apex-code 0.5.0 → 0.5.1

package/dist/core/model-config.d.ts

@@ -16,6 +16,19 @@ declare const ModelDefinitionSchema: Type.TObject<{
         max: Type.TOptional<Type.TUnion<[Type.TString, Type.TNull]>>;
     }>>;
     input: Type.TOptional<Type.TArray<Type.TUnion<[Type.TLiteral<"text">, Type.TLiteral<"image">]>>>;
+    inputLimits: Type.TOptional<Type.TObject<{
+        maxRequestBytes: Type.TOptional<Type.TInteger>;
+        images: Type.TOptional<Type.TObject<{
+            resize: Type.TOptional<Type.TObject<{
+                maxWidth: Type.TOptional<Type.TInteger>;
+                maxHeight: Type.TOptional<Type.TInteger>;
+                maxBytes: Type.TOptional<Type.TInteger>;
+                jpegQuality: Type.TOptional<Type.TInteger>;
+            }>>;
+            maxPerMessage: Type.TOptional<Type.TInteger>;
+            maxPerRequest: Type.TOptional<Type.TInteger>;
+        }>>;
+    }>>;
     cost: Type.TOptional<Type.TObject<{
         input: Type.TNumber;
         output: Type.TNumber;
@@ -29,6 +42,10 @@ declare const ModelDefinitionSchema: Type.TObject<{
             inputTokensAbove: Type.TNumber;
         }>>>;
     }>>;
+    promptCache: Type.TOptional<Type.TObject<{
+        short: Type.TOptional<Type.TNumber>;
+        long: Type.TOptional<Type.TNumber>;
+    }>>;
     contextWindow: Type.TOptional<Type.TNumber>;
     maxTokens: Type.TOptional<Type.TNumber>;
     samplingParams: Type.TOptional<Type.TRecord<"^.*$", Type.TUnknown>>;
@@ -95,17 +112,16 @@ declare const ModelDefinitionSchema: Type.TObject<{
         supportsOpenAIGrammarTools: Type.TOptional<Type.TBoolean>;
         supportsStrictMode: Type.TOptional<Type.TBoolean>;
         sendSessionAffinityHeaders: Type.TOptional<Type.TBoolean>;
-        deferredToolsMode: Type.TOptional<Type.TLiteral<"kimi">>;
         sessionAffinityFormat: Type.TOptional<Type.TUnion<[Type.TLiteral<"openai">, Type.TLiteral<"openai-nosession">, Type.TLiteral<"openrouter">]>>;
         supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
+        vllmPriority: Type.TOptional<Type.TNumber>;
     }>, Type.TObject<{
         supportsDeveloperRole: Type.TOptional<Type.TBoolean>;
         sessionAffinityFormat: Type.TOptional<Type.TUnion<[Type.TLiteral<"openai">, Type.TLiteral<"openai-nosession">, Type.TLiteral<"openrouter">]>>;
         supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
         supportsStrictMode: Type.TOptional<Type.TBoolean>;
         supportsOpenAIGrammarTools: Type.TOptional<Type.TBoolean>;
-        supportsAdditionalTools: Type.TOptional<Type.TBoolean>;
-        supportsToolSearch: Type.TOptional<Type.TBoolean>;
+        supportsMaxOutputTokens: Type.TOptional<Type.TBoolean>;
     }>, Type.TObject<{
         supportsEagerToolInputStreaming: Type.TOptional<Type.TBoolean>;
         supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
@@ -115,7 +131,24 @@ declare const ModelDefinitionSchema: Type.TObject<{
         forceAdaptiveThinking: Type.TOptional<Type.TBoolean>;
         allowEmptySignature: Type.TOptional<Type.TBoolean>;
         supportsStrictTools: Type.TOptional<Type.TBoolean>;
-        supportsToolReferences: Type.TOptional<Type.TBoolean>;
+        supportsMidConvoEffort: Type.TOptional<Type.TBoolean>;
+        allowedFallbackModels: Type.TOptional<Type.TArray<Type.TObject<{
+            provider: Type.TString;
+            model: Type.TString;
+            cost: Type.TObject<{
+                input: Type.TNumber;
+                output: Type.TNumber;
+                cacheRead: Type.TNumber;
+                cacheWrite: Type.TNumber;
+                tiers: Type.TOptional<Type.TArray<Type.TObject<{
+                    input: Type.TNumber;
+                    output: Type.TNumber;
+                    cacheRead: Type.TNumber;
+                    cacheWrite: Type.TNumber;
+                    inputTokensAbove: Type.TNumber;
+                }>>>;
+            }>;
+        }>>>;
     }>]>>;
 }>;
 declare const ModelOverrideSchema: Type.TObject<{
@@ -131,6 +164,19 @@ declare const ModelOverrideSchema: Type.TObject<{
         max: Type.TOptional<Type.TUnion<[Type.TString, Type.TNull]>>;
     }>>;
     input: Type.TOptional<Type.TArray<Type.TUnion<[Type.TLiteral<"text">, Type.TLiteral<"image">]>>>;
+    inputLimits: Type.TOptional<Type.TObject<{
+        maxRequestBytes: Type.TOptional<Type.TInteger>;
+        images: Type.TOptional<Type.TObject<{
+            resize: Type.TOptional<Type.TObject<{
+                maxWidth: Type.TOptional<Type.TInteger>;
+                maxHeight: Type.TOptional<Type.TInteger>;
+                maxBytes: Type.TOptional<Type.TInteger>;
+                jpegQuality: Type.TOptional<Type.TInteger>;
+            }>>;
+            maxPerMessage: Type.TOptional<Type.TInteger>;
+            maxPerRequest: Type.TOptional<Type.TInteger>;
+        }>>;
+    }>>;
     cost: Type.TOptional<Type.TObject<{
         input: Type.TOptional<Type.TNumber>;
         output: Type.TOptional<Type.TNumber>;
@@ -144,6 +190,10 @@ declare const ModelOverrideSchema: Type.TObject<{
             inputTokensAbove: Type.TNumber;
         }>>>;
     }>>;
+    promptCache: Type.TOptional<Type.TObject<{
+        short: Type.TOptional<Type.TNumber>;
+        long: Type.TOptional<Type.TNumber>;
+    }>>;
     contextWindow: Type.TOptional<Type.TNumber>;
     maxTokens: Type.TOptional<Type.TNumber>;
     samplingParams: Type.TOptional<Type.TRecord<"^.*$", Type.TUnknown>>;
@@ -210,17 +260,16 @@ declare const ModelOverrideSchema: Type.TObject<{
         supportsOpenAIGrammarTools: Type.TOptional<Type.TBoolean>;
         supportsStrictMode: Type.TOptional<Type.TBoolean>;
         sendSessionAffinityHeaders: Type.TOptional<Type.TBoolean>;
-        deferredToolsMode: Type.TOptional<Type.TLiteral<"kimi">>;
         sessionAffinityFormat: Type.TOptional<Type.TUnion<[Type.TLiteral<"openai">, Type.TLiteral<"openai-nosession">, Type.TLiteral<"openrouter">]>>;
         supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
+        vllmPriority: Type.TOptional<Type.TNumber>;
     }>, Type.TObject<{
         supportsDeveloperRole: Type.TOptional<Type.TBoolean>;
         sessionAffinityFormat: Type.TOptional<Type.TUnion<[Type.TLiteral<"openai">, Type.TLiteral<"openai-nosession">, Type.TLiteral<"openrouter">]>>;
         supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
         supportsStrictMode: Type.TOptional<Type.TBoolean>;
         supportsOpenAIGrammarTools: Type.TOptional<Type.TBoolean>;
-        supportsAdditionalTools: Type.TOptional<Type.TBoolean>;
-        supportsToolSearch: Type.TOptional<Type.TBoolean>;
+        supportsMaxOutputTokens: Type.TOptional<Type.TBoolean>;
     }>, Type.TObject<{
         supportsEagerToolInputStreaming: Type.TOptional<Type.TBoolean>;
         supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
@@ -230,7 +279,24 @@ declare const ModelOverrideSchema: Type.TObject<{
         forceAdaptiveThinking: Type.TOptional<Type.TBoolean>;
         allowEmptySignature: Type.TOptional<Type.TBoolean>;
         supportsStrictTools: Type.TOptional<Type.TBoolean>;
-        supportsToolReferences: Type.TOptional<Type.TBoolean>;
+        supportsMidConvoEffort: Type.TOptional<Type.TBoolean>;
+        allowedFallbackModels: Type.TOptional<Type.TArray<Type.TObject<{
+            provider: Type.TString;
+            model: Type.TString;
+            cost: Type.TObject<{
+                input: Type.TNumber;
+                output: Type.TNumber;
+                cacheRead: Type.TNumber;
+                cacheWrite: Type.TNumber;
+                tiers: Type.TOptional<Type.TArray<Type.TObject<{
+                    input: Type.TNumber;
+                    output: Type.TNumber;
+                    cacheRead: Type.TNumber;
+                    cacheWrite: Type.TNumber;
+                    inputTokensAbove: Type.TNumber;
+                }>>>;
+            }>;
+        }>>>;
     }>]>>;
 }>;
 declare const ProviderConfigSchema: Type.TObject<{
@@ -302,17 +368,16 @@ declare const ProviderConfigSchema: Type.TObject<{
         supportsOpenAIGrammarTools: Type.TOptional<Type.TBoolean>;
         supportsStrictMode: Type.TOptional<Type.TBoolean>;
         sendSessionAffinityHeaders: Type.TOptional<Type.TBoolean>;
-        deferredToolsMode: Type.TOptional<Type.TLiteral<"kimi">>;
         sessionAffinityFormat: Type.TOptional<Type.TUnion<[Type.TLiteral<"openai">, Type.TLiteral<"openai-nosession">, Type.TLiteral<"openrouter">]>>;
         supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
+        vllmPriority: Type.TOptional<Type.TNumber>;
     }>, Type.TObject<{
         supportsDeveloperRole: Type.TOptional<Type.TBoolean>;
         sessionAffinityFormat: Type.TOptional<Type.TUnion<[Type.TLiteral<"openai">, Type.TLiteral<"openai-nosession">, Type.TLiteral<"openrouter">]>>;
         supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
         supportsStrictMode: Type.TOptional<Type.TBoolean>;
         supportsOpenAIGrammarTools: Type.TOptional<Type.TBoolean>;
-        supportsAdditionalTools: Type.TOptional<Type.TBoolean>;
-        supportsToolSearch: Type.TOptional<Type.TBoolean>;
+        supportsMaxOutputTokens: Type.TOptional<Type.TBoolean>;
     }>, Type.TObject<{
         supportsEagerToolInputStreaming: Type.TOptional<Type.TBoolean>;
         supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
@@ -322,7 +387,24 @@ declare const ProviderConfigSchema: Type.TObject<{
         forceAdaptiveThinking: Type.TOptional<Type.TBoolean>;
         allowEmptySignature: Type.TOptional<Type.TBoolean>;
         supportsStrictTools: Type.TOptional<Type.TBoolean>;
-        supportsToolReferences: Type.TOptional<Type.TBoolean>;
+        supportsMidConvoEffort: Type.TOptional<Type.TBoolean>;
+        allowedFallbackModels: Type.TOptional<Type.TArray<Type.TObject<{
+            provider: Type.TString;
+            model: Type.TString;
+            cost: Type.TObject<{
+                input: Type.TNumber;
+                output: Type.TNumber;
+                cacheRead: Type.TNumber;
+                cacheWrite: Type.TNumber;
+                tiers: Type.TOptional<Type.TArray<Type.TObject<{
+                    input: Type.TNumber;
+                    output: Type.TNumber;
+                    cacheRead: Type.TNumber;
+                    cacheWrite: Type.TNumber;
+                    inputTokensAbove: Type.TNumber;
+                }>>>;
+            }>;
+        }>>>;
     }>]>>;
     authHeader: Type.TOptional<Type.TBoolean>;
     models: Type.TOptional<Type.TArray<Type.TObject<{
@@ -341,6 +423,19 @@ declare const ProviderConfigSchema: Type.TObject<{
             max: Type.TOptional<Type.TUnion<[Type.TString, Type.TNull]>>;
         }>>;
         input: Type.TOptional<Type.TArray<Type.TUnion<[Type.TLiteral<"text">, Type.TLiteral<"image">]>>>;
+        inputLimits: Type.TOptional<Type.TObject<{
+            maxRequestBytes: Type.TOptional<Type.TInteger>;
+            images: Type.TOptional<Type.TObject<{
+                resize: Type.TOptional<Type.TObject<{
+                    maxWidth: Type.TOptional<Type.TInteger>;
+                    maxHeight: Type.TOptional<Type.TInteger>;
+                    maxBytes: Type.TOptional<Type.TInteger>;
+                    jpegQuality: Type.TOptional<Type.TInteger>;
+                }>>;
+                maxPerMessage: Type.TOptional<Type.TInteger>;
+                maxPerRequest: Type.TOptional<Type.TInteger>;
+            }>>;
+        }>>;
         cost: Type.TOptional<Type.TObject<{
             input: Type.TNumber;
             output: Type.TNumber;
@@ -354,6 +449,10 @@ declare const ProviderConfigSchema: Type.TObject<{
                 inputTokensAbove: Type.TNumber;
             }>>>;
         }>>;
+        promptCache: Type.TOptional<Type.TObject<{
+            short: Type.TOptional<Type.TNumber>;
+            long: Type.TOptional<Type.TNumber>;
+        }>>;
         contextWindow: Type.TOptional<Type.TNumber>;
         maxTokens: Type.TOptional<Type.TNumber>;
         samplingParams: Type.TOptional<Type.TRecord<"^.*$", Type.TUnknown>>;
@@ -420,17 +519,16 @@ declare const ProviderConfigSchema: Type.TObject<{
             supportsOpenAIGrammarTools: Type.TOptional<Type.TBoolean>;
             supportsStrictMode: Type.TOptional<Type.TBoolean>;
             sendSessionAffinityHeaders: Type.TOptional<Type.TBoolean>;
-            deferredToolsMode: Type.TOptional<Type.TLiteral<"kimi">>;
             sessionAffinityFormat: Type.TOptional<Type.TUnion<[Type.TLiteral<"openai">, Type.TLiteral<"openai-nosession">, Type.TLiteral<"openrouter">]>>;
             supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
+            vllmPriority: Type.TOptional<Type.TNumber>;
         }>, Type.TObject<{
             supportsDeveloperRole: Type.TOptional<Type.TBoolean>;
             sessionAffinityFormat: Type.TOptional<Type.TUnion<[Type.TLiteral<"openai">, Type.TLiteral<"openai-nosession">, Type.TLiteral<"openrouter">]>>;
             supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
             supportsStrictMode: Type.TOptional<Type.TBoolean>;
             supportsOpenAIGrammarTools: Type.TOptional<Type.TBoolean>;
-            supportsAdditionalTools: Type.TOptional<Type.TBoolean>;
-            supportsToolSearch: Type.TOptional<Type.TBoolean>;
+            supportsMaxOutputTokens: Type.TOptional<Type.TBoolean>;
         }>, Type.TObject<{
             supportsEagerToolInputStreaming: Type.TOptional<Type.TBoolean>;
             supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
@@ -440,7 +538,24 @@ declare const ProviderConfigSchema: Type.TObject<{
             forceAdaptiveThinking: Type.TOptional<Type.TBoolean>;
             allowEmptySignature: Type.TOptional<Type.TBoolean>;
             supportsStrictTools: Type.TOptional<Type.TBoolean>;
-            supportsToolReferences: Type.TOptional<Type.TBoolean>;
+            supportsMidConvoEffort: Type.TOptional<Type.TBoolean>;
+            allowedFallbackModels: Type.TOptional<Type.TArray<Type.TObject<{
+                provider: Type.TString;
+                model: Type.TString;
+                cost: Type.TObject<{
+                    input: Type.TNumber;
+                    output: Type.TNumber;
+                    cacheRead: Type.TNumber;
+                    cacheWrite: Type.TNumber;
+                    tiers: Type.TOptional<Type.TArray<Type.TObject<{
+                        input: Type.TNumber;
+                        output: Type.TNumber;
+                        cacheRead: Type.TNumber;
+                        cacheWrite: Type.TNumber;
+                        inputTokensAbove: Type.TNumber;
+                    }>>>;
+                }>;
+            }>>>;
         }>]>>;
     }>>>;
     modelOverrides: Type.TOptional<Type.TRecord<"^.*$", Type.TObject<{
@@ -456,6 +571,19 @@ declare const ProviderConfigSchema: Type.TObject<{
             max: Type.TOptional<Type.TUnion<[Type.TString, Type.TNull]>>;
         }>>;
         input: Type.TOptional<Type.TArray<Type.TUnion<[Type.TLiteral<"text">, Type.TLiteral<"image">]>>>;
+        inputLimits: Type.TOptional<Type.TObject<{
+            maxRequestBytes: Type.TOptional<Type.TInteger>;
+            images: Type.TOptional<Type.TObject<{
+                resize: Type.TOptional<Type.TObject<{
+                    maxWidth: Type.TOptional<Type.TInteger>;
+                    maxHeight: Type.TOptional<Type.TInteger>;
+                    maxBytes: Type.TOptional<Type.TInteger>;
+                    jpegQuality: Type.TOptional<Type.TInteger>;
+                }>>;
+                maxPerMessage: Type.TOptional<Type.TInteger>;
+                maxPerRequest: Type.TOptional<Type.TInteger>;
+            }>>;
+        }>>;
         cost: Type.TOptional<Type.TObject<{
             input: Type.TOptional<Type.TNumber>;
             output: Type.TOptional<Type.TNumber>;
@@ -469,6 +597,10 @@ declare const ProviderConfigSchema: Type.TObject<{
                 inputTokensAbove: Type.TNumber;
             }>>>;
         }>>;
+        promptCache: Type.TOptional<Type.TObject<{
+            short: Type.TOptional<Type.TNumber>;
+            long: Type.TOptional<Type.TNumber>;
+        }>>;
         contextWindow: Type.TOptional<Type.TNumber>;
         maxTokens: Type.TOptional<Type.TNumber>;
         samplingParams: Type.TOptional<Type.TRecord<"^.*$", Type.TUnknown>>;
@@ -535,17 +667,16 @@ declare const ProviderConfigSchema: Type.TObject<{
             supportsOpenAIGrammarTools: Type.TOptional<Type.TBoolean>;
             supportsStrictMode: Type.TOptional<Type.TBoolean>;
             sendSessionAffinityHeaders: Type.TOptional<Type.TBoolean>;
-            deferredToolsMode: Type.TOptional<Type.TLiteral<"kimi">>;
             sessionAffinityFormat: Type.TOptional<Type.TUnion<[Type.TLiteral<"openai">, Type.TLiteral<"openai-nosession">, Type.TLiteral<"openrouter">]>>;
             supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
+            vllmPriority: Type.TOptional<Type.TNumber>;
         }>, Type.TObject<{
             supportsDeveloperRole: Type.TOptional<Type.TBoolean>;
             sessionAffinityFormat: Type.TOptional<Type.TUnion<[Type.TLiteral<"openai">, Type.TLiteral<"openai-nosession">, Type.TLiteral<"openrouter">]>>;
             supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
             supportsStrictMode: Type.TOptional<Type.TBoolean>;
             supportsOpenAIGrammarTools: Type.TOptional<Type.TBoolean>;
-            supportsAdditionalTools: Type.TOptional<Type.TBoolean>;
-            supportsToolSearch: Type.TOptional<Type.TBoolean>;
+            supportsMaxOutputTokens: Type.TOptional<Type.TBoolean>;
         }>, Type.TObject<{
             supportsEagerToolInputStreaming: Type.TOptional<Type.TBoolean>;
             supportsLongCacheRetention: Type.TOptional<Type.TBoolean>;
@@ -555,7 +686,24 @@ declare const ProviderConfigSchema: Type.TObject<{
             forceAdaptiveThinking: Type.TOptional<Type.TBoolean>;
             allowEmptySignature: Type.TOptional<Type.TBoolean>;
             supportsStrictTools: Type.TOptional<Type.TBoolean>;
-            supportsToolReferences: Type.TOptional<Type.TBoolean>;
+            supportsMidConvoEffort: Type.TOptional<Type.TBoolean>;
+            allowedFallbackModels: Type.TOptional<Type.TArray<Type.TObject<{
+                provider: Type.TString;
+                model: Type.TString;
+                cost: Type.TObject<{
+                    input: Type.TNumber;
+                    output: Type.TNumber;
+                    cacheRead: Type.TNumber;
+                    cacheWrite: Type.TNumber;
+                    tiers: Type.TOptional<Type.TArray<Type.TObject<{
+                        input: Type.TNumber;
+                        output: Type.TNumber;
+                        cacheRead: Type.TNumber;
+                        cacheWrite: Type.TNumber;
+                        inputTokensAbove: Type.TNumber;
+                    }>>>;
+                }>;
+            }>>>;
         }>]>>;
     }>>>;
 }>;

package/dist/core/model-config.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"model-config.d.ts","sourceRoot":"","sources":["../../src/core/model-config.ts"],"names":[],"mappings":"AAAA,wDAAwD;AAGxD,OAAO,EAAE,KAAK,MAAM,EAAE,IAAI,EAAE,MAAM,SAAS,CAAC;AA2J5C,QAAA,MAAM,qBAAqB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAczB,CAAC;AAEH,QAAA,MAAM,mBAAmB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAmBvB,CAAC;AAEH,QAAA,MAAM,oBAAoB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAWxB,CAAC;AAEH,QAAA,MAAM,oBAAoB;IACzB,0EAA0E;;EAEzE,CAAC;AASH,MAAM,MAAM,eAAe,GAAG,MAAM,CAAC,OAAO,qBAAqB,CAAC,CAAC;AACnE,MAAM,MAAM,uBAAuB,GAAG,MAAM,CAAC,OAAO,mBAAmB,CAAC,CAAC;AACzE,MAAM,MAAM,kBAAkB,GAAG,MAAM,CAAC,OAAO,oBAAoB,CAAC,CAAC;AACrE,MAAM,MAAM,wBAAwB,GAAG,MAAM,CAAC,OAAO,oBAAoB,CAAC,CAAC;AAsB3E,yCAAyC;AACzC,qBAAa,WAAW;IACvB,OAAO,CAAC,QAAQ,CAAC,SAAS,CAA0C;IACpE,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAyC;IAC/D,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAqB;IAE3C,OAAO,eAQN;IAED,OAAa,IAAI,CAAC,cAAc,EAAE,MAAM,GAAG,SAAS,GAAG,OAAO,CAAC,WAAW,CAAC,CA6C1E;IAED,WAAW,CAAC,UAAU,EAAE,MAAM,GAAG,kBAAkB,GAAG,SAAS,CAE9D;IAED,cAAc,IAAI,SAAS,MAAM,EAAE,CAElC;IAED,oFAAoF;IACpF,QAAQ,IAAI,WAAW,CAAC,MAAM,EAAE,SAAS,MAAM,EAAE,CAAC,CAEjD;IAED,QAAQ,IAAI,MAAM,GAAG,SAAS,CAE7B;CACD","sourcesContent":["/** Immutable, credential-blind models.json snapshot. */\n\nimport { readFile } from \"node:fs/promises\";\nimport { type Static, Type } from \"typebox\";\nimport { Compile } from \"typebox/compile\";\nimport type { TLocalizedValidationError } from \"typebox/error\";\nimport { stripJsonComments } from \"../utils/json.ts\";\nimport { normalizePath } from \"../utils/paths.ts\";\nimport { stripBom } from \"../utils/text.ts\";\n\nconst PercentileCutoffsSchema = Type.Object({\n\tp50: Type.Optional(Type.Number()),\n\tp75: Type.Optional(Type.Number()),\n\tp90: Type.Optional(Type.Number()),\n\tp99: Type.Optional(Type.Number()),\n});\n\nconst OpenRouterRoutingSchema = Type.Object({\n\tallow_fallbacks: Type.Optional(Type.Boolean()),\n\trequire_parameters: Type.Optional(Type.Boolean()),\n\tdata_collection: Type.Optional(Type.Union([Type.Literal(\"deny\"), Type.Literal(\"allow\")])),\n\tzdr: Type.Optional(Type.Boolean()),\n\tenforce_distillable_text: Type.Optional(Type.Boolean()),\n\torder: Type.Optional(Type.Array(Type.String())),\n\tonly: Type.Optional(Type.Array(Type.String())),\n\tignore: Type.Optional(Type.Array(Type.String())),\n\tquantizations: Type.Optional(Type.Array(Type.String())),\n\tsort: Type.Optional(\n\t\tType.Union([\n\t\t\tType.String(),\n\t\t\tType.Object({\n\t\t\t\tby: Type.Optional(Type.String()),\n\t\t\t\tpartition: Type.Optional(Type.Union([Type.String(), Type.Null()])),\n\t\t\t}),\n\t\t]),\n\t),\n\tmax_price: Type.Optional(\n\t\tType.Object({\n\t\t\tprompt: Type.Optional(Type.Union([Type.Number(), Type.String()])),\n\t\t\tcompletion: Type.Optional(Type.Union([Type.Number(), Type.String()])),\n\t\t\timage: Type.Optional(Type.Union([Type.Number(), Type.String()])),\n\t\t\taudio: Type.Optional(Type.Union([Type.Number(), Type.String()])),\n\t\t\trequest: Type.Optional(Type.Union([Type.Number(), Type.String()])),\n\t\t}),\n\t),\n\tpreferred_min_throughput: Type.Optional(Type.Union([Type.Number(), PercentileCutoffsSchema])),\n\tpreferred_max_latency: Type.Optional(Type.Union([Type.Number(), PercentileCutoffsSchema])),\n});\n\nconst VercelGatewayRoutingSchema = Type.Object({\n\tonly: Type.Optional(Type.Array(Type.String())),\n\torder: Type.Optional(Type.Array(Type.String())),\n});\n\nconst ThinkingLevelMapValueSchema = Type.Union([Type.String(), Type.Null()]);\nconst ThinkingLevelMapSchema = Type.Object({\n\toff: Type.Optional(ThinkingLevelMapValueSchema),\n\tminimal: Type.Optional(ThinkingLevelMapValueSchema),\n\tlow: Type.Optional(ThinkingLevelMapValueSchema),\n\tmedium: Type.Optional(ThinkingLevelMapValueSchema),\n\thigh: Type.Optional(ThinkingLevelMapValueSchema),\n\txhigh: Type.Optional(ThinkingLevelMapValueSchema),\n\tmax: Type.Optional(ThinkingLevelMapValueSchema),\n});\n\nconst ChatTemplateKwargScalarSchema = Type.Union([Type.String(), Type.Number(), Type.Boolean(), Type.Null()]);\nconst ChatTemplateKwargVariableSchema = Type.Object({\n\t$var: Type.Union([Type.Literal(\"thinking.enabled\"), Type.Literal(\"thinking.effort\")]),\n\tomitWhenOff: Type.Optional(Type.Boolean()),\n});\nconst ChatTemplateKwargSchema = Type.Union([ChatTemplateKwargScalarSchema, ChatTemplateKwargVariableSchema]);\n\nconst OpenAICompletionsCompatSchema = Type.Object({\n\tsupportsStore: Type.Optional(Type.Boolean()),\n\tsupportsDeveloperRole: Type.Optional(Type.Boolean()),\n\tsupportsReasoningEffort: Type.Optional(Type.Boolean()),\n\tsupportsUsageInStreaming: Type.Optional(Type.Boolean()),\n\tsupportsFinishReason: Type.Optional(Type.Boolean()),\n\tmaxTokensField: Type.Optional(Type.Union([Type.Literal(\"max_completion_tokens\"), Type.Literal(\"max_tokens\")])),\n\trequiresToolResultName: Type.Optional(Type.Boolean()),\n\trequiresAssistantAfterToolResult: Type.Optional(Type.Boolean()),\n\trequiresThinkingAsText: Type.Optional(Type.Boolean()),\n\trequiresReasoningContentOnAssistantMessages: Type.Optional(Type.Boolean()),\n\tthinkingFormat: Type.Optional(\n\t\tType.Union([\n\t\t\tType.Literal(\"openai\"),\n\t\t\tType.Literal(\"openrouter\"),\n\t\t\tType.Literal(\"together\"),\n\t\t\tType.Literal(\"baseten\"),\n\t\t\tType.Literal(\"deepseek\"),\n\t\t\tType.Literal(\"zai\"),\n\t\t\tType.Literal(\"qwen\"),\n\t\t\tType.Literal(\"chat-template\"),\n\t\t\tType.Literal(\"qwen-chat-template\"),\n\t\t\tType.Literal(\"string-thinking\"),\n\t\t\tType.Literal(\"ant-ling\"),\n\t\t]),\n\t),\n\tchatTemplateKwargs: Type.Optional(Type.Record(Type.String(), ChatTemplateKwargSchema)),\n\tchatTemplateArgs: Type.Optional(Type.Record(Type.String(), ChatTemplateKwargSchema)),\n\tcacheControlFormat: Type.Optional(Type.Literal(\"anthropic\")),\n\topenRouterRouting: Type.Optional(OpenRouterRoutingSchema),\n\tvercelGatewayRouting: Type.Optional(VercelGatewayRoutingSchema),\n\tsupportsOpenAIGrammarTools: Type.Optional(Type.Boolean()),\n\tsupportsStrictMode: Type.Optional(Type.Boolean()),\n\tsendSessionAffinityHeaders: Type.Optional(Type.Boolean()),\n\tdeferredToolsMode: Type.Optional(Type.Literal(\"kimi\")),\n\tsessionAffinityFormat: Type.Optional(\n\t\tType.Union([Type.Literal(\"openai\"), Type.Literal(\"openai-nosession\"), Type.Literal(\"openrouter\")]),\n\t),\n\tsupportsLongCacheRetention: Type.Optional(Type.Boolean()),\n});\n\nconst OpenAIResponsesCompatSchema = Type.Object({\n\tsupportsDeveloperRole: Type.Optional(Type.Boolean()),\n\tsessionAffinityFormat: Type.Optional(\n\t\tType.Union([Type.Literal(\"openai\"), Type.Literal(\"openai-nosession\"), Type.Literal(\"openrouter\")]),\n\t),\n\tsupportsLongCacheRetention: Type.Optional(Type.Boolean()),\n\tsupportsStrictMode: Type.Optional(Type.Boolean()),\n\tsupportsOpenAIGrammarTools: Type.Optional(Type.Boolean()),\n\tsupportsAdditionalTools: Type.Optional(Type.Boolean()),\n\tsupportsToolSearch: Type.Optional(Type.Boolean()),\n});\n\nconst AnthropicMessagesCompatSchema = Type.Object({\n\tsupportsEagerToolInputStreaming: Type.Optional(Type.Boolean()),\n\tsupportsLongCacheRetention: Type.Optional(Type.Boolean()),\n\tsendSessionAffinityHeaders: Type.Optional(Type.Boolean()),\n\tsupportsCacheControlOnTools: Type.Optional(Type.Boolean()),\n\tsupportsTemperature: Type.Optional(Type.Boolean()),\n\tforceAdaptiveThinking: Type.Optional(Type.Boolean()),\n\tallowEmptySignature: Type.Optional(Type.Boolean()),\n\tsupportsStrictTools: Type.Optional(Type.Boolean()),\n\tsupportsToolReferences: Type.Optional(Type.Boolean()),\n});\n\nconst ProviderCompatSchema = Type.Union([\n\tOpenAICompletionsCompatSchema,\n\tOpenAIResponsesCompatSchema,\n\tAnthropicMessagesCompatSchema,\n]);\n\nconst ModelCostRatesSchema = {\n\tinput: Type.Number(),\n\toutput: Type.Number(),\n\tcacheRead: Type.Number(),\n\tcacheWrite: Type.Number(),\n};\nconst ModelCostTierSchema = Type.Object({\n\tinputTokensAbove: Type.Number(),\n\t...ModelCostRatesSchema,\n});\nconst ModelCostSchema = Type.Object({\n\t...ModelCostRatesSchema,\n\ttiers: Type.Optional(Type.Array(ModelCostTierSchema)),\n});\n\nconst ModelDefinitionSchema = Type.Object({\n\tid: Type.String({ minLength: 1 }),\n\tname: Type.Optional(Type.String({ minLength: 1 })),\n\tapi: Type.Optional(Type.String({ minLength: 1 })),\n\tbaseUrl: Type.Optional(Type.String({ minLength: 1 })),\n\treasoning: Type.Optional(Type.Boolean()),\n\tthinkingLevelMap: Type.Optional(ThinkingLevelMapSchema),\n\tinput: Type.Optional(Type.Array(Type.Union([Type.Literal(\"text\"), Type.Literal(\"image\")]))),\n\tcost: Type.Optional(ModelCostSchema),\n\tcontextWindow: Type.Optional(Type.Number()),\n\tmaxTokens: Type.Optional(Type.Number()),\n\tsamplingParams: Type.Optional(Type.Record(Type.String(), Type.Unknown())),\n\theaders: Type.Optional(Type.Record(Type.String(), Type.String())),\n\tcompat: Type.Optional(ProviderCompatSchema),\n});\n\nconst ModelOverrideSchema = Type.Object({\n\tname: Type.Optional(Type.String({ minLength: 1 })),\n\treasoning: Type.Optional(Type.Boolean()),\n\tthinkingLevelMap: Type.Optional(ThinkingLevelMapSchema),\n\tinput: Type.Optional(Type.Array(Type.Union([Type.Literal(\"text\"), Type.Literal(\"image\")]))),\n\tcost: Type.Optional(\n\t\tType.Object({\n\t\t\tinput: Type.Optional(Type.Number()),\n\t\t\toutput: Type.Optional(Type.Number()),\n\t\t\tcacheRead: Type.Optional(Type.Number()),\n\t\t\tcacheWrite: Type.Optional(Type.Number()),\n\t\t\ttiers: Type.Optional(Type.Array(ModelCostTierSchema)),\n\t\t}),\n\t),\n\tcontextWindow: Type.Optional(Type.Number()),\n\tmaxTokens: Type.Optional(Type.Number()),\n\tsamplingParams: Type.Optional(Type.Record(Type.String(), Type.Unknown())),\n\theaders: Type.Optional(Type.Record(Type.String(), Type.String())),\n\tcompat: Type.Optional(ProviderCompatSchema),\n});\n\nconst ProviderConfigSchema = Type.Object({\n\tname: Type.Optional(Type.String({ minLength: 1 })),\n\tbaseUrl: Type.Optional(Type.String({ minLength: 1 })),\n\tapiKey: Type.Optional(Type.String({ minLength: 1 })),\n\tapi: Type.Optional(Type.String({ minLength: 1 })),\n\toauth: Type.Optional(Type.Literal(\"radius\")),\n\theaders: Type.Optional(Type.Record(Type.String(), Type.String())),\n\tcompat: Type.Optional(ProviderCompatSchema),\n\tauthHeader: Type.Optional(Type.Boolean()),\n\tmodels: Type.Optional(Type.Array(ModelDefinitionSchema)),\n\tmodelOverrides: Type.Optional(Type.Record(Type.String(), ModelOverrideSchema)),\n});\n\nconst RoleDefinitionSchema = Type.Object({\n\t/** Ordered, non-empty list of canonical \"provider/modelId\" candidates. */\n\tmodels: Type.Array(Type.String({ minLength: 1 }), { minItems: 1 }),\n});\n\nconst ModelsConfigSchema = Type.Object({\n\tproviders: Type.Record(Type.String(), ProviderConfigSchema),\n\t/** Additive, credential-blind role -> ordered model-candidate mapping. */\n\troles: Type.Optional(Type.Record(Type.String(), RoleDefinitionSchema)),\n});\nconst validateModelsConfig = Compile(ModelsConfigSchema);\n\nexport type ModelsJsonModel = Static<typeof ModelDefinitionSchema>;\nexport type ModelsJsonModelOverride = Static<typeof ModelOverrideSchema>;\nexport type ModelsJsonProvider = Static<typeof ProviderConfigSchema>;\nexport type ModelsJsonRoleDefinition = Static<typeof RoleDefinitionSchema>;\ntype ModelsJson = Static<typeof ModelsConfigSchema>;\n\nfunction formatValidationPath(error: TLocalizedValidationError): string {\n\tif (error.keyword === \"required\") {\n\t\tconst requiredProperties = (error.params as { requiredProperties?: string[] }).requiredProperties;\n\t\tconst requiredProperty = requiredProperties?.[0];\n\t\tif (requiredProperty) {\n\t\t\tconst basePath = error.instancePath.replace(/^\\//, \"\").replace(/\\//g, \".\");\n\t\t\treturn basePath ? `${basePath}.${requiredProperty}` : requiredProperty;\n\t\t}\n\t}\n\tconst path = error.instancePath.replace(/^\\//, \"\").replace(/\\//g, \".\");\n\treturn path || \"root\";\n}\n\nfunction deepFreeze<T>(value: T): T {\n\tif (typeof value !== \"object\" || value === null || Object.isFrozen(value)) return value;\n\tfor (const child of Object.values(value)) deepFreeze(child);\n\treturn Object.freeze(value);\n}\n\n/** One immutable load of models.json. */\nexport class ModelConfig {\n\tprivate readonly providers: ReadonlyMap<string, ModelsJsonProvider>;\n\tprivate readonly roles: ReadonlyMap<string, readonly string[]>;\n\tprivate readonly error: string | undefined;\n\n\tprivate constructor(\n\t\tproviders: ReadonlyMap<string, ModelsJsonProvider>,\n\t\troles: ReadonlyMap<string, readonly string[]> = new Map(),\n\t\terror?: string,\n\t) {\n\t\tthis.providers = providers;\n\t\tthis.roles = roles;\n\t\tthis.error = error;\n\t}\n\n\tstatic async load(modelsJsonPath: string | undefined): Promise<ModelConfig> {\n\t\tif (!modelsJsonPath) return new ModelConfig(new Map());\n\t\tconst path = normalizePath(modelsJsonPath);\n\t\tlet content: string;\n\t\ttry {\n\t\t\tcontent = await readFile(path, \"utf-8\");\n\t\t} catch (error) {\n\t\t\tif ((error as NodeJS.ErrnoException).code === \"ENOENT\") return new ModelConfig(new Map());\n\t\t\treturn new ModelConfig(\n\t\t\t\tnew Map(),\n\t\t\t\tnew Map(),\n\t\t\t\t`Failed to load models.json: ${error instanceof Error ? error.message : error}\\n\\nFile: ${path}`,\n\t\t\t);\n\t\t}\n\n\t\tlet parsed: unknown;\n\t\ttry {\n\t\t\tparsed = JSON.parse(stripJsonComments(stripBom(content)));\n\t\t} catch (error) {\n\t\t\treturn new ModelConfig(\n\t\t\t\tnew Map(),\n\t\t\t\tnew Map(),\n\t\t\t\t`Failed to parse models.json: ${error instanceof Error ? error.message : error}\\n\\nFile: ${path}`,\n\t\t\t);\n\t\t}\n\n\t\tif (!validateModelsConfig.Check(parsed)) {\n\t\t\tconst errors =\n\t\t\t\tvalidateModelsConfig\n\t\t\t\t\t.Errors(parsed)\n\t\t\t\t\t.map((error) => `  - ${formatValidationPath(error)}: ${error.message}`)\n\t\t\t\t\t.join(\"\\n\") || \"Unknown schema error\";\n\t\t\treturn new ModelConfig(new Map(), new Map(), `Invalid models.json schema:\\n${errors}\\n\\nFile: ${path}`);\n\t\t}\n\n\t\tconst config = parsed as ModelsJson;\n\t\tconst providers = new Map<string, ModelsJsonProvider>();\n\t\tfor (const [providerId, provider] of Object.entries(config.providers)) {\n\t\t\tproviders.set(providerId, deepFreeze(structuredClone(provider)));\n\t\t}\n\t\tconst roles = new Map<string, readonly string[]>();\n\t\tfor (const [roleName, role] of Object.entries(config.roles ?? {})) {\n\t\t\troles.set(roleName, Object.freeze([...role.models]));\n\t\t}\n\t\treturn new ModelConfig(providers, roles);\n\t}\n\n\tgetProvider(providerId: string): ModelsJsonProvider | undefined {\n\t\treturn this.providers.get(providerId);\n\t}\n\n\tgetProviderIds(): readonly string[] {\n\t\treturn [...this.providers.keys()];\n\t}\n\n\t/** Additive role -> ordered \"provider/modelId\" candidate references, unresolved. */\n\tgetRoles(): ReadonlyMap<string, readonly string[]> {\n\t\treturn this.roles;\n\t}\n\n\tgetError(): string | undefined {\n\t\treturn this.error;\n\t}\n}\n"]}
+{"version":3,"file":"model-config.d.ts","sourceRoot":"","sources":["../../src/core/model-config.ts"],"names":[],"mappings":"AAAA,wDAAwD;AAGxD,OAAO,EAAE,KAAK,MAAM,EAAE,IAAI,EAAE,MAAM,SAAS,CAAC;AAwL5C,QAAA,MAAM,qBAAqB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAgBzB,CAAC;AAEH,QAAA,MAAM,mBAAmB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAqBvB,CAAC;AAEH,QAAA,MAAM,oBAAoB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAWxB,CAAC;AAEH,QAAA,MAAM,oBAAoB;IACzB,0EAA0E;;EAEzE,CAAC;AASH,MAAM,MAAM,eAAe,GAAG,MAAM,CAAC,OAAO,qBAAqB,CAAC,CAAC;AACnE,MAAM,MAAM,uBAAuB,GAAG,MAAM,CAAC,OAAO,mBAAmB,CAAC,CAAC;AACzE,MAAM,MAAM,kBAAkB,GAAG,MAAM,CAAC,OAAO,oBAAoB,CAAC,CAAC;AACrE,MAAM,MAAM,wBAAwB,GAAG,MAAM,CAAC,OAAO,oBAAoB,CAAC,CAAC;AAsB3E,yCAAyC;AACzC,qBAAa,WAAW;IACvB,OAAO,CAAC,QAAQ,CAAC,SAAS,CAA0C;IACpE,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAyC;IAC/D,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAqB;IAE3C,OAAO,eAQN;IAED,OAAa,IAAI,CAAC,cAAc,EAAE,MAAM,GAAG,SAAS,GAAG,OAAO,CAAC,WAAW,CAAC,CA6C1E;IAED,WAAW,CAAC,UAAU,EAAE,MAAM,GAAG,kBAAkB,GAAG,SAAS,CAE9D;IAED,cAAc,IAAI,SAAS,MAAM,EAAE,CAElC;IAED,oFAAoF;IACpF,QAAQ,IAAI,WAAW,CAAC,MAAM,EAAE,SAAS,MAAM,EAAE,CAAC,CAEjD;IAED,QAAQ,IAAI,MAAM,GAAG,SAAS,CAE7B;CACD","sourcesContent":["/** Immutable, credential-blind models.json snapshot. */\n\nimport { readFile } from \"node:fs/promises\";\nimport { type Static, Type } from \"typebox\";\nimport { Compile } from \"typebox/compile\";\nimport type { TLocalizedValidationError } from \"typebox/error\";\nimport { stripJsonComments } from \"../utils/json.ts\";\nimport { normalizePath } from \"../utils/paths.ts\";\nimport { stripBom } from \"../utils/text.ts\";\n\nconst PercentileCutoffsSchema = Type.Object({\n\tp50: Type.Optional(Type.Number()),\n\tp75: Type.Optional(Type.Number()),\n\tp90: Type.Optional(Type.Number()),\n\tp99: Type.Optional(Type.Number()),\n});\n\nconst OpenRouterRoutingSchema = Type.Object({\n\tallow_fallbacks: Type.Optional(Type.Boolean()),\n\trequire_parameters: Type.Optional(Type.Boolean()),\n\tdata_collection: Type.Optional(Type.Union([Type.Literal(\"deny\"), Type.Literal(\"allow\")])),\n\tzdr: Type.Optional(Type.Boolean()),\n\tenforce_distillable_text: Type.Optional(Type.Boolean()),\n\torder: Type.Optional(Type.Array(Type.String())),\n\tonly: Type.Optional(Type.Array(Type.String())),\n\tignore: Type.Optional(Type.Array(Type.String())),\n\tquantizations: Type.Optional(Type.Array(Type.String())),\n\tsort: Type.Optional(\n\t\tType.Union([\n\t\t\tType.String(),\n\t\t\tType.Object({\n\t\t\t\tby: Type.Optional(Type.String()),\n\t\t\t\tpartition: Type.Optional(Type.Union([Type.String(), Type.Null()])),\n\t\t\t}),\n\t\t]),\n\t),\n\tmax_price: Type.Optional(\n\t\tType.Object({\n\t\t\tprompt: Type.Optional(Type.Union([Type.Number(), Type.String()])),\n\t\t\tcompletion: Type.Optional(Type.Union([Type.Number(), Type.String()])),\n\t\t\timage: Type.Optional(Type.Union([Type.Number(), Type.String()])),\n\t\t\taudio: Type.Optional(Type.Union([Type.Number(), Type.String()])),\n\t\t\trequest: Type.Optional(Type.Union([Type.Number(), Type.String()])),\n\t\t}),\n\t),\n\tpreferred_min_throughput: Type.Optional(Type.Union([Type.Number(), PercentileCutoffsSchema])),\n\tpreferred_max_latency: Type.Optional(Type.Union([Type.Number(), PercentileCutoffsSchema])),\n});\n\nconst VercelGatewayRoutingSchema = Type.Object({\n\tonly: Type.Optional(Type.Array(Type.String())),\n\torder: Type.Optional(Type.Array(Type.String())),\n});\n\nconst ThinkingLevelMapValueSchema = Type.Union([Type.String(), Type.Null()]);\nconst ThinkingLevelMapSchema = Type.Object({\n\toff: Type.Optional(ThinkingLevelMapValueSchema),\n\tminimal: Type.Optional(ThinkingLevelMapValueSchema),\n\tlow: Type.Optional(ThinkingLevelMapValueSchema),\n\tmedium: Type.Optional(ThinkingLevelMapValueSchema),\n\thigh: Type.Optional(ThinkingLevelMapValueSchema),\n\txhigh: Type.Optional(ThinkingLevelMapValueSchema),\n\tmax: Type.Optional(ThinkingLevelMapValueSchema),\n});\n\nconst ChatTemplateKwargScalarSchema = Type.Union([Type.String(), Type.Number(), Type.Boolean(), Type.Null()]);\nconst ChatTemplateKwargVariableSchema = Type.Object({\n\t$var: Type.Union([Type.Literal(\"thinking.enabled\"), Type.Literal(\"thinking.effort\")]),\n\tomitWhenOff: Type.Optional(Type.Boolean()),\n});\nconst ChatTemplateKwargSchema = Type.Union([ChatTemplateKwargScalarSchema, ChatTemplateKwargVariableSchema]);\n\nconst OpenAICompletionsCompatSchema = Type.Object({\n\tsupportsStore: Type.Optional(Type.Boolean()),\n\tsupportsDeveloperRole: Type.Optional(Type.Boolean()),\n\tsupportsReasoningEffort: Type.Optional(Type.Boolean()),\n\tsupportsUsageInStreaming: Type.Optional(Type.Boolean()),\n\tsupportsFinishReason: Type.Optional(Type.Boolean()),\n\tmaxTokensField: Type.Optional(Type.Union([Type.Literal(\"max_completion_tokens\"), Type.Literal(\"max_tokens\")])),\n\trequiresToolResultName: Type.Optional(Type.Boolean()),\n\trequiresAssistantAfterToolResult: Type.Optional(Type.Boolean()),\n\trequiresThinkingAsText: Type.Optional(Type.Boolean()),\n\trequiresReasoningContentOnAssistantMessages: Type.Optional(Type.Boolean()),\n\tthinkingFormat: Type.Optional(\n\t\tType.Union([\n\t\t\tType.Literal(\"openai\"),\n\t\t\tType.Literal(\"openrouter\"),\n\t\t\tType.Literal(\"together\"),\n\t\t\tType.Literal(\"baseten\"),\n\t\t\tType.Literal(\"deepseek\"),\n\t\t\tType.Literal(\"zai\"),\n\t\t\tType.Literal(\"qwen\"),\n\t\t\tType.Literal(\"chat-template\"),\n\t\t\tType.Literal(\"qwen-chat-template\"),\n\t\t\tType.Literal(\"string-thinking\"),\n\t\t\tType.Literal(\"ant-ling\"),\n\t\t]),\n\t),\n\tchatTemplateKwargs: Type.Optional(Type.Record(Type.String(), ChatTemplateKwargSchema)),\n\tchatTemplateArgs: Type.Optional(Type.Record(Type.String(), ChatTemplateKwargSchema)),\n\tcacheControlFormat: Type.Optional(Type.Literal(\"anthropic\")),\n\topenRouterRouting: Type.Optional(OpenRouterRoutingSchema),\n\tvercelGatewayRouting: Type.Optional(VercelGatewayRoutingSchema),\n\tsupportsOpenAIGrammarTools: Type.Optional(Type.Boolean()),\n\tsupportsStrictMode: Type.Optional(Type.Boolean()),\n\tsendSessionAffinityHeaders: Type.Optional(Type.Boolean()),\n\tsessionAffinityFormat: Type.Optional(\n\t\tType.Union([Type.Literal(\"openai\"), Type.Literal(\"openai-nosession\"), Type.Literal(\"openrouter\")]),\n\t),\n\tsupportsLongCacheRetention: Type.Optional(Type.Boolean()),\n\tvllmPriority: Type.Optional(Type.Number()),\n});\n\nconst OpenAIResponsesCompatSchema = Type.Object({\n\tsupportsDeveloperRole: Type.Optional(Type.Boolean()),\n\tsessionAffinityFormat: Type.Optional(\n\t\tType.Union([Type.Literal(\"openai\"), Type.Literal(\"openai-nosession\"), Type.Literal(\"openrouter\")]),\n\t),\n\tsupportsLongCacheRetention: Type.Optional(Type.Boolean()),\n\tsupportsStrictMode: Type.Optional(Type.Boolean()),\n\tsupportsOpenAIGrammarTools: Type.Optional(Type.Boolean()),\n\tsupportsMaxOutputTokens: Type.Optional(Type.Boolean()),\n});\n\nconst ModelCostRatesSchema = {\n\tinput: Type.Number(),\n\toutput: Type.Number(),\n\tcacheRead: Type.Number(),\n\tcacheWrite: Type.Number(),\n};\nconst ModelCostTierSchema = Type.Object({\n\tinputTokensAbove: Type.Number(),\n\t...ModelCostRatesSchema,\n});\nconst ModelCostSchema = Type.Object({\n\t...ModelCostRatesSchema,\n\ttiers: Type.Optional(Type.Array(ModelCostTierSchema)),\n});\nconst ModelPromptCacheSchema = Type.Object({\n\tshort: Type.Optional(Type.Number({ exclusiveMinimum: 0 })),\n\tlong: Type.Optional(Type.Number({ exclusiveMinimum: 0 })),\n});\nconst ImageResizeSchema = Type.Object({\n\tmaxWidth: Type.Optional(Type.Integer({ minimum: 1 })),\n\tmaxHeight: Type.Optional(Type.Integer({ minimum: 1 })),\n\tmaxBytes: Type.Optional(Type.Integer({ minimum: 1 })),\n\tjpegQuality: Type.Optional(Type.Integer({ minimum: 1, maximum: 100 })),\n});\nconst ModelInputLimitsSchema = Type.Object({\n\tmaxRequestBytes: Type.Optional(Type.Integer({ minimum: 1 })),\n\timages: Type.Optional(\n\t\tType.Object({\n\t\t\tresize: Type.Optional(ImageResizeSchema),\n\t\t\tmaxPerMessage: Type.Optional(Type.Integer({ minimum: 1 })),\n\t\t\tmaxPerRequest: Type.Optional(Type.Integer({ minimum: 1 })),\n\t\t}),\n\t),\n});\n\nconst AnthropicMessagesCompatSchema = Type.Object({\n\tsupportsEagerToolInputStreaming: Type.Optional(Type.Boolean()),\n\tsupportsLongCacheRetention: Type.Optional(Type.Boolean()),\n\tsendSessionAffinityHeaders: Type.Optional(Type.Boolean()),\n\tsupportsCacheControlOnTools: Type.Optional(Type.Boolean()),\n\tsupportsTemperature: Type.Optional(Type.Boolean()),\n\tforceAdaptiveThinking: Type.Optional(Type.Boolean()),\n\tallowEmptySignature: Type.Optional(Type.Boolean()),\n\tsupportsStrictTools: Type.Optional(Type.Boolean()),\n\tsupportsMidConvoEffort: Type.Optional(Type.Boolean()),\n\tallowedFallbackModels: Type.Optional(\n\t\tType.Array(\n\t\t\tType.Object({\n\t\t\t\tprovider: Type.String({ minLength: 1 }),\n\t\t\t\tmodel: Type.String({ minLength: 1 }),\n\t\t\t\tcost: ModelCostSchema,\n\t\t\t}),\n\t\t\t{ maxItems: 3 },\n\t\t),\n\t),\n});\n\nconst ProviderCompatSchema = Type.Union([\n\tOpenAICompletionsCompatSchema,\n\tOpenAIResponsesCompatSchema,\n\tAnthropicMessagesCompatSchema,\n]);\n\nconst ModelDefinitionSchema = Type.Object({\n\tid: Type.String({ minLength: 1 }),\n\tname: Type.Optional(Type.String({ minLength: 1 })),\n\tapi: Type.Optional(Type.String({ minLength: 1 })),\n\tbaseUrl: Type.Optional(Type.String({ minLength: 1 })),\n\treasoning: Type.Optional(Type.Boolean()),\n\tthinkingLevelMap: Type.Optional(ThinkingLevelMapSchema),\n\tinput: Type.Optional(Type.Array(Type.Union([Type.Literal(\"text\"), Type.Literal(\"image\")]))),\n\tinputLimits: Type.Optional(ModelInputLimitsSchema),\n\tcost: Type.Optional(ModelCostSchema),\n\tpromptCache: Type.Optional(ModelPromptCacheSchema),\n\tcontextWindow: Type.Optional(Type.Number()),\n\tmaxTokens: Type.Optional(Type.Number()),\n\tsamplingParams: Type.Optional(Type.Record(Type.String(), Type.Unknown())),\n\theaders: Type.Optional(Type.Record(Type.String(), Type.String())),\n\tcompat: Type.Optional(ProviderCompatSchema),\n});\n\nconst ModelOverrideSchema = Type.Object({\n\tname: Type.Optional(Type.String({ minLength: 1 })),\n\treasoning: Type.Optional(Type.Boolean()),\n\tthinkingLevelMap: Type.Optional(ThinkingLevelMapSchema),\n\tinput: Type.Optional(Type.Array(Type.Union([Type.Literal(\"text\"), Type.Literal(\"image\")]))),\n\tinputLimits: Type.Optional(ModelInputLimitsSchema),\n\tcost: Type.Optional(\n\t\tType.Object({\n\t\t\tinput: Type.Optional(Type.Number()),\n\t\t\toutput: Type.Optional(Type.Number()),\n\t\t\tcacheRead: Type.Optional(Type.Number()),\n\t\t\tcacheWrite: Type.Optional(Type.Number()),\n\t\t\ttiers: Type.Optional(Type.Array(ModelCostTierSchema)),\n\t\t}),\n\t),\n\tpromptCache: Type.Optional(ModelPromptCacheSchema),\n\tcontextWindow: Type.Optional(Type.Number()),\n\tmaxTokens: Type.Optional(Type.Number()),\n\tsamplingParams: Type.Optional(Type.Record(Type.String(), Type.Unknown())),\n\theaders: Type.Optional(Type.Record(Type.String(), Type.String())),\n\tcompat: Type.Optional(ProviderCompatSchema),\n});\n\nconst ProviderConfigSchema = Type.Object({\n\tname: Type.Optional(Type.String({ minLength: 1 })),\n\tbaseUrl: Type.Optional(Type.String({ minLength: 1 })),\n\tapiKey: Type.Optional(Type.String({ minLength: 1 })),\n\tapi: Type.Optional(Type.String({ minLength: 1 })),\n\toauth: Type.Optional(Type.Literal(\"radius\")),\n\theaders: Type.Optional(Type.Record(Type.String(), Type.String())),\n\tcompat: Type.Optional(ProviderCompatSchema),\n\tauthHeader: Type.Optional(Type.Boolean()),\n\tmodels: Type.Optional(Type.Array(ModelDefinitionSchema)),\n\tmodelOverrides: Type.Optional(Type.Record(Type.String(), ModelOverrideSchema)),\n});\n\nconst RoleDefinitionSchema = Type.Object({\n\t/** Ordered, non-empty list of canonical \"provider/modelId\" candidates. */\n\tmodels: Type.Array(Type.String({ minLength: 1 }), { minItems: 1 }),\n});\n\nconst ModelsConfigSchema = Type.Object({\n\tproviders: Type.Record(Type.String(), ProviderConfigSchema),\n\t/** Additive, credential-blind role -> ordered model-candidate mapping. */\n\troles: Type.Optional(Type.Record(Type.String(), RoleDefinitionSchema)),\n});\nconst validateModelsConfig = Compile(ModelsConfigSchema);\n\nexport type ModelsJsonModel = Static<typeof ModelDefinitionSchema>;\nexport type ModelsJsonModelOverride = Static<typeof ModelOverrideSchema>;\nexport type ModelsJsonProvider = Static<typeof ProviderConfigSchema>;\nexport type ModelsJsonRoleDefinition = Static<typeof RoleDefinitionSchema>;\ntype ModelsJson = Static<typeof ModelsConfigSchema>;\n\nfunction formatValidationPath(error: TLocalizedValidationError): string {\n\tif (error.keyword === \"required\") {\n\t\tconst requiredProperties = (error.params as { requiredProperties?: string[] }).requiredProperties;\n\t\tconst requiredProperty = requiredProperties?.[0];\n\t\tif (requiredProperty) {\n\t\t\tconst basePath = error.instancePath.replace(/^\\//, \"\").replace(/\\//g, \".\");\n\t\t\treturn basePath ? `${basePath}.${requiredProperty}` : requiredProperty;\n\t\t}\n\t}\n\tconst path = error.instancePath.replace(/^\\//, \"\").replace(/\\//g, \".\");\n\treturn path || \"root\";\n}\n\nfunction deepFreeze<T>(value: T): T {\n\tif (typeof value !== \"object\" || value === null || Object.isFrozen(value)) return value;\n\tfor (const child of Object.values(value)) deepFreeze(child);\n\treturn Object.freeze(value);\n}\n\n/** One immutable load of models.json. */\nexport class ModelConfig {\n\tprivate readonly providers: ReadonlyMap<string, ModelsJsonProvider>;\n\tprivate readonly roles: ReadonlyMap<string, readonly string[]>;\n\tprivate readonly error: string | undefined;\n\n\tprivate constructor(\n\t\tproviders: ReadonlyMap<string, ModelsJsonProvider>,\n\t\troles: ReadonlyMap<string, readonly string[]> = new Map(),\n\t\terror?: string,\n\t) {\n\t\tthis.providers = providers;\n\t\tthis.roles = roles;\n\t\tthis.error = error;\n\t}\n\n\tstatic async load(modelsJsonPath: string | undefined): Promise<ModelConfig> {\n\t\tif (!modelsJsonPath) return new ModelConfig(new Map());\n\t\tconst path = normalizePath(modelsJsonPath);\n\t\tlet content: string;\n\t\ttry {\n\t\t\tcontent = await readFile(path, \"utf-8\");\n\t\t} catch (error) {\n\t\t\tif ((error as NodeJS.ErrnoException).code === \"ENOENT\") return new ModelConfig(new Map());\n\t\t\treturn new ModelConfig(\n\t\t\t\tnew Map(),\n\t\t\t\tnew Map(),\n\t\t\t\t`Failed to load models.json: ${error instanceof Error ? error.message : error}\\n\\nFile: ${path}`,\n\t\t\t);\n\t\t}\n\n\t\tlet parsed: unknown;\n\t\ttry {\n\t\t\tparsed = JSON.parse(stripJsonComments(stripBom(content)));\n\t\t} catch (error) {\n\t\t\treturn new ModelConfig(\n\t\t\t\tnew Map(),\n\t\t\t\tnew Map(),\n\t\t\t\t`Failed to parse models.json: ${error instanceof Error ? error.message : error}\\n\\nFile: ${path}`,\n\t\t\t);\n\t\t}\n\n\t\tif (!validateModelsConfig.Check(parsed)) {\n\t\t\tconst errors =\n\t\t\t\tvalidateModelsConfig\n\t\t\t\t\t.Errors(parsed)\n\t\t\t\t\t.map((error) => `  - ${formatValidationPath(error)}: ${error.message}`)\n\t\t\t\t\t.join(\"\\n\") || \"Unknown schema error\";\n\t\t\treturn new ModelConfig(new Map(), new Map(), `Invalid models.json schema:\\n${errors}\\n\\nFile: ${path}`);\n\t\t}\n\n\t\tconst config = parsed as ModelsJson;\n\t\tconst providers = new Map<string, ModelsJsonProvider>();\n\t\tfor (const [providerId, provider] of Object.entries(config.providers)) {\n\t\t\tproviders.set(providerId, deepFreeze(structuredClone(provider)));\n\t\t}\n\t\tconst roles = new Map<string, readonly string[]>();\n\t\tfor (const [roleName, role] of Object.entries(config.roles ?? {})) {\n\t\t\troles.set(roleName, Object.freeze([...role.models]));\n\t\t}\n\t\treturn new ModelConfig(providers, roles);\n\t}\n\n\tgetProvider(providerId: string): ModelsJsonProvider | undefined {\n\t\treturn this.providers.get(providerId);\n\t}\n\n\tgetProviderIds(): readonly string[] {\n\t\treturn [...this.providers.keys()];\n\t}\n\n\t/** Additive role -> ordered \"provider/modelId\" candidate references, unresolved. */\n\tgetRoles(): ReadonlyMap<string, readonly string[]> {\n\t\treturn this.roles;\n\t}\n\n\tgetError(): string | undefined {\n\t\treturn this.error;\n\t}\n}\n"]}

package/dist/core/model-config.js

@@ -90,9 +90,9 @@ const OpenAICompletionsCompatSchema = Type.Object({
     supportsOpenAIGrammarTools: Type.Optional(Type.Boolean()),
     supportsStrictMode: Type.Optional(Type.Boolean()),
     sendSessionAffinityHeaders: Type.Optional(Type.Boolean()),
-    deferredToolsMode: Type.Optional(Type.Literal("kimi")),
     sessionAffinityFormat: Type.Optional(Type.Union([Type.Literal("openai"), Type.Literal("openai-nosession"), Type.Literal("openrouter")])),
     supportsLongCacheRetention: Type.Optional(Type.Boolean()),
+    vllmPriority: Type.Optional(Type.Number()),
 });
 const OpenAIResponsesCompatSchema = Type.Object({
     supportsDeveloperRole: Type.Optional(Type.Boolean()),
@@ -100,8 +100,39 @@ const OpenAIResponsesCompatSchema = Type.Object({
     supportsLongCacheRetention: Type.Optional(Type.Boolean()),
     supportsStrictMode: Type.Optional(Type.Boolean()),
     supportsOpenAIGrammarTools: Type.Optional(Type.Boolean()),
-    supportsAdditionalTools: Type.Optional(Type.Boolean()),
-    supportsToolSearch: Type.Optional(Type.Boolean()),
+    supportsMaxOutputTokens: Type.Optional(Type.Boolean()),
+});
+const ModelCostRatesSchema = {
+    input: Type.Number(),
+    output: Type.Number(),
+    cacheRead: Type.Number(),
+    cacheWrite: Type.Number(),
+};
+const ModelCostTierSchema = Type.Object({
+    inputTokensAbove: Type.Number(),
+    ...ModelCostRatesSchema,
+});
+const ModelCostSchema = Type.Object({
+    ...ModelCostRatesSchema,
+    tiers: Type.Optional(Type.Array(ModelCostTierSchema)),
+});
+const ModelPromptCacheSchema = Type.Object({
+    short: Type.Optional(Type.Number({ exclusiveMinimum: 0 })),
+    long: Type.Optional(Type.Number({ exclusiveMinimum: 0 })),
+});
+const ImageResizeSchema = Type.Object({
+    maxWidth: Type.Optional(Type.Integer({ minimum: 1 })),
+    maxHeight: Type.Optional(Type.Integer({ minimum: 1 })),
+    maxBytes: Type.Optional(Type.Integer({ minimum: 1 })),
+    jpegQuality: Type.Optional(Type.Integer({ minimum: 1, maximum: 100 })),
+});
+const ModelInputLimitsSchema = Type.Object({
+    maxRequestBytes: Type.Optional(Type.Integer({ minimum: 1 })),
+    images: Type.Optional(Type.Object({
+        resize: Type.Optional(ImageResizeSchema),
+        maxPerMessage: Type.Optional(Type.Integer({ minimum: 1 })),
+        maxPerRequest: Type.Optional(Type.Integer({ minimum: 1 })),
+    })),
 });
 const AnthropicMessagesCompatSchema = Type.Object({
     supportsEagerToolInputStreaming: Type.Optional(Type.Boolean()),
@@ -112,27 +143,18 @@ const AnthropicMessagesCompatSchema = Type.Object({
     forceAdaptiveThinking: Type.Optional(Type.Boolean()),
     allowEmptySignature: Type.Optional(Type.Boolean()),
     supportsStrictTools: Type.Optional(Type.Boolean()),
-    supportsToolReferences: Type.Optional(Type.Boolean()),
+    supportsMidConvoEffort: Type.Optional(Type.Boolean()),
+    allowedFallbackModels: Type.Optional(Type.Array(Type.Object({
+        provider: Type.String({ minLength: 1 }),
+        model: Type.String({ minLength: 1 }),
+        cost: ModelCostSchema,
+    }), { maxItems: 3 })),
 });
 const ProviderCompatSchema = Type.Union([
     OpenAICompletionsCompatSchema,
     OpenAIResponsesCompatSchema,
     AnthropicMessagesCompatSchema,
 ]);
-const ModelCostRatesSchema = {
-    input: Type.Number(),
-    output: Type.Number(),
-    cacheRead: Type.Number(),
-    cacheWrite: Type.Number(),
-};
-const ModelCostTierSchema = Type.Object({
-    inputTokensAbove: Type.Number(),
-    ...ModelCostRatesSchema,
-});
-const ModelCostSchema = Type.Object({
-    ...ModelCostRatesSchema,
-    tiers: Type.Optional(Type.Array(ModelCostTierSchema)),
-});
 const ModelDefinitionSchema = Type.Object({
     id: Type.String({ minLength: 1 }),
     name: Type.Optional(Type.String({ minLength: 1 })),
@@ -141,7 +163,9 @@ const ModelDefinitionSchema = Type.Object({
     reasoning: Type.Optional(Type.Boolean()),
     thinkingLevelMap: Type.Optional(ThinkingLevelMapSchema),
     input: Type.Optional(Type.Array(Type.Union([Type.Literal("text"), Type.Literal("image")]))),
+    inputLimits: Type.Optional(ModelInputLimitsSchema),
     cost: Type.Optional(ModelCostSchema),
+    promptCache: Type.Optional(ModelPromptCacheSchema),
     contextWindow: Type.Optional(Type.Number()),
     maxTokens: Type.Optional(Type.Number()),
     samplingParams: Type.Optional(Type.Record(Type.String(), Type.Unknown())),
@@ -153,6 +177,7 @@ const ModelOverrideSchema = Type.Object({
     reasoning: Type.Optional(Type.Boolean()),
     thinkingLevelMap: Type.Optional(ThinkingLevelMapSchema),
     input: Type.Optional(Type.Array(Type.Union([Type.Literal("text"), Type.Literal("image")]))),
+    inputLimits: Type.Optional(ModelInputLimitsSchema),
     cost: Type.Optional(Type.Object({
         input: Type.Optional(Type.Number()),
         output: Type.Optional(Type.Number()),
@@ -160,6 +185,7 @@ const ModelOverrideSchema = Type.Object({
         cacheWrite: Type.Optional(Type.Number()),
         tiers: Type.Optional(Type.Array(ModelCostTierSchema)),
     })),
+    promptCache: Type.Optional(ModelPromptCacheSchema),
     contextWindow: Type.Optional(Type.Number()),
     maxTokens: Type.Optional(Type.Number()),
     samplingParams: Type.Optional(Type.Record(Type.String(), Type.Unknown())),