smoltalk 0.0.19 → 0.0.20

package/dist/client.js

@@ -4,25 +4,29 @@ export * from "./clients/openaiResponses.js";
 import { SmolGoogle } from "./clients/google.js";
 import { SmolOpenAi } from "./clients/openai.js";
 import { SmolOpenAiResponses } from "./clients/openaiResponses.js";
-import { getModel, isTextModel } from "./models.js";
+import { getModel, isModelConfig, isTextModel, pickModel } from "./models.js";
 import { SmolError } from "./smolError.js";
 import { getLogger } from "./logger.js";
 import { SmolOllama } from "./clients/ollama.js";
 export function getClient(config) {
     // Initialize logger singleton with desired log level
     const logger = getLogger(config.logLevel);
+    // Resolve ModelConfig to a concrete model name
+    const modelName = isModelConfig(config.model)
+        ? pickModel(config.model)
+        : config.model;
     let provider = config.provider;
     if (!provider) {
-        const model = getModel(config.model);
+        const model = getModel(modelName);
         if (model === undefined) {
-            throw new SmolError(`Model ${config.model} is not recognized. Please specify a known model, or explicitly set the provider option in the config.`);
+            throw new SmolError(`Model ${modelName} is not recognized. Please specify a known model, or explicitly set the provider option in the config.`);
         }
         if (!isTextModel(model)) {
-            throw new SmolError(`Only text models are supported currently. ${config.model} is a ${model?.type} model.`);
+            throw new SmolError(`Only text models are supported currently. ${modelName} is a ${model?.type} model.`);
         }
         provider = model.provider;
     }
-    const clientConfig = { ...config };
+    const clientConfig = { ...config, model: modelName };
     switch (provider) {
         case "openai":
             if (!config.openAiApiKey) {

package/dist/functions.js

@@ -1,6 +1,8 @@
 import { getClient } from "./client.js";
+import { isModelConfig, pickModel } from "./models.js";
 function splitConfig(config) {
-    const { openAiApiKey, googleApiKey, ollamaApiKey, ollamaHost, model, provider, logLevel, toolLoopDetection, ...promptConfig } = config;
+    const { openAiApiKey, googleApiKey, ollamaApiKey, ollamaHost, model: rawModel, provider, logLevel, toolLoopDetection, ...promptConfig } = config;
+    const model = isModelConfig(rawModel) ? pickModel(rawModel) : rawModel;
     return {
         smolConfig: {
             openAiApiKey,

package/dist/models.d.ts

@@ -1,7 +1,7 @@
-export type ModelSource = "local" | "ollama" | "openai" | "openai-responses" | "anthropic" | "google" | "replicate" | "modal";
+export type Provider = "local" | "ollama" | "openai" | "openai-responses" | "anthropic" | "google" | "replicate" | "modal";
 export type BaseModel = {
     modelName: string;
-    provider: ModelSource;
+    provider: Provider;
     description?: string;
     inputTokenCost?: number;
     cachedInputTokenCost?: number;
@@ -23,6 +23,7 @@ export type TextModel = BaseModel & {
     modelName: string;
     maxInputTokens: number;
     maxOutputTokens: number;
+    outputTokensPerSecond?: number;
 };
 export type EmbeddingsModel = {
     type: "embeddings";
@@ -49,6 +50,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 0.15;
     readonly cachedInputTokenCost: 0.075;
     readonly outputTokenCost: 0.6;
+    readonly outputTokensPerSecond: 65;
     readonly provider: "openai";
 }, {
     readonly type: "text";
@@ -59,6 +61,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 2.5;
     readonly cachedInputTokenCost: 1.25;
     readonly outputTokenCost: 10;
+    readonly outputTokensPerSecond: 143;
     readonly provider: "openai";
 }, {
     readonly type: "text";
@@ -69,6 +72,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 2;
     readonly cachedInputTokenCost: 0.5;
     readonly outputTokenCost: 8;
+    readonly outputTokensPerSecond: 94;
     readonly provider: "openai";
 }, {
     readonly type: "text";
@@ -79,6 +83,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 1.1;
     readonly cachedInputTokenCost: 0.55;
     readonly outputTokenCost: 4.4;
+    readonly outputTokensPerSecond: 214;
     readonly provider: "openai";
 }, {
     readonly type: "text";
@@ -89,6 +94,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 1.1;
     readonly cachedInputTokenCost: 0.55;
     readonly outputTokenCost: 4.4;
+    readonly outputTokensPerSecond: 135;
     readonly provider: "openai";
 }, {
     readonly type: "text";
@@ -99,6 +105,7 @@ export declare const textModels: readonly [{
     readonly inputTokenCost: 15;
     readonly cachedInputTokenCost: 7.5;
     readonly outputTokenCost: 60;
+    readonly outputTokensPerSecond: 100;
     readonly provider: "openai";
 }, {
     readonly type: "text";
@@ -166,6 +173,7 @@ export declare const textModels: readonly [{
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 1.25;
     readonly outputTokenCost: 10;
+    readonly outputTokensPerSecond: 175;
     readonly provider: "google";
 }, {
     readonly type: "text";
@@ -175,6 +183,7 @@ export declare const textModels: readonly [{
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.3;
     readonly outputTokenCost: 2.5;
+    readonly outputTokensPerSecond: 225;
     readonly provider: "google";
 }, {
     readonly type: "text";
@@ -184,6 +193,7 @@ export declare const textModels: readonly [{
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.1;
     readonly outputTokenCost: 0.4;
+    readonly outputTokensPerSecond: 400;
     readonly provider: "google";
 }, {
     readonly type: "text";
@@ -193,6 +203,7 @@ export declare const textModels: readonly [{
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.1;
     readonly outputTokenCost: 0.4;
+    readonly outputTokensPerSecond: 213;
     readonly disabled: true;
     readonly provider: "google";
 }, {
@@ -222,6 +233,7 @@ export declare const textModels: readonly [{
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.01875;
     readonly outputTokenCost: 0.075;
+    readonly outputTokensPerSecond: 178;
     readonly costUnit: "characters";
     readonly provider: "google";
 }, {
@@ -232,6 +244,7 @@ export declare const textModels: readonly [{
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.3125;
     readonly outputTokenCost: 1.25;
+    readonly outputTokensPerSecond: 59;
     readonly costUnit: "characters";
     readonly provider: "google";
 }, {
@@ -252,6 +265,7 @@ export declare const textModels: readonly [{
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 3;
     readonly outputTokenCost: 15;
+    readonly outputTokensPerSecond: 78;
     readonly provider: "anthropic";
 }, {
     readonly type: "text";
@@ -261,6 +275,7 @@ export declare const textModels: readonly [{
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.8;
     readonly outputTokenCost: 4;
+    readonly outputTokensPerSecond: 66;
     readonly provider: "anthropic";
 }, {
     readonly type: "text";
@@ -352,6 +367,7 @@ export declare function getModel(modelName: ModelName): {
     readonly inputTokenCost: 0.15;
     readonly cachedInputTokenCost: 0.075;
     readonly outputTokenCost: 0.6;
+    readonly outputTokensPerSecond: 65;
     readonly provider: "openai";
 } | {
     readonly type: "text";
@@ -362,6 +378,7 @@ export declare function getModel(modelName: ModelName): {
     readonly inputTokenCost: 2.5;
     readonly cachedInputTokenCost: 1.25;
     readonly outputTokenCost: 10;
+    readonly outputTokensPerSecond: 143;
     readonly provider: "openai";
 } | {
     readonly type: "text";
@@ -372,6 +389,7 @@ export declare function getModel(modelName: ModelName): {
     readonly inputTokenCost: 2;
     readonly cachedInputTokenCost: 0.5;
     readonly outputTokenCost: 8;
+    readonly outputTokensPerSecond: 94;
     readonly provider: "openai";
 } | {
     readonly type: "text";
@@ -382,6 +400,7 @@ export declare function getModel(modelName: ModelName): {
     readonly inputTokenCost: 1.1;
     readonly cachedInputTokenCost: 0.55;
     readonly outputTokenCost: 4.4;
+    readonly outputTokensPerSecond: 214;
     readonly provider: "openai";
 } | {
     readonly type: "text";
@@ -392,6 +411,7 @@ export declare function getModel(modelName: ModelName): {
     readonly inputTokenCost: 1.1;
     readonly cachedInputTokenCost: 0.55;
     readonly outputTokenCost: 4.4;
+    readonly outputTokensPerSecond: 135;
     readonly provider: "openai";
 } | {
     readonly type: "text";
@@ -402,6 +422,7 @@ export declare function getModel(modelName: ModelName): {
     readonly inputTokenCost: 15;
     readonly cachedInputTokenCost: 7.5;
     readonly outputTokenCost: 60;
+    readonly outputTokensPerSecond: 100;
     readonly provider: "openai";
 } | {
     readonly type: "text";
@@ -469,6 +490,7 @@ export declare function getModel(modelName: ModelName): {
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 1.25;
     readonly outputTokenCost: 10;
+    readonly outputTokensPerSecond: 175;
     readonly provider: "google";
 } | {
     readonly type: "text";
@@ -478,6 +500,7 @@ export declare function getModel(modelName: ModelName): {
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.3;
     readonly outputTokenCost: 2.5;
+    readonly outputTokensPerSecond: 225;
     readonly provider: "google";
 } | {
     readonly type: "text";
@@ -487,6 +510,7 @@ export declare function getModel(modelName: ModelName): {
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.1;
     readonly outputTokenCost: 0.4;
+    readonly outputTokensPerSecond: 400;
     readonly provider: "google";
 } | {
     readonly type: "text";
@@ -496,6 +520,7 @@ export declare function getModel(modelName: ModelName): {
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.1;
     readonly outputTokenCost: 0.4;
+    readonly outputTokensPerSecond: 213;
     readonly disabled: true;
     readonly provider: "google";
 } | {
@@ -525,6 +550,7 @@ export declare function getModel(modelName: ModelName): {
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.01875;
     readonly outputTokenCost: 0.075;
+    readonly outputTokensPerSecond: 178;
     readonly costUnit: "characters";
     readonly provider: "google";
 } | {
@@ -535,6 +561,7 @@ export declare function getModel(modelName: ModelName): {
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.3125;
     readonly outputTokenCost: 1.25;
+    readonly outputTokensPerSecond: 59;
     readonly costUnit: "characters";
     readonly provider: "google";
 } | {
@@ -555,6 +582,7 @@ export declare function getModel(modelName: ModelName): {
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 3;
     readonly outputTokenCost: 15;
+    readonly outputTokensPerSecond: 78;
     readonly provider: "anthropic";
 } | {
     readonly type: "text";
@@ -564,6 +592,7 @@ export declare function getModel(modelName: ModelName): {
     readonly maxOutputTokens: 8192;
     readonly inputTokenCost: 0.8;
     readonly outputTokenCost: 4;
+    readonly outputTokensPerSecond: 66;
     readonly provider: "anthropic";
 } | {
     readonly type: "text";
@@ -630,6 +659,13 @@ export declare function isImageModel(model: Model): model is ImageModel;
 export declare function isTextModel(model: Model): model is TextModel;
 export declare function isSpeechToTextModel(model: Model): model is SpeechToTextModel;
 export declare function isEmbeddingsModel(model: Model): model is EmbeddingsModel;
+export type Optimization = "speed" | "accuracy" | "cost" | "large-context";
+export type ModelConfig = {
+    optimizeFor: Optimization[];
+    providers: Provider[];
+};
+export declare function isModelConfig(model: ModelName | ModelConfig): model is ModelConfig;
+export declare function pickModel(config: ModelConfig, models?: readonly (typeof textModels)[number][]): TextModelName;
 export declare function calculateCost(modelName: ModelName, usage: {
     inputTokens: number;
     outputTokens: number;

package/dist/models.js

@@ -1,3 +1,4 @@
+import { SmolError } from "./smolError.js";
 import { round } from "./util.js";
 export const speechToTextModels = [
     { type: "speech-to-text", modelName: "whisper-local", provider: "local" },
@@ -28,6 +29,7 @@ export const textModels = [
         inputTokenCost: 0.15,
         cachedInputTokenCost: 0.075,
         outputTokenCost: 0.6,
+        outputTokensPerSecond: 65,
         provider: "openai",
     },
     {
@@ -39,6 +41,7 @@ export const textModels = [
         inputTokenCost: 2.5,
         cachedInputTokenCost: 1.25,
         outputTokenCost: 10,
+        outputTokensPerSecond: 143,
         provider: "openai",
     },
     {
@@ -50,6 +53,7 @@ export const textModels = [
         inputTokenCost: 2,
         cachedInputTokenCost: 0.5,
         outputTokenCost: 8,
+        outputTokensPerSecond: 94,
         provider: "openai",
     },
     {
@@ -61,6 +65,7 @@ export const textModels = [
         inputTokenCost: 1.1,
         cachedInputTokenCost: 0.55,
         outputTokenCost: 4.4,
+        outputTokensPerSecond: 214,
         provider: "openai",
     },
     {
@@ -72,6 +77,7 @@ export const textModels = [
         inputTokenCost: 1.1,
         cachedInputTokenCost: 0.55,
         outputTokenCost: 4.4,
+        outputTokensPerSecond: 135,
         provider: "openai",
     },
     {
@@ -83,6 +89,7 @@ export const textModels = [
         inputTokenCost: 15,
         cachedInputTokenCost: 7.5,
         outputTokenCost: 60,
+        outputTokensPerSecond: 100,
         provider: "openai",
     },
     {
@@ -157,6 +164,7 @@ export const textModels = [
         maxOutputTokens: 8192,
         inputTokenCost: 1.25,
         outputTokenCost: 10.0,
+        outputTokensPerSecond: 175,
         provider: "google",
     },
     {
@@ -167,6 +175,7 @@ export const textModels = [
         maxOutputTokens: 8192,
         inputTokenCost: 0.3,
         outputTokenCost: 2.5,
+        outputTokensPerSecond: 225,
         provider: "google",
     },
     {
@@ -177,6 +186,7 @@ export const textModels = [
         maxOutputTokens: 8192,
         inputTokenCost: 0.1,
         outputTokenCost: 0.4,
+        outputTokensPerSecond: 400,
         provider: "google",
     },
     {
@@ -187,6 +197,7 @@ export const textModels = [
         maxOutputTokens: 8192,
         inputTokenCost: 0.1,
         outputTokenCost: 0.4,
+        outputTokensPerSecond: 213,
         disabled: true,
         provider: "google",
     },
@@ -219,6 +230,7 @@ export const textModels = [
         maxOutputTokens: 8192,
         inputTokenCost: 0.01875,
         outputTokenCost: 0.075,
+        outputTokensPerSecond: 178,
         costUnit: "characters",
         provider: "google",
     },
@@ -230,6 +242,7 @@ export const textModels = [
         maxOutputTokens: 8192,
         inputTokenCost: 0.3125,
         outputTokenCost: 1.25,
+        outputTokensPerSecond: 59,
         costUnit: "characters",
         provider: "google",
     },
@@ -252,6 +265,7 @@ export const textModels = [
         maxOutputTokens: 8192,
         inputTokenCost: 3,
         outputTokenCost: 15,
+        outputTokensPerSecond: 78,
         provider: "anthropic",
     },
     {
@@ -262,6 +276,7 @@ export const textModels = [
         maxOutputTokens: 8192,
         inputTokenCost: 0.8,
         outputTokenCost: 4,
+        outputTokensPerSecond: 66,
         provider: "anthropic",
     },
     /*  {
@@ -388,6 +403,81 @@ export function isSpeechToTextModel(model) {
 export function isEmbeddingsModel(model) {
     return model.type === "embeddings";
 }
+export function isModelConfig(model) {
+    return typeof model === "object" && "optimizeFor" in model;
+}
+const WEIGHTS = {
+    1: [1],
+    2: [0.6, 0.4],
+    3: [0.5, 0.3, 0.2],
+    4: [0.4, 0.3, 0.2, 0.1],
+};
+function getRawMetric(model, optimization) {
+    const m = model;
+    switch (optimization) {
+        case "cost":
+            return (m.inputTokenCost ?? 0) + (m.outputTokenCost ?? 0);
+        case "speed":
+            return m.outputTokensPerSecond ?? 0;
+        case "accuracy":
+            return (m.inputTokenCost ?? 0) + (m.outputTokenCost ?? 0);
+        case "large-context":
+            return m.maxInputTokens;
+    }
+}
+function isLowerBetter(optimization) {
+    return optimization === "cost";
+}
+export function pickModel(config, models = textModels) {
+    const candidates = models.filter((m) => config.providers.includes(m.provider) &&
+        !("disabled" in m && m.disabled));
+    if (candidates.length === 0) {
+        throw new SmolError("No models available for providers: " +
+            config.providers.join(", ") +
+            ". Check that the providers have non-disabled models.");
+    }
+    if (candidates.length === 1) {
+        return candidates[0].modelName;
+    }
+    const optimizations = config.optimizeFor;
+    const weights = WEIGHTS[optimizations.length] ?? WEIGHTS[4];
+    const scores = new Map();
+    for (const c of candidates) {
+        scores.set(c.modelName, 0);
+    }
+    for (let i = 0; i < optimizations.length; i++) {
+        const opt = optimizations[i];
+        const weight = weights[i];
+        const rawValues = candidates.map((c) => getRawMetric(c, opt));
+        const min = Math.min(...rawValues);
+        const max = Math.max(...rawValues);
+        const range = max - min;
+        for (let j = 0; j < candidates.length; j++) {
+            const raw = rawValues[j];
+            let normalized;
+            if (range === 0) {
+                normalized = 0;
+            }
+            else if (isLowerBetter(opt)) {
+                normalized = (raw - min) / range;
+            }
+            else {
+                normalized = (max - raw) / range;
+            }
+            scores.set(candidates[j].modelName, scores.get(candidates[j].modelName) + weight * normalized);
+        }
+    }
+    let bestModel = candidates[0];
+    let bestScore = scores.get(candidates[0].modelName);
+    for (let i = 1; i < candidates.length; i++) {
+        const score = scores.get(candidates[i].modelName);
+        if (score < bestScore) {
+            bestScore = score;
+            bestModel = candidates[i];
+        }
+    }
+    return bestModel.modelName;
+}
 export function calculateCost(modelName, usage) {
     const model = getModel(modelName);
     if (!model || !isTextModel(model)) {

package/dist/types.d.ts

@@ -3,7 +3,7 @@ import { LogLevel } from "egonlog";
 import { ZodType } from "zod";
 import { Message } from "./classes/message/index.js";
 import { ToolCall } from "./classes/ToolCall.js";
-import { ModelName, ModelSource } from "./models.js";
+import { ModelConfig, ModelName, Provider } from "./models.js";
 import { Result } from "./types/result.js";
 export type PromptConfig = {
     messages: Message[];
@@ -31,8 +31,8 @@ export type SmolConfig = {
     googleApiKey?: string;
     ollamaApiKey?: string;
     ollamaHost?: string;
-    model: ModelName;
-    provider?: ModelSource;
+    model: ModelName | ModelConfig;
+    provider?: Provider;
     logLevel?: LogLevel;
     toolLoopDetection?: ToolLoopDetection;
 };
@@ -42,7 +42,10 @@ export type ToolLoopDetection = {
     intervention?: "remove-tool" | "remove-all-tools" | "throw-error" | "halt-execution";
     excludeTools?: string[];
 };
-export type BaseClientConfig = SmolConfig & {};
+export type ResolvedSmolConfig = Omit<SmolConfig, "model"> & {
+    model: ModelName;
+};
+export type BaseClientConfig = ResolvedSmolConfig & {};
 export type TokenUsage = {
     inputTokens: number;
     outputTokens: number;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "smoltalk",
-  "version": "0.0.19",
+  "version": "0.0.20",
   "description": "A common interface for LLM APIs",
   "homepage": "https://github.com/egonSchiele/smoltalk",
   "scripts": {
@@ -17,9 +17,9 @@
   ],
   "exports": {
     ".": {
+      "types": "./dist/index.d.ts",
       "import": "./dist/index.js",
-      "require": "./dist/index.js",
-      "types": "./dist/index.d.ts"
+      "require": "./dist/index.js"
     }
   },
   "type": "module",