLLM', url: defaultDescriptorUrl(VLLM_ENGINE_ID) },
|
|
31
|
+
adapter: vllmAdapter,
|
|
32
|
+
settings: vllmSettings,
|
|
33
|
+
check: vllmCheckEngine,
|
|
34
|
+
launchPlan: vllmLaunchPlan,
|
|
35
|
+
routePolicy: vllmRoutePolicy,
|
|
36
|
+
};
|
|
37
|
+
//# sourceMappingURL=engine.js.map
|