omnius 1.0.623 → 1.0.625

package/docs/DISCOVERY.json

@@ -3157,11 +3157,14 @@
           "tags": [
             "ASR"
           ],
-          "description": "Body: {engineId, modelId?, setup?, device?}. Equivalent to PATCH /v1/asr/selection.",
+          "description": "Body: {engineId, modelId?, setup?, device?}. Equivalent to PATCH /v1/asr/selection. If weights are absent, returns 202 without replacing the active model; poll statusUrl until deployment.state is active or error.",
           "responses": {
             "200": {
               "description": "Exact selection activated"
             },
+            "202": {
+              "description": "Selection is pending weight download and CUDA validation"
+            },
             "400": {
               "description": "Unknown selection"
             },
@@ -3179,11 +3182,332 @@
         "packages/cli/src/api/openapi.ts"
       ]
     },
+    {
+      "id": "api.v1-asr-downloads",
+      "kind": "api",
+      "title": "/v1/asr/downloads",
+      "summary": "List persistent ASR weight download and deployment transitions; Start or resume an ASR weight download",
+      "aliases": [
+        "/v1/asr/downloads"
+      ],
+      "keywords": [
+        "rest",
+        "openapi",
+        "GET",
+        "POST",
+        "ASR",
+        "v1",
+        "asr",
+        "downloads"
+      ],
+      "maturity": "stable",
+      "layer": "interface",
+      "audiences": [
+        "integrator",
+        "service-agent",
+        "coding-agent"
+      ],
+      "interfaces": [
+        {
+          "type": "rest",
+          "target": "GET /v1/asr/downloads"
+        },
+        {
+          "type": "rest",
+          "target": "POST /v1/asr/downloads"
+        },
+        {
+          "type": "openapi",
+          "target": "/openapi.json"
+        }
+      ],
+      "references": [
+        {
+          "type": "source",
+          "target": "packages/cli/src/api/openapi.ts",
+          "relation": "openapi-source"
+        },
+        {
+          "type": "documentation",
+          "target": "docs/reference/rest-api.md",
+          "relation": "endpoint-inventory"
+        }
+      ],
+      "methods": [
+        "GET",
+        "POST"
+      ],
+      "tags": [
+        "ASR"
+      ],
+      "operations": {
+        "get": {
+          "summary": "List persistent ASR weight download and deployment transitions",
+          "tags": [
+            "ASR"
+          ],
+          "responses": {
+            "200": {
+              "description": "Jobs with downloadedBytes, totalBytes when available, progressPercent, bytesPerSecond, etaSeconds, deployment state, errors, and links"
+            }
+          }
+        },
+        "post": {
+          "summary": "Start or resume an ASR weight download",
+          "tags": [
+            "ASR"
+          ],
+          "description": "Body: {engineId, modelId, device?}. Returns 202 immediately. Duplicate requests coalesce and Hugging Face snapshots resume after retry.",
+          "responses": {
+            "200": {
+              "description": "Weights already ready"
+            },
+            "202": {
+              "description": "pending_download with statusUrl, retryUrl, and pollAfterMs"
+            },
+            "400": {
+              "description": "Unknown or unmanaged model"
+            }
+          }
+        }
+      },
+      "source_of_truth": [
+        "GET /openapi.json",
+        "packages/cli/src/api/openapi.ts"
+      ]
+    },
+    {
+      "id": "api.v1-asr-downloads-engine-id-model-id",
+      "kind": "api",
+      "title": "/v1/asr/downloads/{engineId}/{modelId}",
+      "summary": "Poll one ASR download and deployment transition; Start or resume this model download",
+      "aliases": [
+        "/v1/asr/downloads/{engineId}/{modelId}"
+      ],
+      "keywords": [
+        "rest",
+        "openapi",
+        "GET",
+        "POST",
+        "ASR",
+        "v1",
+        "asr",
+        "downloads",
+        "{engineId}",
+        "{modelId}"
+      ],
+      "maturity": "stable",
+      "layer": "interface",
+      "audiences": [
+        "integrator",
+        "service-agent",
+        "coding-agent"
+      ],
+      "interfaces": [
+        {
+          "type": "rest",
+          "target": "GET /v1/asr/downloads/{engineId}/{modelId}"
+        },
+        {
+          "type": "rest",
+          "target": "POST /v1/asr/downloads/{engineId}/{modelId}"
+        },
+        {
+          "type": "openapi",
+          "target": "/openapi.json"
+        }
+      ],
+      "references": [
+        {
+          "type": "source",
+          "target": "packages/cli/src/api/openapi.ts",
+          "relation": "openapi-source"
+        },
+        {
+          "type": "documentation",
+          "target": "docs/reference/rest-api.md",
+          "relation": "endpoint-inventory"
+        }
+      ],
+      "methods": [
+        "GET",
+        "POST"
+      ],
+      "tags": [
+        "ASR"
+      ],
+      "operations": {
+        "get": {
+          "summary": "Poll one ASR download and deployment transition",
+          "tags": [
+            "ASR"
+          ],
+          "description": "Poll at pollAfterMs until download.state=ready and download.deployment.state=active for deploy/select operations. Terminal failures use state=error or interrupted and include retry links.",
+          "parameters": [
+            {
+              "name": "engineId",
+              "in": "path",
+              "required": true,
+              "schema": {
+                "type": "string"
+              }
+            },
+            {
+              "name": "modelId",
+              "in": "path",
+              "required": true,
+              "schema": {
+                "type": "string"
+              }
+            }
+          ],
+          "responses": {
+            "200": {
+              "description": "Current persistent transition state"
+            },
+            "404": {
+              "description": "No job exists"
+            }
+          }
+        },
+        "post": {
+          "summary": "Start or resume this model download",
+          "tags": [
+            "ASR"
+          ],
+          "parameters": [
+            {
+              "name": "engineId",
+              "in": "path",
+              "required": true,
+              "schema": {
+                "type": "string"
+              }
+            },
+            {
+              "name": "modelId",
+              "in": "path",
+              "required": true,
+              "schema": {
+                "type": "string"
+              }
+            }
+          ],
+          "responses": {
+            "200": {
+              "description": "Weights already ready"
+            },
+            "202": {
+              "description": "Download accepted or already running"
+            }
+          }
+        }
+      },
+      "source_of_truth": [
+        "GET /openapi.json",
+        "packages/cli/src/api/openapi.ts"
+      ]
+    },
+    {
+      "id": "api.v1-asr-downloads-engine-id-model-id-retry",
+      "kind": "api",
+      "title": "/v1/asr/downloads/{engineId}/{modelId}/retry",
+      "summary": "Retry or resume a failed/interrupted ASR weight download",
+      "aliases": [
+        "/v1/asr/downloads/{engineId}/{modelId}/retry"
+      ],
+      "keywords": [
+        "rest",
+        "openapi",
+        "POST",
+        "ASR",
+        "v1",
+        "asr",
+        "downloads",
+        "{engineId}",
+        "{modelId}",
+        "retry"
+      ],
+      "maturity": "stable",
+      "layer": "interface",
+      "audiences": [
+        "integrator",
+        "service-agent",
+        "coding-agent"
+      ],
+      "interfaces": [
+        {
+          "type": "rest",
+          "target": "POST /v1/asr/downloads/{engineId}/{modelId}/retry"
+        },
+        {
+          "type": "openapi",
+          "target": "/openapi.json"
+        }
+      ],
+      "references": [
+        {
+          "type": "source",
+          "target": "packages/cli/src/api/openapi.ts",
+          "relation": "openapi-source"
+        },
+        {
+          "type": "documentation",
+          "target": "docs/reference/rest-api.md",
+          "relation": "endpoint-inventory"
+        }
+      ],
+      "methods": [
+        "POST"
+      ],
+      "tags": [
+        "ASR"
+      ],
+      "operations": {
+        "post": {
+          "summary": "Retry or resume a failed/interrupted ASR weight download",
+          "tags": [
+            "ASR"
+          ],
+          "description": "Safe to call repeatedly. Existing active work is deduplicated.",
+          "parameters": [
+            {
+              "name": "engineId",
+              "in": "path",
+              "required": true,
+              "schema": {
+                "type": "string"
+              }
+            },
+            {
+              "name": "modelId",
+              "in": "path",
+              "required": true,
+              "schema": {
+                "type": "string"
+              }
+            }
+          ],
+          "responses": {
+            "200": {
+              "description": "Weights already ready"
+            },
+            "202": {
+              "description": "Retry accepted"
+            }
+          }
+        }
+      },
+      "source_of_truth": [
+        "GET /openapi.json",
+        "packages/cli/src/api/openapi.ts"
+      ]
+    },
     {
       "id": "api.v1-asr-engines",
       "kind": "api",
       "title": "/v1/asr/engines",
-      "summary": "List all ASR engines, models, capabilities, resource requirements, readiness, and active selection",
+      "summary": "List all ASR engines, models, capabilities, resource requirements, readiness, download transitions, and active selection",
       "aliases": [
         "/v1/asr/engines"
       ],
@@ -3233,13 +3557,14 @@
       ],
       "operations": {
         "get": {
-          "summary": "List all ASR engines, models, capabilities, resource requirements, readiness, and active selection",
+          "summary": "List all ASR engines, models, capabilities, resource requirements, readiness, download transitions, and active selection",
           "tags": [
             "ASR"
           ],
+          "description": "Each managed model includes a nullable download object with state, byte progress, rate, ETA, deployment state, and agent-followable links.",
           "responses": {
             "200": {
-              "description": "Canonical ASR registry and runtime readiness"
+              "description": "Canonical ASR registry, runtime readiness, and transitional download state"
             }
           }
         }
@@ -3332,7 +3657,10 @@
           ],
           "responses": {
             "200": {
-              "description": "Model deployed and selected"
+              "description": "Model already deployed and selected"
+            },
+            "202": {
+              "description": "Deployment pending; poll statusUrl until deployment.state=active"
             },
             "500": {
               "description": "Deploy or CUDA validation failed"
@@ -3407,7 +3735,7 @@
           "tags": [
             "ASR"
           ],
-          "description": "Optional body: {device}. Downloads model weights into Omnius-managed storage and verifies CUDA placement for Whisper and Nemotron.",
+          "description": "Optional body: {device}. Downloads model weights into Omnius-managed storage and verifies CUDA placement for Whisper, Nemotron, and Voxtral. Voxtral model IDs are voxtral-mini-4b-realtime-2602, voxtral-mini-3b-2507, and voxtral-small-24b-2507.",
           "parameters": [
             {
               "name": "engineId",
@@ -3428,7 +3756,10 @@
           ],
           "responses": {
             "200": {
-              "description": "Weights are ready"
+              "description": "Weights already ready"
+            },
+            "202": {
+              "description": "Download accepted with polling and retry links"
             },
             "500": {
               "description": "Runtime, download, or CUDA validation failed"
@@ -3501,7 +3832,7 @@
           "tags": [
             "ASR"
           ],
-          "description": "Body: {modelId?, device?}. Installs the managed runtime and pulls the requested model. Whisper and Nemotron loads validate their selected CUDA device; VibeVoice stores its pinned model and tokenizer snapshot under Omnius' ASR runtime directories.",
+          "description": "Body: {modelId?, device?}. Installs the managed runtime and pulls the requested model. Whisper, Nemotron, and Voxtral load the requested weights to validate exact CUDA placement; VibeVoice stores its pinned model and tokenizer snapshot under Omnius' ASR runtime directories.",
           "parameters": [
             {
               "name": "engineId",
@@ -3610,6 +3941,9 @@
             "200": {
               "description": "Exact selection activated"
             },
+            "202": {
+              "description": "Model accepted; pending_download includes statusUrl, retryUrl, pollAfterMs, and the still-active prior selection"
+            },
             "400": {
               "description": "Unknown engine or model"
             },
@@ -15276,7 +15610,7 @@
       "id": "api.v1-voice-models",
       "kind": "api",
       "title": "/v1/voice/models",
-      "summary": "List TTS voice models with backend-specific metadata",
+      "summary": "List TTS voice models with backend metadata and managed readiness",
       "aliases": [
         "/v1/voice/models"
       ],
@@ -15326,13 +15660,191 @@
       ],
       "operations": {
         "get": {
-          "summary": "List TTS voice models with backend-specific metadata",
+          "summary": "List TTS voice models with backend metadata and managed readiness",
+          "tags": [
+            "Voice"
+          ],
+          "responses": {
+            "200": {
+              "description": "Voice model list, active selection, and Voxtral pull/deploy state"
+            }
+          }
+        }
+      },
+      "source_of_truth": [
+        "GET /openapi.json",
+        "packages/cli/src/api/openapi.ts"
+      ]
+    },
+    {
+      "id": "api.v1-voice-models-model-id-deploy",
+      "kind": "api",
+      "title": "/v1/voice/models/{modelId}/deploy",
+      "summary": "Pull and deploy one managed CUDA TTS model",
+      "aliases": [
+        "/v1/voice/models/{modelId}/deploy"
+      ],
+      "keywords": [
+        "rest",
+        "openapi",
+        "POST",
+        "Voice",
+        "v1",
+        "voice",
+        "models",
+        "{modelId}",
+        "deploy"
+      ],
+      "maturity": "stable",
+      "layer": "interface",
+      "audiences": [
+        "integrator",
+        "service-agent",
+        "coding-agent"
+      ],
+      "interfaces": [
+        {
+          "type": "rest",
+          "target": "POST /v1/voice/models/{modelId}/deploy"
+        },
+        {
+          "type": "openapi",
+          "target": "/openapi.json"
+        }
+      ],
+      "references": [
+        {
+          "type": "source",
+          "target": "packages/cli/src/api/openapi.ts",
+          "relation": "openapi-source"
+        },
+        {
+          "type": "documentation",
+          "target": "docs/reference/rest-api.md",
+          "relation": "endpoint-inventory"
+        }
+      ],
+      "methods": [
+        "POST"
+      ],
+      "tags": [
+        "Voice"
+      ],
+      "operations": {
+        "post": {
+          "summary": "Pull and deploy one managed CUDA TTS model",
+          "tags": [
+            "Voice"
+          ],
+          "description": "For voxtral-4b-tts-2603 this verifies the selected CUDA device and starts a persistent local vLLM-Omni server. Set OMNIUS_TTS_CUDA_DEVICE to choose the physical GPU.",
+          "parameters": [
+            {
+              "name": "modelId",
+              "in": "path",
+              "required": true,
+              "schema": {
+                "type": "string"
+              }
+            }
+          ],
+          "responses": {
+            "200": {
+              "description": "Runtime deployed and ready"
+            },
+            "404": {
+              "description": "No managed deploy adapter"
+            },
+            "500": {
+              "description": "Runtime installation, CUDA preflight, or startup failed"
+            }
+          }
+        }
+      },
+      "source_of_truth": [
+        "GET /openapi.json",
+        "packages/cli/src/api/openapi.ts"
+      ]
+    },
+    {
+      "id": "api.v1-voice-models-model-id-pull",
+      "kind": "api",
+      "title": "/v1/voice/models/{modelId}/pull",
+      "summary": "Install runtime prerequisites and pull one managed TTS model",
+      "aliases": [
+        "/v1/voice/models/{modelId}/pull"
+      ],
+      "keywords": [
+        "rest",
+        "openapi",
+        "POST",
+        "Voice",
+        "v1",
+        "voice",
+        "models",
+        "{modelId}",
+        "pull"
+      ],
+      "maturity": "stable",
+      "layer": "interface",
+      "audiences": [
+        "integrator",
+        "service-agent",
+        "coding-agent"
+      ],
+      "interfaces": [
+        {
+          "type": "rest",
+          "target": "POST /v1/voice/models/{modelId}/pull"
+        },
+        {
+          "type": "openapi",
+          "target": "/openapi.json"
+        }
+      ],
+      "references": [
+        {
+          "type": "source",
+          "target": "packages/cli/src/api/openapi.ts",
+          "relation": "openapi-source"
+        },
+        {
+          "type": "documentation",
+          "target": "docs/reference/rest-api.md",
+          "relation": "endpoint-inventory"
+        }
+      ],
+      "methods": [
+        "POST"
+      ],
+      "tags": [
+        "Voice"
+      ],
+      "operations": {
+        "post": {
+          "summary": "Install runtime prerequisites and pull one managed TTS model",
           "tags": [
             "Voice"
           ],
+          "description": "For voxtral-4b-tts-2603 this creates an Omnius-managed environment and snapshots mistralai/Voxtral-4B-TTS-2603. It does not require shell setup and does not start inference.",
+          "parameters": [
+            {
+              "name": "modelId",
+              "in": "path",
+              "required": true,
+              "schema": {
+                "type": "string"
+              }
+            }
+          ],
           "responses": {
             "200": {
-              "description": "Voice model list"
+              "description": "Weights ready"
+            },
+            "404": {
+              "description": "No managed pull adapter"
+            },
+            "500": {
+              "description": "Dependency installation or download failed"
             }
           }
         }
@@ -16017,7 +16529,7 @@
           "tags": [
             "Voice"
           ],
-          "description": "Returns synthesized audio and automatically enables/warms the daemon voice runtime when needed. Format `wav` (default) ships a complete WAV with header; `pcm` ships raw Int16 LE PCM. Optional `model` selects an exact one-shot renderer, including glados, overwatch, luxtts:announcer-testchamber03, and enabled Voicebox sub-model IDs. Model transactions are serialized, fail closed with no silent fallback, and restore the persistent model plus clone reference. Headers report the actual X-Voice-Model, X-Voice-Backend, and X-Sample-Rate.",
+          "description": "Returns synthesized audio and automatically enables/warms the selected runtime. Format `wav` (default) ships a complete WAV with header; `pcm` ships raw Int16 LE PCM. Set model=voxtral-4b-tts-2603 and voice=<preset> for the managed CUDA Voxtral renderer; the OpenAI-compatible alias also accepts model=mistralai/Voxtral-4B-TTS-2603. Other model transactions remain serialized and restore the persistent renderer. Headers report X-Voice-Model, X-Voice-Backend, and X-Sample-Rate.",
           "responses": {
             "200": {
               "description": "Audio bytes (audio/wav or audio/L16). Headers: X-Voice-Model, X-Sample-Rate."

package/docs/DISCOVERY.md

@@ -87,7 +87,10 @@ Daemon equivalents are `GET /v1/discovery/bootstrap`, `GET /v1/discovery?q=<inte
 | `api.v1-aiwg-skills-name` | /v1/aiwg/skills/{name} | Skill content |
 | `api.v1-aiwg-use` | /v1/aiwg/use | AIWG cascade activation bundle (aiwg use all equivalent, model-tier sized) |
 | `api.v1-asr-activate` | /v1/asr/activate | Activate and persist an exact ASR engine/model |
-| `api.v1-asr-engines` | /v1/asr/engines | List all ASR engines, models, capabilities, resource requirements, readiness, and active selection |
+| `api.v1-asr-downloads` | /v1/asr/downloads | List persistent ASR weight download and deployment transitions; Start or resume an ASR weight download |
+| `api.v1-asr-downloads-engine-id-model-id` | /v1/asr/downloads/{engineId}/{modelId} | Poll one ASR download and deployment transition; Start or resume this model download |
+| `api.v1-asr-downloads-engine-id-model-id-retry` | /v1/asr/downloads/{engineId}/{modelId}/retry | Retry or resume a failed/interrupted ASR weight download |
+| `api.v1-asr-engines` | /v1/asr/engines | List all ASR engines, models, capabilities, resource requirements, readiness, download transitions, and active selection |
 | `api.v1-asr-engines-engine-id-models-model-id-deploy` | /v1/asr/engines/{engineId}/models/{modelId}/deploy | Pull, select, and activate one exact ASR model |
 | `api.v1-asr-engines-engine-id-models-model-id-pull` | /v1/asr/engines/{engineId}/models/{modelId}/pull | Pull one exact ASR model and validate its managed runtime |
 | `api.v1-asr-engines-engine-id-setup` | /v1/asr/engines/{engineId}/setup | Install a managed ASR runtime and its pinned weights |
@@ -219,7 +222,9 @@ Daemon equivalents are `GET /v1/discovery/bootstrap`, `GET /v1/discovery?q=<inte
 | `api.v1-voice-clone-refs-filename-rename` | /v1/voice/clone-refs/{filename}/rename | Rename the friendly name (filename unchanged) — body {name} |
 | `api.v1-voice-clone-refs-from-url` | /v1/voice/clone-refs/from-url | Fetch a voice clone reference from a URL |
 | `api.v1-voice-clone-refs-upload` | /v1/voice/clone-refs/upload | Upload a voice clone reference (raw bytes) |
-| `api.v1-voice-models` | /v1/voice/models | List TTS voice models with backend-specific metadata |
+| `api.v1-voice-models` | /v1/voice/models | List TTS voice models with backend metadata and managed readiness |
+| `api.v1-voice-models-model-id-deploy` | /v1/voice/models/{modelId}/deploy | Pull and deploy one managed CUDA TTS model |
+| `api.v1-voice-models-model-id-pull` | /v1/voice/models/{modelId}/pull | Install runtime prerequisites and pull one managed TTS model |
 | `api.v1-voice-models-switch` | /v1/voice/models/switch | Switch and enable the active TTS model — body {modelId, enable?} |
 | `api.v1-voice-speak` | /v1/voice/speak | Synthesize text and broadcast to /v1/voicechat/ws clients (no audio in response body) |
 | `api.v1-voice-start` | /v1/voice/start | Enable and warm the daemon voice runtime; optional body {modelId\|model\|voice} |

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.623",
+  "version": "1.0.625",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.623",
+      "version": "1.0.625",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.623",
+  "version": "1.0.625",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/library.js",