@x12i/ai-profiles 1.7.0 → 1.8.0

package/dist/data/models-profiles.json

@@ -2,7 +2,7 @@
   "schema": "x12i.ai-profiles.model-registry.v3",
   "registryId": "models-profiles",
   "generatedAt": "2026-05-29T00:00:00.000Z",
-  "lastVerifiedAt": "2026-05-30T20:44:08Z",
+  "lastVerifiedAt": "2026-05-30T21:56:06Z",
   "version": "3.0.0",
   "currency": "USD",
   "pricingUnit": "1M_tokens",
@@ -13,7 +13,8 @@
     "openRouterModelId": "Resolved from models-catalog.json at runtime; never stored on profile choices.",
     "modelStatus": "verified = stable vendor SKU; predicted = forward-looking placeholder — not guaranteed callable.",
     "requiredCapabilities": "Profile intent contract; concrete model picks remain suggestions until Optimixer selects dynamically.",
-    "deploymentType": "How the model is operated (proprietary API vs hosted open weights vs self-hosted).",
+    "deploymentType": "How the model is operated (proprietary API vs self-hosted CPU). Remote profiles require OpenRouter catalog backing.",
+    "cpuLocalProfile": "Only the \"local\" profile runs on-device CPU (funcx / Transformers.js). No OpenRouter.",
     "inputModalities": "Required input media types for profile intent (text, image, audio, video, file).",
     "outputModalities": "Required output media types for profile intent (text, image, audio, video, file).",
     "choiceModalities": "Optional per-choice modality override when catalog linkage is missing."
@@ -253,7 +254,7 @@
               "openrouter:catalog"
             ]
           },
-          "modelStatus": "predicted"
+          "modelStatus": "verified"
         },
         "anthropic_fast": {
           "choice": "anthropic_fast",
@@ -418,7 +419,7 @@
               "openrouter:catalog"
             ]
           },
-          "modelStatus": "predicted"
+          "modelStatus": "verified"
         },
         "minimax_balanced": {
           "choice": "minimax_balanced",
@@ -539,7 +540,7 @@
               "openrouter:catalog"
             ]
           },
-          "modelStatus": "predicted"
+          "modelStatus": "verified"
         },
         "deepseek_pro": {
           "choice": "deepseek_pro",
@@ -789,7 +790,7 @@
               "openrouter:catalog"
             ]
           },
-          "modelStatus": "predicted"
+          "modelStatus": "verified"
         },
         "anthropic_json": {
           "choice": "anthropic_json",
@@ -910,7 +911,7 @@
               "openrouter:catalog"
             ]
           },
-          "modelStatus": "predicted"
+          "modelStatus": "verified"
         },
         "minimax_agentic": {
           "choice": "minimax_agentic",
@@ -1045,7 +1046,7 @@
               "openrouter:catalog"
             ]
           },
-          "modelStatus": "predicted"
+          "modelStatus": "verified"
         }
       },
       "fallbackChoiceOrder": [
@@ -1120,7 +1121,7 @@
               "openrouter:catalog"
             ]
           },
-          "modelStatus": "predicted"
+          "modelStatus": "verified"
         },
         "deep_forensics": {
           "choice": "deep_forensics",
@@ -1219,7 +1220,7 @@
               "openrouter:catalog"
             ]
           },
-          "modelStatus": "predicted"
+          "modelStatus": "verified"
         },
         "brand_voice": {
           "choice": "brand_voice",
@@ -1300,13 +1301,13 @@
           "displayName": "Mega Context Summarization",
           "description": "Compressing entire repositories, books, or multi-year document histories.",
           "provider": "google",
-          "modelId": "gemini-2.5-pro",
+          "modelId": "gemini-3.5-flash",
           "runtime": {
             "reasoningEffort": "high",
             "toolPolicy": "none",
             "outputMode": "text"
           },
-          "reason": "When input exceeds ~200k tokens, Pro-tier long context handles needle-in-haystack retrieval best.",
+          "reason": "When input exceeds ~200k tokens, Gemini 3.5 Flash long context handles needle-in-haystack retrieval best.",
           "verification": {
             "lastVerifiedAt": "2026-05-29T17:08:03Z",
             "sourceRefs": [
@@ -1407,98 +1408,10 @@
         ]
       }
     },
-    "open_weights": {
-      "profile": "open_weights",
-      "displayName": "Open Weights Ecosystem",
-      "description": "Hosted open-source models as a hedge against proprietary lock-in, with pathways to self-hosting.",
-      "category": "resilience",
-      "aliases": [
-        "llama",
-        "qwen",
-        "mistral",
-        "self-hostable"
-      ],
-      "defaultChoice": "default",
-      "choices": {
-        "default": {
-          "choice": "default",
-          "displayName": "Llama 3.3 70B",
-          "description": "Standard open-weights workhorse with a clear self-hosting path when compliance requires it.",
-          "provider": "together",
-          "modelId": "meta-llama/Llama-3.3-70B-Instruct-Turbo",
-          "runtime": {
-            "instructionTier": "default",
-            "backend": "openrouter",
-            "reasoningEffort": "medium",
-            "toolPolicy": "none",
-            "outputMode": "text"
-          },
-          "metadata": {
-            "openRouter": "vendor-direct"
-          },
-          "reason": "Direct pathway to self-hosting if privacy or vendor independence is required.",
-          "verification": {
-            "lastVerifiedAt": "2026-05-29T17:08:03Z",
-            "sourceRefs": [
-              "vendor:together",
-              "openrouter:catalog"
-            ]
-          },
-          "modelStatus": "verified",
-          "deploymentType": "openWeightsHosted"
-        },
-        "multilingual_code": {
-          "choice": "multilingual_code",
-          "displayName": "Qwen 2.5 72B",
-          "description": "Strong open-weights option for coding and non-English workloads.",
-          "provider": "together",
-          "modelId": "Qwen/Qwen3-235B-A22B-Instruct-2507-tput",
-          "runtime": {
-            "instructionTier": "default",
-            "backend": "openrouter",
-            "reasoningEffort": "high",
-            "toolPolicy": "none",
-            "outputMode": "text"
-          },
-          "metadata": {
-            "openRouter": "vendor-direct"
-          },
-          "reason": "Top-tier Chinese open-weights model; exceptionally strong at coding and multilingual tasks.",
-          "verification": {
-            "lastVerifiedAt": "2026-05-29T17:08:03Z",
-            "sourceRefs": [
-              "vendor:together",
-              "openrouter:catalog"
-            ]
-          },
-          "modelStatus": "verified",
-          "deploymentType": "openWeightsHosted"
-        }
-      },
-      "fallbackChoiceOrder": [
-        "default",
-        "multilingual_code"
-      ],
-      "requiredCapabilities": {
-        "reasoning": "medium",
-        "coding": true,
-        "cost": "medium",
-        "inputModalities": [
-          "text"
-        ],
-        "outputModalities": [
-          "text"
-        ]
-      },
-      "deploymentType": "openWeightsHosted",
-      "metadata": {
-        "deploymentStrategy": "openWeightsHosted"
-      }
-    },
     "local": {
       "profile": "local",
-      "displayName": "Local / CPU",
-      "description": "On-device inference via FuncX CPU backends. No OpenRouter. Uses cpu instruction tier.",
+      "displayName": "Local / CPU (no OpenRouter)",
+      "description": "On-device inference only — llama.cpp GGUF or Transformers.js on CPU. Does not call OpenRouter or any cloud LLM API.",
       "category": "local",
       "aliases": [
         "weak",
@@ -1506,56 +1419,66 @@
         "on-device"
       ],
       "defaultChoice": "llama_cpp_gguf",
+      "deploymentType": "selfHosted",
+      "metadata": {
+        "transport": "cpu-local",
+        "openRouter": "cpu-local",
+        "billingModel": "self_hosted"
+      },
       "choices": {
         "llama_cpp_gguf": {
           "choice": "llama_cpp_gguf",
-          "displayName": "llama.cpp (GGUF)",
-          "description": "Local GGUF via node-llama-cpp. Path from LLAMA_CPP_MODEL_PATH.",
+          "displayName": "llama.cpp GGUF (CPU, offline)",
+          "description": "Runs a local GGUF weights file via node-llama-cpp. Configure LLAMA_CPP_MODEL_PATH. No network LLM calls.",
           "provider": "local",
           "modelId": "funcx:llama-cpp:gguf",
           "runtime": {
             "instructionTier": "cpu",
+            "executionMode": "offline",
             "backend": "llama-cpp",
             "reasoningEffort": "minimal",
-            "executionMode": "offline",
             "temperature": 0.1,
             "toolPolicy": "none",
             "outputMode": "text"
           },
           "metadata": {
+            "openRouter": "cpu-local",
             "funcxBackend": "llama-cpp",
             "envModelPath": "LLAMA_CPP_MODEL_PATH",
             "envThreads": "LLAMA_CPP_THREADS",
             "envContextSize": "LLAMA_CPP_CONTEXT_SIZE",
             "defaultModelPath": "./models/model.gguf",
             "optionalDep": "node-llama-cpp",
-            "billingModel": "self_hosted"
+            "billingModel": "self_hosted",
+            "device": "cpu"
           },
           "verification": {
             "lastVerifiedAt": "2026-05-29T17:08:03Z",
             "sourceRefs": [
               "vendor:local",
-              "openrouter:catalog"
+              "cpu:self-hosted"
             ]
           },
-          "modelStatus": "verified"
+          "modelStatus": "verified",
+          "deploymentType": "selfHosted"
         },
         "transformersjs_default": {
           "choice": "transformersjs_default",
-          "displayName": "Transformers.js (CPU)",
-          "description": "Hugging Face model via @huggingface/transformers on CPU.",
+          "displayName": "Transformers.js (CPU, offline)",
+          "description": "Runs a Hugging Face model in-browser/on-node via @huggingface/transformers on CPU only. No OpenRouter.",
           "provider": "local",
           "modelId": "Xenova/distilbart-cnn-6-6",
           "runtime": {
             "instructionTier": "cpu",
+            "executionMode": "offline",
             "backend": "transformersjs",
             "reasoningEffort": "minimal",
-            "executionMode": "offline",
             "temperature": 0.1,
             "toolPolicy": "none",
             "outputMode": "text"
           },
           "metadata": {
+            "openRouter": "cpu-local",
             "funcxBackend": "transformersjs",
             "envModelId": "TRANSFORMERS_JS_MODEL_ID",
             "envCacheDir": "TRANSFORMERS_JS_CACHE_DIR",
@@ -1569,10 +1492,11 @@
             "lastVerifiedAt": "2026-05-29T17:08:03Z",
             "sourceRefs": [
               "vendor:local",
-              "openrouter:catalog"
+              "cpu:self-hosted"
             ]
           },
-          "modelStatus": "verified"
+          "modelStatus": "verified",
+          "deploymentType": "selfHosted"
         }
       },
       "fallbackChoiceOrder": [