omk-ai 1.5.2 → 1.5.4

package/dist/providers/bedrock-thinking.js.map

@@ -1 +1 @@
-{"version":3,"file":"bedrock-thinking.js","sourceRoot":"","sources":["../../src/providers/bedrock-thinking.ts"],"names":[],"mappings":"AAIA,8FAA8F;AAC9F,MAAM,UAAU,uBAAuB,CAAC,OAAe,EAAE,SAAkB,EAAY;IACtF,MAAM,MAAM,GAAG,SAAS,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC;IAC5D,OAAO,MAAM,CAAC,OAAO,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC;QAChC,MAAM,KAAK,GAAG,KAAK,CAAC,WAAW,EAAE,CAAC;QAClC,OAAO,CAAC,KAAK,EAAE,KAAK,CAAC,OAAO,CAAC,WAAW,EAAE,GAAG,CAAC,CAAC,CAAC;IAAA,CAChD,CAAC,CAAC;AAAA,CACH;AAED,MAAM,UAAU,wBAAwB,CAAC,OAAe,EAAE,SAAkB,EAAW;IACtF,OAAO,uBAAuB,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,IAAI,CACtD,CAAC,KAAK,EAAE,EAAE,CAAC,kCAAkC,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,KAAK,CAAC,QAAQ,CAAC,YAAY,CAAC,CACzF,CAAC;AAAA,CACF;AAED,MAAM,UAAU,wBAAwB,CACvC,KAAuC,EACvC,KAAuC,EAC9B;IACT,IACC,KAAK,KAAK,OAAO;QACjB,uBAAuB,CAAC,KAAK,CAAC,EAAE,EAAE,KAAK,CAAC,IAAI,CAAC,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,iCAAiC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QAE5G,OAAO,OAAO,CAAC;IAChB,MAAM,MAAM,GAAG,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,gBAAgB,EAAE,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IACnE,IAAI,MAAM,KAAK,KAAK,IAAI,MAAM,KAAK,QAAQ,IAAI,MAAM,KAAK,MAAM,IAAI,MAAM,KAAK,OAAO,IAAI,MAAM,KAAK,KAAK;QACzG,OAAO,MAAM,CAAC;IACf,QAAQ,KAAK,EAAE,CAAC;QACf,KAAK,SAAS,CAAC;QACf,KAAK,KAAK;YACT,OAAO,KAAK,CAAC;QACd,KAAK,QAAQ;YACZ,OAAO,QAAQ,CAAC;QACjB;YACC,OAAO,MAAM,CAAC;IAChB,CAAC;AAAA,CACD","sourcesContent":["import type { Model, SimpleStreamOptions } from \"../types.ts\";\n\ntype Effort = \"low\" | \"medium\" | \"high\" | \"xhigh\" | \"max\";\n\n/** Application inference profiles may identify their model by display name instead of ARN. */\nexport function getModelMatchCandidates(modelId: string, modelName?: string): string[] {\n\tconst values = modelName ? [modelId, modelName] : [modelId];\n\treturn values.flatMap((value) => {\n\t\tconst lower = value.toLowerCase();\n\t\treturn [lower, lower.replace(/[\\s_.:]+/g, \"-\")];\n\t});\n}\n\nexport function supportsAdaptiveThinking(modelId: string, modelName?: string): boolean {\n\treturn getModelMatchCandidates(modelId, modelName).some(\n\t\t(value) => /opus-(?:4-[678]|5(?:-5)?)(?:-|$)/.test(value) || value.includes(\"sonnet-4-6\"),\n\t);\n}\n\nexport function mapThinkingLevelToEffort(\n\tmodel: Model<\"bedrock-converse-stream\">,\n\tlevel: SimpleStreamOptions[\"reasoning\"],\n): Effort {\n\tif (\n\t\tlevel === \"xhigh\" &&\n\t\tgetModelMatchCandidates(model.id, model.name).some((value) => /opus-(?:4-[78]|5(?:-5)?)(?:-|$)/.test(value))\n\t)\n\t\treturn \"xhigh\";\n\tconst mapped = level ? model.thinkingLevelMap?.[level] : undefined;\n\tif (mapped === \"low\" || mapped === \"medium\" || mapped === \"high\" || mapped === \"xhigh\" || mapped === \"max\")\n\t\treturn mapped;\n\tswitch (level) {\n\t\tcase \"minimal\":\n\t\tcase \"low\":\n\t\t\treturn \"low\";\n\t\tcase \"medium\":\n\t\t\treturn \"medium\";\n\t\tdefault:\n\t\t\treturn \"high\";\n\t}\n}\n"]}
+{"version":3,"file":"bedrock-thinking.js","sourceRoot":"","sources":["../../src/providers/bedrock-thinking.ts"],"names":[],"mappings":"AAIA;;;;GAIG;AACH,MAAM,UAAU,sBAAsB,CAAC,KAAuC,EAAW;IACxF,MAAM,EAAE,GAAG,KAAK,CAAC,EAAE,CAAC,WAAW,EAAE,CAAC;IAClC,MAAM,IAAI,GAAG,KAAK,CAAC,IAAI,EAAE,WAAW,EAAE,IAAI,EAAE,CAAC;IAC7C,OAAO,CACN,EAAE,CAAC,QAAQ,CAAC,kBAAkB,CAAC;QAC/B,EAAE,CAAC,QAAQ,CAAC,kBAAkB,CAAC;QAC/B,IAAI,CAAC,QAAQ,CAAC,kBAAkB,CAAC;QACjC,IAAI,CAAC,QAAQ,CAAC,kBAAkB,CAAC;QACjC,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC,CACvB,CAAC;AAAA,CACF;AAED,8FAA8F;AAC9F,MAAM,UAAU,uBAAuB,CAAC,OAAe,EAAE,SAAkB,EAAY;IACtF,MAAM,MAAM,GAAG,SAAS,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC;IAC5D,OAAO,MAAM,CAAC,OAAO,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC;QAChC,MAAM,KAAK,GAAG,KAAK,CAAC,WAAW,EAAE,CAAC;QAClC,OAAO,CAAC,KAAK,EAAE,KAAK,CAAC,OAAO,CAAC,WAAW,EAAE,GAAG,CAAC,CAAC,CAAC;IAAA,CAChD,CAAC,CAAC;AAAA,CACH;AAED,MAAM,UAAU,wBAAwB,CAAC,OAAe,EAAE,SAAkB,EAAW;IACtF,OAAO,uBAAuB,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,IAAI,CACtD,CAAC,KAAK,EAAE,EAAE,CAAC,kCAAkC,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,KAAK,CAAC,QAAQ,CAAC,YAAY,CAAC,CACzF,CAAC;AAAA,CACF;AAED,MAAM,UAAU,wBAAwB,CACvC,KAAuC,EACvC,KAAuC,EAC9B;IACT,IACC,KAAK,KAAK,OAAO;QACjB,uBAAuB,CAAC,KAAK,CAAC,EAAE,EAAE,KAAK,CAAC,IAAI,CAAC,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,iCAAiC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QAE5G,OAAO,OAAO,CAAC;IAChB,MAAM,MAAM,GAAG,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,gBAAgB,EAAE,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IACnE,IAAI,MAAM,KAAK,KAAK,IAAI,MAAM,KAAK,QAAQ,IAAI,MAAM,KAAK,MAAM,IAAI,MAAM,KAAK,OAAO,IAAI,MAAM,KAAK,KAAK;QACzG,OAAO,MAAM,CAAC;IACf,QAAQ,KAAK,EAAE,CAAC;QACf,KAAK,SAAS,CAAC;QACf,KAAK,KAAK;YACT,OAAO,KAAK,CAAC;QACd,KAAK,QAAQ;YACZ,OAAO,QAAQ,CAAC;QACjB;YACC,OAAO,MAAM,CAAC;IAChB,CAAC;AAAA,CACD","sourcesContent":["import type { Model, SimpleStreamOptions } from \"../types.ts\";\n\ntype Effort = \"low\" | \"medium\" | \"high\" | \"xhigh\" | \"max\";\n\n/**\n * Check if the model is an Anthropic Claude model on Bedrock.\n * Checks both model ID and model name to support application inference profiles\n * whose ARNs don't contain the model name.\n */\nexport function isAnthropicClaudeModel(model: Model<\"bedrock-converse-stream\">): boolean {\n\tconst id = model.id.toLowerCase();\n\tconst name = model.name?.toLowerCase() ?? \"\";\n\treturn (\n\t\tid.includes(\"anthropic.claude\") ||\n\t\tid.includes(\"anthropic/claude\") ||\n\t\tname.includes(\"anthropic.claude\") ||\n\t\tname.includes(\"anthropic/claude\") ||\n\t\tname.includes(\"claude\")\n\t);\n}\n\n/** Application inference profiles may identify their model by display name instead of ARN. */\nexport function getModelMatchCandidates(modelId: string, modelName?: string): string[] {\n\tconst values = modelName ? [modelId, modelName] : [modelId];\n\treturn values.flatMap((value) => {\n\t\tconst lower = value.toLowerCase();\n\t\treturn [lower, lower.replace(/[\\s_.:]+/g, \"-\")];\n\t});\n}\n\nexport function supportsAdaptiveThinking(modelId: string, modelName?: string): boolean {\n\treturn getModelMatchCandidates(modelId, modelName).some(\n\t\t(value) => /opus-(?:4-[678]|5(?:-5)?)(?:-|$)/.test(value) || value.includes(\"sonnet-4-6\"),\n\t);\n}\n\nexport function mapThinkingLevelToEffort(\n\tmodel: Model<\"bedrock-converse-stream\">,\n\tlevel: SimpleStreamOptions[\"reasoning\"],\n): Effort {\n\tif (\n\t\tlevel === \"xhigh\" &&\n\t\tgetModelMatchCandidates(model.id, model.name).some((value) => /opus-(?:4-[78]|5(?:-5)?)(?:-|$)/.test(value))\n\t)\n\t\treturn \"xhigh\";\n\tconst mapped = level ? model.thinkingLevelMap?.[level] : undefined;\n\tif (mapped === \"low\" || mapped === \"medium\" || mapped === \"high\" || mapped === \"xhigh\" || mapped === \"max\")\n\t\treturn mapped;\n\tswitch (level) {\n\t\tcase \"minimal\":\n\t\tcase \"low\":\n\t\t\treturn \"low\";\n\t\tcase \"medium\":\n\t\t\treturn \"medium\";\n\t\tdefault:\n\t\t\treturn \"high\";\n\t}\n}\n"]}

package/dist/providers/simple-options.d.ts

@@ -1,4 +1,12 @@
 import type { Api, Model, SimpleStreamOptions, StreamOptions, ThinkingBudgets, ThinkingLevel } from "../types.ts";
+/** Default token budgets per thinking level (token-based providers). */
+export declare const DEFAULT_THINKING_BUDGETS: Readonly<Required<ThinkingBudgets>>;
+/**
+ * True when the provider's simple stream adds the thinking budget on top of an
+ * explicit `maxTokens` via {@link adjustMaxTokensForThinking} (so a caller cap
+ * sizes only the answer).
+ */
+export declare function addsThinkingBudgetToMaxTokens(model: Model<Api>): boolean;
 export declare function buildBaseOptions(_model: Model<Api>, options?: SimpleStreamOptions, apiKey?: string): StreamOptions;
 export declare function clampReasoning(effort: ThinkingLevel | undefined): Exclude<ThinkingLevel, "xhigh" | "max" | "ultra"> | undefined;
 export declare function adjustMaxTokensForThinking(baseMaxTokens: number | undefined, modelMaxTokens: number, reasoningLevel: ThinkingLevel, customBudgets?: ThinkingBudgets): {

package/dist/providers/simple-options.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"simple-options.d.ts","sourceRoot":"","sources":["../../src/providers/simple-options.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,GAAG,EAAE,KAAK,EAAE,mBAAmB,EAAE,aAAa,EAAE,eAAe,EAAE,aAAa,EAAE,MAAM,aAAa,CAAC;AAElH,wBAAgB,gBAAgB,CAAC,MAAM,EAAE,KAAK,CAAC,GAAG,CAAC,EAAE,OAAO,CAAC,EAAE,mBAAmB,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,aAAa,CAoBlH;AAED,wBAAgB,cAAc,CAC7B,MAAM,EAAE,aAAa,GAAG,SAAS,GAC/B,OAAO,CAAC,aAAa,EAAE,OAAO,GAAG,KAAK,GAAG,OAAO,CAAC,GAAG,SAAS,CAE/D;AAED,wBAAgB,0BAA0B,CAEzC,aAAa,EAAE,MAAM,GAAG,SAAS,EACjC,cAAc,EAAE,MAAM,EACtB,cAAc,EAAE,aAAa,EAC7B,aAAa,CAAC,EAAE,eAAe,GAC7B;IAAE,SAAS,EAAE,MAAM,CAAC;IAAC,cAAc,EAAE,MAAM,CAAA;CAAE,CAoB/C","sourcesContent":["import type { Api, Model, SimpleStreamOptions, StreamOptions, ThinkingBudgets, ThinkingLevel } from \"../types.ts\";\n\nexport function buildBaseOptions(_model: Model<Api>, options?: SimpleStreamOptions, apiKey?: string): StreamOptions {\n\treturn {\n\t\ttemperature: options?.temperature,\n\t\tmaxTokens: options?.maxTokens,\n\t\tsignal: options?.signal,\n\t\tapiKey: apiKey || options?.apiKey,\n\t\ttransport: options?.transport,\n\t\tcacheRetention: options?.cacheRetention,\n\t\tsessionId: options?.sessionId,\n\t\tcwd: options?.cwd,\n\t\theaders: options?.headers,\n\t\tonPayload: options?.onPayload,\n\t\tonResponse: options?.onResponse,\n\t\tonRateLimit: options?.onRateLimit,\n\t\ttimeoutMs: options?.timeoutMs,\n\t\twebsocketConnectTimeoutMs: options?.websocketConnectTimeoutMs,\n\t\tmaxRetries: options?.maxRetries,\n\t\tmaxRetryDelayMs: options?.maxRetryDelayMs,\n\t\tmetadata: options?.metadata,\n\t};\n}\n\nexport function clampReasoning(\n\teffort: ThinkingLevel | undefined,\n): Exclude<ThinkingLevel, \"xhigh\" | \"max\" | \"ultra\"> | undefined {\n\treturn effort === \"xhigh\" || effort === \"max\" || effort === \"ultra\" ? \"high\" : effort;\n}\n\nexport function adjustMaxTokensForThinking(\n\t// Undefined means no explicit caller cap. Use the model cap and fit thinking inside it.\n\tbaseMaxTokens: number | undefined,\n\tmodelMaxTokens: number,\n\treasoningLevel: ThinkingLevel,\n\tcustomBudgets?: ThinkingBudgets,\n): { maxTokens: number; thinkingBudget: number } {\n\tconst defaultBudgets: ThinkingBudgets = {\n\t\tminimal: 1024,\n\t\tlow: 2048,\n\t\tmedium: 8192,\n\t\thigh: 16384,\n\t};\n\tconst budgets = { ...defaultBudgets, ...customBudgets };\n\n\tconst minOutputTokens = 1024;\n\tconst level = clampReasoning(reasoningLevel)!;\n\tlet thinkingBudget = budgets[level]!;\n\tconst maxTokens =\n\t\tbaseMaxTokens === undefined ? modelMaxTokens : Math.min(baseMaxTokens + thinkingBudget, modelMaxTokens);\n\n\tif (maxTokens <= thinkingBudget) {\n\t\tthinkingBudget = Math.max(0, maxTokens - minOutputTokens);\n\t}\n\n\treturn { maxTokens, thinkingBudget };\n}\n"]}
+{"version":3,"file":"simple-options.d.ts","sourceRoot":"","sources":["../../src/providers/simple-options.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,GAAG,EAAE,KAAK,EAAE,mBAAmB,EAAE,aAAa,EAAE,eAAe,EAAE,aAAa,EAAE,MAAM,aAAa,CAAC;AAGlH,wEAAwE;AACxE,eAAO,MAAM,wBAAwB,EAAE,QAAQ,CAAC,QAAQ,CAAC,eAAe,CAAC,CAKxE,CAAC;AAEF;;;;GAIG;AACH,wBAAgB,6BAA6B,CAAC,KAAK,EAAE,KAAK,CAAC,GAAG,CAAC,GAAG,OAAO,CAOxE;AAED,wBAAgB,gBAAgB,CAAC,MAAM,EAAE,KAAK,CAAC,GAAG,CAAC,EAAE,OAAO,CAAC,EAAE,mBAAmB,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,aAAa,CAoBlH;AAED,wBAAgB,cAAc,CAC7B,MAAM,EAAE,aAAa,GAAG,SAAS,GAC/B,OAAO,CAAC,aAAa,EAAE,OAAO,GAAG,KAAK,GAAG,OAAO,CAAC,GAAG,SAAS,CAE/D;AAED,wBAAgB,0BAA0B,CAEzC,aAAa,EAAE,MAAM,GAAG,SAAS,EACjC,cAAc,EAAE,MAAM,EACtB,cAAc,EAAE,aAAa,EAC7B,aAAa,CAAC,EAAE,eAAe,GAC7B;IAAE,SAAS,EAAE,MAAM,CAAC;IAAC,cAAc,EAAE,MAAM,CAAA;CAAE,CAc/C","sourcesContent":["import type { Api, Model, SimpleStreamOptions, StreamOptions, ThinkingBudgets, ThinkingLevel } from \"../types.ts\";\nimport { isAnthropicClaudeModel, supportsAdaptiveThinking } from \"./bedrock-thinking.ts\";\n\n/** Default token budgets per thinking level (token-based providers). */\nexport const DEFAULT_THINKING_BUDGETS: Readonly<Required<ThinkingBudgets>> = {\n\tminimal: 1024,\n\tlow: 2048,\n\tmedium: 8192,\n\thigh: 16384,\n};\n\n/**\n * True when the provider's simple stream adds the thinking budget on top of an\n * explicit `maxTokens` via {@link adjustMaxTokensForThinking} (so a caller cap\n * sizes only the answer).\n */\nexport function addsThinkingBudgetToMaxTokens(model: Model<Api>): boolean {\n\tif (model.api === \"anthropic-messages\") return true;\n\tif (model.api === \"bedrock-converse-stream\") {\n\t\tconst bedrock = model as Model<\"bedrock-converse-stream\">;\n\t\treturn isAnthropicClaudeModel(bedrock) && !supportsAdaptiveThinking(bedrock.id, bedrock.name);\n\t}\n\treturn false;\n}\n\nexport function buildBaseOptions(_model: Model<Api>, options?: SimpleStreamOptions, apiKey?: string): StreamOptions {\n\treturn {\n\t\ttemperature: options?.temperature,\n\t\tmaxTokens: options?.maxTokens,\n\t\tsignal: options?.signal,\n\t\tapiKey: apiKey || options?.apiKey,\n\t\ttransport: options?.transport,\n\t\tcacheRetention: options?.cacheRetention,\n\t\tsessionId: options?.sessionId,\n\t\tcwd: options?.cwd,\n\t\theaders: options?.headers,\n\t\tonPayload: options?.onPayload,\n\t\tonResponse: options?.onResponse,\n\t\tonRateLimit: options?.onRateLimit,\n\t\ttimeoutMs: options?.timeoutMs,\n\t\twebsocketConnectTimeoutMs: options?.websocketConnectTimeoutMs,\n\t\tmaxRetries: options?.maxRetries,\n\t\tmaxRetryDelayMs: options?.maxRetryDelayMs,\n\t\tmetadata: options?.metadata,\n\t};\n}\n\nexport function clampReasoning(\n\teffort: ThinkingLevel | undefined,\n): Exclude<ThinkingLevel, \"xhigh\" | \"max\" | \"ultra\"> | undefined {\n\treturn effort === \"xhigh\" || effort === \"max\" || effort === \"ultra\" ? \"high\" : effort;\n}\n\nexport function adjustMaxTokensForThinking(\n\t// Undefined means no explicit caller cap. Use the model cap and fit thinking inside it.\n\tbaseMaxTokens: number | undefined,\n\tmodelMaxTokens: number,\n\treasoningLevel: ThinkingLevel,\n\tcustomBudgets?: ThinkingBudgets,\n): { maxTokens: number; thinkingBudget: number } {\n\tconst budgets = { ...DEFAULT_THINKING_BUDGETS, ...customBudgets };\n\n\tconst minOutputTokens = 1024;\n\tconst level = clampReasoning(reasoningLevel)!;\n\tlet thinkingBudget = budgets[level]!;\n\tconst maxTokens =\n\t\tbaseMaxTokens === undefined ? modelMaxTokens : Math.min(baseMaxTokens + thinkingBudget, modelMaxTokens);\n\n\tif (maxTokens <= thinkingBudget) {\n\t\tthinkingBudget = Math.max(0, maxTokens - minOutputTokens);\n\t}\n\n\treturn { maxTokens, thinkingBudget };\n}\n"]}

package/dist/providers/simple-options.js

@@ -1,3 +1,25 @@
+import { isAnthropicClaudeModel, supportsAdaptiveThinking } from "./bedrock-thinking.js";
+/** Default token budgets per thinking level (token-based providers). */
+export const DEFAULT_THINKING_BUDGETS = {
+    minimal: 1024,
+    low: 2048,
+    medium: 8192,
+    high: 16384,
+};
+/**
+ * True when the provider's simple stream adds the thinking budget on top of an
+ * explicit `maxTokens` via {@link adjustMaxTokensForThinking} (so a caller cap
+ * sizes only the answer).
+ */
+export function addsThinkingBudgetToMaxTokens(model) {
+    if (model.api === "anthropic-messages")
+        return true;
+    if (model.api === "bedrock-converse-stream") {
+        const bedrock = model;
+        return isAnthropicClaudeModel(bedrock) && !supportsAdaptiveThinking(bedrock.id, bedrock.name);
+    }
+    return false;
+}
 export function buildBaseOptions(_model, options, apiKey) {
     return {
         temperature: options?.temperature,
@@ -25,13 +47,7 @@ export function clampReasoning(effort) {
 export function adjustMaxTokensForThinking(
 // Undefined means no explicit caller cap. Use the model cap and fit thinking inside it.
 baseMaxTokens, modelMaxTokens, reasoningLevel, customBudgets) {
-    const defaultBudgets = {
-        minimal: 1024,
-        low: 2048,
-        medium: 8192,
-        high: 16384,
-    };
-    const budgets = { ...defaultBudgets, ...customBudgets };
+    const budgets = { ...DEFAULT_THINKING_BUDGETS, ...customBudgets };
     const minOutputTokens = 1024;
     const level = clampReasoning(reasoningLevel);
     let thinkingBudget = budgets[level];

package/dist/providers/simple-options.js.map

@@ -1 +1 @@
-{"version":3,"file":"simple-options.js","sourceRoot":"","sources":["../../src/providers/simple-options.ts"],"names":[],"mappings":"AAEA,MAAM,UAAU,gBAAgB,CAAC,MAAkB,EAAE,OAA6B,EAAE,MAAe,EAAiB;IACnH,OAAO;QACN,WAAW,EAAE,OAAO,EAAE,WAAW;QACjC,SAAS,EAAE,OAAO,EAAE,SAAS;QAC7B,MAAM,EAAE,OAAO,EAAE,MAAM;QACvB,MAAM,EAAE,MAAM,IAAI,OAAO,EAAE,MAAM;QACjC,SAAS,EAAE,OAAO,EAAE,SAAS;QAC7B,cAAc,EAAE,OAAO,EAAE,cAAc;QACvC,SAAS,EAAE,OAAO,EAAE,SAAS;QAC7B,GAAG,EAAE,OAAO,EAAE,GAAG;QACjB,OAAO,EAAE,OAAO,EAAE,OAAO;QACzB,SAAS,EAAE,OAAO,EAAE,SAAS;QAC7B,UAAU,EAAE,OAAO,EAAE,UAAU;QAC/B,WAAW,EAAE,OAAO,EAAE,WAAW;QACjC,SAAS,EAAE,OAAO,EAAE,SAAS;QAC7B,yBAAyB,EAAE,OAAO,EAAE,yBAAyB;QAC7D,UAAU,EAAE,OAAO,EAAE,UAAU;QAC/B,eAAe,EAAE,OAAO,EAAE,eAAe;QACzC,QAAQ,EAAE,OAAO,EAAE,QAAQ;KAC3B,CAAC;AAAA,CACF;AAED,MAAM,UAAU,cAAc,CAC7B,MAAiC,EAC+B;IAChE,OAAO,MAAM,KAAK,OAAO,IAAI,MAAM,KAAK,KAAK,IAAI,MAAM,KAAK,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC;AAAA,CACtF;AAED,MAAM,UAAU,0BAA0B;AACzC,wFAAwF;AACxF,aAAiC,EACjC,cAAsB,EACtB,cAA6B,EAC7B,aAA+B,EACiB;IAChD,MAAM,cAAc,GAAoB;QACvC,OAAO,EAAE,IAAI;QACb,GAAG,EAAE,IAAI;QACT,MAAM,EAAE,IAAI;QACZ,IAAI,EAAE,KAAK;KACX,CAAC;IACF,MAAM,OAAO,GAAG,EAAE,GAAG,cAAc,EAAE,GAAG,aAAa,EAAE,CAAC;IAExD,MAAM,eAAe,GAAG,IAAI,CAAC;IAC7B,MAAM,KAAK,GAAG,cAAc,CAAC,cAAc,CAAE,CAAC;IAC9C,IAAI,cAAc,GAAG,OAAO,CAAC,KAAK,CAAE,CAAC;IACrC,MAAM,SAAS,GACd,aAAa,KAAK,SAAS,CAAC,CAAC,CAAC,cAAc,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,aAAa,GAAG,cAAc,EAAE,cAAc,CAAC,CAAC;IAEzG,IAAI,SAAS,IAAI,cAAc,EAAE,CAAC;QACjC,cAAc,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,GAAG,eAAe,CAAC,CAAC;IAC3D,CAAC;IAED,OAAO,EAAE,SAAS,EAAE,cAAc,EAAE,CAAC;AAAA,CACrC","sourcesContent":["import type { Api, Model, SimpleStreamOptions, StreamOptions, ThinkingBudgets, ThinkingLevel } from \"../types.ts\";\n\nexport function buildBaseOptions(_model: Model<Api>, options?: SimpleStreamOptions, apiKey?: string): StreamOptions {\n\treturn {\n\t\ttemperature: options?.temperature,\n\t\tmaxTokens: options?.maxTokens,\n\t\tsignal: options?.signal,\n\t\tapiKey: apiKey || options?.apiKey,\n\t\ttransport: options?.transport,\n\t\tcacheRetention: options?.cacheRetention,\n\t\tsessionId: options?.sessionId,\n\t\tcwd: options?.cwd,\n\t\theaders: options?.headers,\n\t\tonPayload: options?.onPayload,\n\t\tonResponse: options?.onResponse,\n\t\tonRateLimit: options?.onRateLimit,\n\t\ttimeoutMs: options?.timeoutMs,\n\t\twebsocketConnectTimeoutMs: options?.websocketConnectTimeoutMs,\n\t\tmaxRetries: options?.maxRetries,\n\t\tmaxRetryDelayMs: options?.maxRetryDelayMs,\n\t\tmetadata: options?.metadata,\n\t};\n}\n\nexport function clampReasoning(\n\teffort: ThinkingLevel | undefined,\n): Exclude<ThinkingLevel, \"xhigh\" | \"max\" | \"ultra\"> | undefined {\n\treturn effort === \"xhigh\" || effort === \"max\" || effort === \"ultra\" ? \"high\" : effort;\n}\n\nexport function adjustMaxTokensForThinking(\n\t// Undefined means no explicit caller cap. Use the model cap and fit thinking inside it.\n\tbaseMaxTokens: number | undefined,\n\tmodelMaxTokens: number,\n\treasoningLevel: ThinkingLevel,\n\tcustomBudgets?: ThinkingBudgets,\n): { maxTokens: number; thinkingBudget: number } {\n\tconst defaultBudgets: ThinkingBudgets = {\n\t\tminimal: 1024,\n\t\tlow: 2048,\n\t\tmedium: 8192,\n\t\thigh: 16384,\n\t};\n\tconst budgets = { ...defaultBudgets, ...customBudgets };\n\n\tconst minOutputTokens = 1024;\n\tconst level = clampReasoning(reasoningLevel)!;\n\tlet thinkingBudget = budgets[level]!;\n\tconst maxTokens =\n\t\tbaseMaxTokens === undefined ? modelMaxTokens : Math.min(baseMaxTokens + thinkingBudget, modelMaxTokens);\n\n\tif (maxTokens <= thinkingBudget) {\n\t\tthinkingBudget = Math.max(0, maxTokens - minOutputTokens);\n\t}\n\n\treturn { maxTokens, thinkingBudget };\n}\n"]}
+{"version":3,"file":"simple-options.js","sourceRoot":"","sources":["../../src/providers/simple-options.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,sBAAsB,EAAE,wBAAwB,EAAE,MAAM,uBAAuB,CAAC;AAEzF,wEAAwE;AACxE,MAAM,CAAC,MAAM,wBAAwB,GAAwC;IAC5E,OAAO,EAAE,IAAI;IACb,GAAG,EAAE,IAAI;IACT,MAAM,EAAE,IAAI;IACZ,IAAI,EAAE,KAAK;CACX,CAAC;AAEF;;;;GAIG;AACH,MAAM,UAAU,6BAA6B,CAAC,KAAiB,EAAW;IACzE,IAAI,KAAK,CAAC,GAAG,KAAK,oBAAoB;QAAE,OAAO,IAAI,CAAC;IACpD,IAAI,KAAK,CAAC,GAAG,KAAK,yBAAyB,EAAE,CAAC;QAC7C,MAAM,OAAO,GAAG,KAAyC,CAAC;QAC1D,OAAO,sBAAsB,CAAC,OAAO,CAAC,IAAI,CAAC,wBAAwB,CAAC,OAAO,CAAC,EAAE,EAAE,OAAO,CAAC,IAAI,CAAC,CAAC;IAC/F,CAAC;IACD,OAAO,KAAK,CAAC;AAAA,CACb;AAED,MAAM,UAAU,gBAAgB,CAAC,MAAkB,EAAE,OAA6B,EAAE,MAAe,EAAiB;IACnH,OAAO;QACN,WAAW,EAAE,OAAO,EAAE,WAAW;QACjC,SAAS,EAAE,OAAO,EAAE,SAAS;QAC7B,MAAM,EAAE,OAAO,EAAE,MAAM;QACvB,MAAM,EAAE,MAAM,IAAI,OAAO,EAAE,MAAM;QACjC,SAAS,EAAE,OAAO,EAAE,SAAS;QAC7B,cAAc,EAAE,OAAO,EAAE,cAAc;QACvC,SAAS,EAAE,OAAO,EAAE,SAAS;QAC7B,GAAG,EAAE,OAAO,EAAE,GAAG;QACjB,OAAO,EAAE,OAAO,EAAE,OAAO;QACzB,SAAS,EAAE,OAAO,EAAE,SAAS;QAC7B,UAAU,EAAE,OAAO,EAAE,UAAU;QAC/B,WAAW,EAAE,OAAO,EAAE,WAAW;QACjC,SAAS,EAAE,OAAO,EAAE,SAAS;QAC7B,yBAAyB,EAAE,OAAO,EAAE,yBAAyB;QAC7D,UAAU,EAAE,OAAO,EAAE,UAAU;QAC/B,eAAe,EAAE,OAAO,EAAE,eAAe;QACzC,QAAQ,EAAE,OAAO,EAAE,QAAQ;KAC3B,CAAC;AAAA,CACF;AAED,MAAM,UAAU,cAAc,CAC7B,MAAiC,EAC+B;IAChE,OAAO,MAAM,KAAK,OAAO,IAAI,MAAM,KAAK,KAAK,IAAI,MAAM,KAAK,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC;AAAA,CACtF;AAED,MAAM,UAAU,0BAA0B;AACzC,wFAAwF;AACxF,aAAiC,EACjC,cAAsB,EACtB,cAA6B,EAC7B,aAA+B,EACiB;IAChD,MAAM,OAAO,GAAG,EAAE,GAAG,wBAAwB,EAAE,GAAG,aAAa,EAAE,CAAC;IAElE,MAAM,eAAe,GAAG,IAAI,CAAC;IAC7B,MAAM,KAAK,GAAG,cAAc,CAAC,cAAc,CAAE,CAAC;IAC9C,IAAI,cAAc,GAAG,OAAO,CAAC,KAAK,CAAE,CAAC;IACrC,MAAM,SAAS,GACd,aAAa,KAAK,SAAS,CAAC,CAAC,CAAC,cAAc,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,aAAa,GAAG,cAAc,EAAE,cAAc,CAAC,CAAC;IAEzG,IAAI,SAAS,IAAI,cAAc,EAAE,CAAC;QACjC,cAAc,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,GAAG,eAAe,CAAC,CAAC;IAC3D,CAAC;IAED,OAAO,EAAE,SAAS,EAAE,cAAc,EAAE,CAAC;AAAA,CACrC","sourcesContent":["import type { Api, Model, SimpleStreamOptions, StreamOptions, ThinkingBudgets, ThinkingLevel } from \"../types.ts\";\nimport { isAnthropicClaudeModel, supportsAdaptiveThinking } from \"./bedrock-thinking.ts\";\n\n/** Default token budgets per thinking level (token-based providers). */\nexport const DEFAULT_THINKING_BUDGETS: Readonly<Required<ThinkingBudgets>> = {\n\tminimal: 1024,\n\tlow: 2048,\n\tmedium: 8192,\n\thigh: 16384,\n};\n\n/**\n * True when the provider's simple stream adds the thinking budget on top of an\n * explicit `maxTokens` via {@link adjustMaxTokensForThinking} (so a caller cap\n * sizes only the answer).\n */\nexport function addsThinkingBudgetToMaxTokens(model: Model<Api>): boolean {\n\tif (model.api === \"anthropic-messages\") return true;\n\tif (model.api === \"bedrock-converse-stream\") {\n\t\tconst bedrock = model as Model<\"bedrock-converse-stream\">;\n\t\treturn isAnthropicClaudeModel(bedrock) && !supportsAdaptiveThinking(bedrock.id, bedrock.name);\n\t}\n\treturn false;\n}\n\nexport function buildBaseOptions(_model: Model<Api>, options?: SimpleStreamOptions, apiKey?: string): StreamOptions {\n\treturn {\n\t\ttemperature: options?.temperature,\n\t\tmaxTokens: options?.maxTokens,\n\t\tsignal: options?.signal,\n\t\tapiKey: apiKey || options?.apiKey,\n\t\ttransport: options?.transport,\n\t\tcacheRetention: options?.cacheRetention,\n\t\tsessionId: options?.sessionId,\n\t\tcwd: options?.cwd,\n\t\theaders: options?.headers,\n\t\tonPayload: options?.onPayload,\n\t\tonResponse: options?.onResponse,\n\t\tonRateLimit: options?.onRateLimit,\n\t\ttimeoutMs: options?.timeoutMs,\n\t\twebsocketConnectTimeoutMs: options?.websocketConnectTimeoutMs,\n\t\tmaxRetries: options?.maxRetries,\n\t\tmaxRetryDelayMs: options?.maxRetryDelayMs,\n\t\tmetadata: options?.metadata,\n\t};\n}\n\nexport function clampReasoning(\n\teffort: ThinkingLevel | undefined,\n): Exclude<ThinkingLevel, \"xhigh\" | \"max\" | \"ultra\"> | undefined {\n\treturn effort === \"xhigh\" || effort === \"max\" || effort === \"ultra\" ? \"high\" : effort;\n}\n\nexport function adjustMaxTokensForThinking(\n\t// Undefined means no explicit caller cap. Use the model cap and fit thinking inside it.\n\tbaseMaxTokens: number | undefined,\n\tmodelMaxTokens: number,\n\treasoningLevel: ThinkingLevel,\n\tcustomBudgets?: ThinkingBudgets,\n): { maxTokens: number; thinkingBudget: number } {\n\tconst budgets = { ...DEFAULT_THINKING_BUDGETS, ...customBudgets };\n\n\tconst minOutputTokens = 1024;\n\tconst level = clampReasoning(reasoningLevel)!;\n\tlet thinkingBudget = budgets[level]!;\n\tconst maxTokens =\n\t\tbaseMaxTokens === undefined ? modelMaxTokens : Math.min(baseMaxTokens + thinkingBudget, modelMaxTokens);\n\n\tif (maxTokens <= thinkingBudget) {\n\t\tthinkingBudget = Math.max(0, maxTokens - minOutputTokens);\n\t}\n\n\treturn { maxTokens, thinkingBudget };\n}\n"]}

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "omk-ai",
-	"version": "1.5.2",
+	"version": "1.5.4",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"type": "module",
 	"main": "./dist/index.js",