modelfusion 0.101.0 → 0.103.0

package/model-provider/automatic1111/Automatic1111ImageGenerationModel.js

@@ -34,13 +34,15 @@ export class Automatic1111ImageGenerationModel extends AbstractModel {
                 url: api.assembleUrl(`/txt2img`),
                 headers: api.headers,
                 body: {
-                    height: this.settings.height,
-                    width: this.settings.width,
                     prompt: input.prompt,
                     negative_prompt: input.negativePrompt,
+                    seed: this.settings.seed,
+                    batch_size: this.settings.numberOfGenerations,
+                    height: this.settings.height,
+                    width: this.settings.width,
+                    cfg_scale: this.settings.cfgScale,
                     sampler_index: this.settings.sampler,
                     steps: this.settings.steps,
-                    seed: input.seed,
                     override_settings: {
                         sd_model_checkpoint: this.settings.model,
                     },
@@ -59,11 +61,11 @@ export class Automatic1111ImageGenerationModel extends AbstractModel {
             steps: this.settings.steps,
         };
     }
-    async doGenerateImage(prompt, options) {
+    async doGenerateImages(prompt, options) {
         const response = await this.callAPI(prompt, options);
         return {
             response,
-            base64Image: response.images[0],
+            base64Images: response.images,
         };
     }
     withTextPrompt() {

package/model-provider/automatic1111/Automatic1111ImageGenerationPrompt.d.ts

@@ -2,7 +2,6 @@ import { PromptTemplate } from "../../model-function/PromptTemplate.js";
 export type Automatic1111ImageGenerationPrompt = {
     prompt: string;
     negativePrompt?: string;
-    seed?: number;
 };
 /**
  * Formats a basic text prompt as an Automatic1111 prompt.

package/model-provider/cohere/CohereTextGenerationModel.cjs

@@ -7,6 +7,7 @@ const postToApi_js_1 = require("../../core/api/postToApi.cjs");
 const ZodSchema_js_1 = require("../../core/schema/ZodSchema.cjs");
 const AbstractModel_js_1 = require("../../model-function/AbstractModel.cjs");
 const PromptTemplateTextStreamingModel_js_1 = require("../../model-function/generate-text/PromptTemplateTextStreamingModel.cjs");
+const TextGenerationModel_js_1 = require("../../model-function/generate-text/TextGenerationModel.cjs");
 const TextPromptTemplate_js_1 = require("../../model-function/generate-text/prompt-template/TextPromptTemplate.cjs");
 const countTokens_js_1 = require("../../model-function/tokenize-text/countTokens.cjs");
 const AsyncQueue_js_1 = require("../../util/AsyncQueue.cjs");
@@ -84,8 +85,8 @@ class CohereTextGenerationModel extends AbstractModel_js_1.AbstractModel {
         const responseFormat = options.responseFormat;
         const abortSignal = options.run?.abortSignal;
         return (0, callWithRetryAndThrottle_js_1.callWithRetryAndThrottle)({
-            retry: this.settings.api?.retry,
-            throttle: this.settings.api?.throttle,
+            retry: api.retry,
+            throttle: api.throttle,
             call: async () => {
                 return (0, postToApi_js_1.postJsonToApi)({
                     url: api.assembleUrl(`/generate`),
@@ -116,9 +117,7 @@ class CohereTextGenerationModel extends AbstractModel_js_1.AbstractModel {
     }
     get settingsForEvent() {
         const eventSettingProperties = [
-            "maxGenerationTokens",
-            "stopSequences",
-            "numberOfGenerations",
+            ...TextGenerationModel_js_1.textGenerationModelProperties,
             "temperature",
             "k",
             "p",
@@ -138,9 +137,26 @@ class CohereTextGenerationModel extends AbstractModel_js_1.AbstractModel {
         });
         return {
             response,
-            texts: response.generations.map((generation) => generation.text),
+            textGenerationResults: response.generations.map((generation) => ({
+                text: generation.text,
+                finishReason: this.translateFinishReason(generation.finish_reason),
+            })),
         };
     }
+    translateFinishReason(finishReason) {
+        switch (finishReason) {
+            case "COMPLETE":
+                return "stop";
+            case "MAX_TOKENS":
+                return "length";
+            case "ERROR_TOXIC":
+                return "content-filter";
+            case "ERROR":
+                return "error";
+            default:
+                return "unknown";
+        }
+    }
     doStreamText(prompt, options) {
         return this.callAPI(prompt, {
             ...options,

package/model-provider/cohere/CohereTextGenerationModel.d.ts

@@ -7,6 +7,7 @@ import { Delta } from "../../model-function/Delta.js";
 import { PromptTemplateTextStreamingModel } from "../../model-function/generate-text/PromptTemplateTextStreamingModel.js";
 import { TextGenerationModelSettings, TextStreamingModel } from "../../model-function/generate-text/TextGenerationModel.js";
 import { TextGenerationPromptTemplate } from "../../model-function/generate-text/TextGenerationPromptTemplate.js";
+import { TextGenerationFinishReason } from "../../model-function/generate-text/TextGenerationResult.js";
 import { CohereTokenizer } from "./CohereTokenizer.js";
 export declare const COHERE_TEXT_GENERATION_MODELS: {
     command: {
@@ -79,8 +80,12 @@ export declare class CohereTextGenerationModel extends AbstractModel<CohereTextG
                 };
             } | undefined;
         };
-        texts: string[];
+        textGenerationResults: {
+            text: string;
+            finishReason: TextGenerationFinishReason;
+        }[];
     }>;
+    private translateFinishReason;
     doStreamText(prompt: string, options?: FunctionOptions): Promise<AsyncIterable<Delta<string>>>;
     extractTextDelta(fullDelta: CohereTextGenerationDelta): string | undefined;
     /**

package/model-provider/cohere/CohereTextGenerationModel.js

@@ -4,6 +4,7 @@ import { createJsonResponseHandler, postJsonToApi, } from "../../core/api/postTo
 import { ZodSchema } from "../../core/schema/ZodSchema.js";
 import { AbstractModel } from "../../model-function/AbstractModel.js";
 import { PromptTemplateTextStreamingModel } from "../../model-function/generate-text/PromptTemplateTextStreamingModel.js";
+import { textGenerationModelProperties, } from "../../model-function/generate-text/TextGenerationModel.js";
 import { chat, instruction, } from "../../model-function/generate-text/prompt-template/TextPromptTemplate.js";
 import { countTokens } from "../../model-function/tokenize-text/countTokens.js";
 import { AsyncQueue } from "../../util/AsyncQueue.js";
@@ -81,8 +82,8 @@ export class CohereTextGenerationModel extends AbstractModel {
         const responseFormat = options.responseFormat;
         const abortSignal = options.run?.abortSignal;
         return callWithRetryAndThrottle({
-            retry: this.settings.api?.retry,
-            throttle: this.settings.api?.throttle,
+            retry: api.retry,
+            throttle: api.throttle,
             call: async () => {
                 return postJsonToApi({
                     url: api.assembleUrl(`/generate`),
@@ -113,9 +114,7 @@ export class CohereTextGenerationModel extends AbstractModel {
     }
     get settingsForEvent() {
         const eventSettingProperties = [
-            "maxGenerationTokens",
-            "stopSequences",
-            "numberOfGenerations",
+            ...textGenerationModelProperties,
             "temperature",
             "k",
             "p",
@@ -135,9 +134,26 @@ export class CohereTextGenerationModel extends AbstractModel {
         });
         return {
             response,
-            texts: response.generations.map((generation) => generation.text),
+            textGenerationResults: response.generations.map((generation) => ({
+                text: generation.text,
+                finishReason: this.translateFinishReason(generation.finish_reason),
+            })),
         };
     }
+    translateFinishReason(finishReason) {
+        switch (finishReason) {
+            case "COMPLETE":
+                return "stop";
+            case "MAX_TOKENS":
+                return "length";
+            case "ERROR_TOXIC":
+                return "content-filter";
+            case "ERROR":
+                return "error";
+            default:
+                return "unknown";
+        }
+    }
     doStreamText(prompt, options) {
         return this.callAPI(prompt, {
             ...options,

package/model-provider/elevenlabs/ElevenLabsSpeechModel.cjs

@@ -101,7 +101,7 @@ class ElevenLabsSpeechModel extends AbstractModel_js_1.AbstractModel {
                 // The JS WebSocket API does not support authorization headers, so we send the API key in the BOS message.
                 // See https://stackoverflow.com/questions/4361173/http-headers-in-websockets-client-api
                 xi_api_key: api.apiKey,
-                text: " ",
+                text: " ", // first message
                 voice_settings: toApiVoiceSettings(this.settings.voiceSettings),
                 generation_config: toGenerationConfig(this.settings.generationConfig),
             }));
@@ -125,7 +125,7 @@ class ElevenLabsSpeechModel extends AbstractModel_js_1.AbstractModel {
             // send remaining text:
             if (textBuffer.length > 0) {
                 socket.send(JSON.stringify({
-                    text: `${textBuffer} `,
+                    text: `${textBuffer} `, // append space
                     try_trigger_generation: true,
                 }));
             }

package/model-provider/elevenlabs/ElevenLabsSpeechModel.js

@@ -98,7 +98,7 @@ export class ElevenLabsSpeechModel extends AbstractModel {
                 // The JS WebSocket API does not support authorization headers, so we send the API key in the BOS message.
                 // See https://stackoverflow.com/questions/4361173/http-headers-in-websockets-client-api
                 xi_api_key: api.apiKey,
-                text: " ",
+                text: " ", // first message
                 voice_settings: toApiVoiceSettings(this.settings.voiceSettings),
                 generation_config: toGenerationConfig(this.settings.generationConfig),
             }));
@@ -122,7 +122,7 @@ export class ElevenLabsSpeechModel extends AbstractModel {
             // send remaining text:
             if (textBuffer.length > 0) {
                 socket.send(JSON.stringify({
-                    text: `${textBuffer} `,
+                    text: `${textBuffer} `, // append space
                     try_trigger_generation: true,
                 }));
             }

package/model-provider/huggingface/HuggingFaceTextGenerationModel.cjs

@@ -6,6 +6,7 @@ const callWithRetryAndThrottle_js_1 = require("../../core/api/callWithRetryAndTh
 const postToApi_js_1 = require("../../core/api/postToApi.cjs");
 const AbstractModel_js_1 = require("../../model-function/AbstractModel.cjs");
 const PromptTemplateTextGenerationModel_js_1 = require("../../model-function/generate-text/PromptTemplateTextGenerationModel.cjs");
+const TextGenerationModel_js_1 = require("../../model-function/generate-text/TextGenerationModel.cjs");
 const HuggingFaceApiConfiguration_js_1 = require("./HuggingFaceApiConfiguration.cjs");
 const HuggingFaceError_js_1 = require("./HuggingFaceError.cjs");
 /**
@@ -61,8 +62,8 @@ class HuggingFaceTextGenerationModel extends AbstractModel_js_1.AbstractModel {
         const api = this.settings.api ?? new HuggingFaceApiConfiguration_js_1.HuggingFaceApiConfiguration();
         const abortSignal = options?.run?.abortSignal;
         return (0, callWithRetryAndThrottle_js_1.callWithRetryAndThrottle)({
-            retry: this.settings.api?.retry,
-            throttle: this.settings.api?.throttle,
+            retry: api.retry,
+            throttle: api.throttle,
             call: async () => {
                 return (0, postToApi_js_1.postJsonToApi)({
                     url: api.assembleUrl(`/${this.settings.model}`),
@@ -91,16 +92,13 @@ class HuggingFaceTextGenerationModel extends AbstractModel_js_1.AbstractModel {
     }
     get settingsForEvent() {
         const eventSettingProperties = [
-            "stopSequences",
-            "maxGenerationTokens",
-            "numberOfGenerations",
+            ...TextGenerationModel_js_1.textGenerationModelProperties,
             "topK",
             "topP",
             "temperature",
             "repetitionPenalty",
             "maxTime",
             "doSample",
-            "options",
         ];
         return Object.fromEntries(Object.entries(this.settings).filter(([key]) => eventSettingProperties.includes(key)));
     }
@@ -108,12 +106,15 @@ class HuggingFaceTextGenerationModel extends AbstractModel_js_1.AbstractModel {
         const response = await this.callAPI(prompt, options);
         return {
             response,
-            texts: response.map((response) => response.generated_text),
+            textGenerationResults: response.map((response) => ({
+                text: response.generated_text,
+                finishReason: "unknown",
+            })),
         };
     }
     withPromptTemplate(promptTemplate) {
         return new PromptTemplateTextGenerationModel_js_1.PromptTemplateTextGenerationModel({
-            model: this,
+            model: this, // stop tokens are not supported by this model
             promptTemplate,
         });
     }

package/model-provider/huggingface/HuggingFaceTextGenerationModel.d.ts

@@ -14,10 +14,6 @@ export interface HuggingFaceTextGenerationModelSettings extends TextGenerationMo
     repetitionPenalty?: number;
     maxTime?: number;
     doSample?: boolean;
-    options?: {
-        useCache?: boolean;
-        waitForModel?: boolean;
-    };
 }
 /**
  * Create a text generation model that calls a Hugging Face Inference API Text Generation Task.
@@ -50,7 +46,10 @@ export declare class HuggingFaceTextGenerationModel extends AbstractModel<Huggin
         response: {
             generated_text: string;
         }[];
-        texts: string[];
+        textGenerationResults: {
+            text: string;
+            finishReason: "unknown";
+        }[];
     }>;
     withPromptTemplate<INPUT_PROMPT>(promptTemplate: TextGenerationPromptTemplate<INPUT_PROMPT, string>): PromptTemplateTextGenerationModel<INPUT_PROMPT, string, HuggingFaceTextGenerationModelSettings, this>;
     withSettings(additionalSettings: Partial<HuggingFaceTextGenerationModelSettings>): this;

package/model-provider/huggingface/HuggingFaceTextGenerationModel.js

@@ -3,6 +3,7 @@ import { callWithRetryAndThrottle } from "../../core/api/callWithRetryAndThrottl
 import { createJsonResponseHandler, postJsonToApi, } from "../../core/api/postToApi.js";
 import { AbstractModel } from "../../model-function/AbstractModel.js";
 import { PromptTemplateTextGenerationModel } from "../../model-function/generate-text/PromptTemplateTextGenerationModel.js";
+import { textGenerationModelProperties, } from "../../model-function/generate-text/TextGenerationModel.js";
 import { HuggingFaceApiConfiguration } from "./HuggingFaceApiConfiguration.js";
 import { failedHuggingFaceCallResponseHandler } from "./HuggingFaceError.js";
 /**
@@ -58,8 +59,8 @@ export class HuggingFaceTextGenerationModel extends AbstractModel {
         const api = this.settings.api ?? new HuggingFaceApiConfiguration();
         const abortSignal = options?.run?.abortSignal;
         return callWithRetryAndThrottle({
-            retry: this.settings.api?.retry,
-            throttle: this.settings.api?.throttle,
+            retry: api.retry,
+            throttle: api.throttle,
             call: async () => {
                 return postJsonToApi({
                     url: api.assembleUrl(`/${this.settings.model}`),
@@ -88,16 +89,13 @@ export class HuggingFaceTextGenerationModel extends AbstractModel {
     }
     get settingsForEvent() {
         const eventSettingProperties = [
-            "stopSequences",
-            "maxGenerationTokens",
-            "numberOfGenerations",
+            ...textGenerationModelProperties,
             "topK",
             "topP",
             "temperature",
             "repetitionPenalty",
             "maxTime",
             "doSample",
-            "options",
         ];
         return Object.fromEntries(Object.entries(this.settings).filter(([key]) => eventSettingProperties.includes(key)));
     }
@@ -105,12 +103,15 @@ export class HuggingFaceTextGenerationModel extends AbstractModel {
         const response = await this.callAPI(prompt, options);
         return {
             response,
-            texts: response.map((response) => response.generated_text),
+            textGenerationResults: response.map((response) => ({
+                text: response.generated_text,
+                finishReason: "unknown",
+            })),
         };
     }
     withPromptTemplate(promptTemplate) {
         return new PromptTemplateTextGenerationModel({
-            model: this,
+            model: this, // stop tokens are not supported by this model
             promptTemplate,
         });
     }

package/model-provider/llamacpp/LlamaCppTextGenerationModel.cjs

@@ -7,6 +7,7 @@ const postToApi_js_1 = require("../../core/api/postToApi.cjs");
 const ZodSchema_js_1 = require("../../core/schema/ZodSchema.cjs");
 const AbstractModel_js_1 = require("../../model-function/AbstractModel.cjs");
 const PromptTemplateTextStreamingModel_js_1 = require("../../model-function/generate-text/PromptTemplateTextStreamingModel.cjs");
+const TextGenerationModel_js_1 = require("../../model-function/generate-text/TextGenerationModel.cjs");
 const AsyncQueue_js_1 = require("../../util/AsyncQueue.cjs");
 const parseJSON_js_1 = require("../../core/schema/parseJSON.cjs");
 const parseEventSourceStream_js_1 = require("../../util/streaming/parseEventSourceStream.cjs");
@@ -37,25 +38,52 @@ class LlamaCppTextGenerationModel extends AbstractModel_js_1.AbstractModel {
         return this.settings.contextWindowSize;
     }
     async callAPI(prompt, options) {
+        const api = this.settings.api ?? new LlamaCppApiConfiguration_js_1.LlamaCppApiConfiguration();
+        const responseFormat = options.responseFormat;
+        const abortSignal = options.run?.abortSignal;
         return (0, callWithRetryAndThrottle_js_1.callWithRetryAndThrottle)({
-            retry: this.settings.api?.retry,
-            throttle: this.settings.api?.throttle,
-            call: async () => callLlamaCppTextGenerationAPI({
-                ...this.settings,
-                // mapping
-                nPredict: this.settings.maxGenerationTokens,
-                stop: this.settings.stopSequences,
-                // other
-                abortSignal: options.run?.abortSignal,
-                prompt,
-                responseFormat: options.responseFormat,
+            retry: api.retry,
+            throttle: api.throttle,
+            call: async () => (0, postToApi_js_1.postJsonToApi)({
+                url: api.assembleUrl(`/completion`),
+                headers: api.headers,
+                body: {
+                    stream: responseFormat.stream,
+                    prompt: prompt.text,
+                    image_data: prompt.images != null
+                        ? Object.entries(prompt.images).map(([id, data]) => ({
+                            id: +id,
+                            data,
+                        }))
+                        : undefined,
+                    cache_prompt: this.settings.cachePrompt,
+                    temperature: this.settings.temperature,
+                    top_k: this.settings.topK,
+                    top_p: this.settings.topP,
+                    n_predict: this.settings.maxGenerationTokens,
+                    n_keep: this.settings.nKeep,
+                    stop: this.settings.stopSequences,
+                    tfs_z: this.settings.tfsZ,
+                    typical_p: this.settings.typicalP,
+                    repeat_penalty: this.settings.repeatPenalty,
+                    repeat_last_n: this.settings.repeatLastN,
+                    penalize_nl: this.settings.penalizeNl,
+                    mirostat: this.settings.mirostat,
+                    mirostat_tau: this.settings.mirostatTau,
+                    mirostat_eta: this.settings.mirostatEta,
+                    seed: this.settings.seed,
+                    ignore_eos: this.settings.ignoreEos,
+                    logit_bias: this.settings.logitBias,
+                },
+                failedResponseHandler: LlamaCppError_js_1.failedLlamaCppCallResponseHandler,
+                successfulResponseHandler: responseFormat.handler,
+                abortSignal,
             }),
         });
     }
     get settingsForEvent() {
         const eventSettingProperties = [
-            "maxGenerationTokens",
-            "stopSequences",
+            ...TextGenerationModel_js_1.textGenerationModelProperties,
             "contextWindowSize",
             "cachePrompt",
             "temperature",
@@ -87,7 +115,16 @@ class LlamaCppTextGenerationModel extends AbstractModel_js_1.AbstractModel {
         });
         return {
             response,
-            texts: [response.content],
+            textGenerationResults: [
+                {
+                    text: response.content,
+                    finishReason: response.stopped_eos || response.stopped_word
+                        ? "stop"
+                        : response.stopped_limit
+                            ? "length"
+                            : "unknown",
+                },
+            ],
             usage: {
                 promptTokens: response.tokens_evaluated,
                 completionTokens: response.tokens_predicted,
@@ -198,43 +235,6 @@ const llamaCppTextStreamingResponseSchema = new ZodSchema_js_1.ZodSchema(zod_1.z
     }),
     llamaCppTextGenerationResponseSchema,
 ]));
-async function callLlamaCppTextGenerationAPI({ api = new LlamaCppApiConfiguration_js_1.LlamaCppApiConfiguration(), abortSignal, responseFormat, prompt, cachePrompt, temperature, topK, topP, nPredict, nKeep, stop, tfsZ, typicalP, repeatPenalty, repeatLastN, penalizeNl, mirostat, mirostatTau, mirostatEta, seed, ignoreEos, logitBias, }) {
-    return (0, postToApi_js_1.postJsonToApi)({
-        url: api.assembleUrl(`/completion`),
-        headers: api.headers,
-        body: {
-            stream: responseFormat.stream,
-            prompt: prompt.text,
-            cache_prompt: cachePrompt,
-            temperature,
-            top_k: topK,
-            top_p: topP,
-            n_predict: nPredict,
-            n_keep: nKeep,
-            stop,
-            tfs_z: tfsZ,
-            typical_p: typicalP,
-            repeat_penalty: repeatPenalty,
-            repeat_last_n: repeatLastN,
-            penalize_nl: penalizeNl,
-            mirostat,
-            mirostat_tau: mirostatTau,
-            mirostat_eta: mirostatEta,
-            seed,
-            ignore_eos: ignoreEos,
-            logit_bias: logitBias,
-            image_data: prompt.images != null
-                ? Object.entries(prompt.images).map(([id, data]) => ({
-                    id: +id,
-                    data,
-                }))
-                : undefined,
-        },
-        failedResponseHandler: LlamaCppError_js_1.failedLlamaCppCallResponseHandler,
-        successfulResponseHandler: responseFormat.handler,
-        abortSignal,
-    });
-}
 async function createLlamaCppFullDeltaIterableQueue(stream) {
     const queue = new AsyncQueue_js_1.AsyncQueue();
     let content = "";

package/model-provider/llamacpp/LlamaCppTextGenerationModel.d.ts

@@ -58,16 +58,16 @@ export declare class LlamaCppTextGenerationModel<CONTEXT_WINDOW_SIZE extends num
     countPromptTokens(prompt: LlamaCppTextGenerationPrompt): Promise<number>;
     doGenerateTexts(prompt: LlamaCppTextGenerationPrompt, options?: FunctionOptions): Promise<{
         response: {
+            stop: true;
             model: string;
             prompt: string;
             content: string;
-            stop: true;
             generation_settings: {
+                stop: string[];
                 model: string;
                 stream: boolean;
-                mirostat: number;
                 seed: number;
-                stop: string[];
+                mirostat: number;
                 frequency_penalty: number;
                 ignore_eos: boolean;
                 logit_bias: number[];
@@ -106,7 +106,10 @@ export declare class LlamaCppTextGenerationModel<CONTEXT_WINDOW_SIZE extends num
             tokens_predicted: number;
             truncated: boolean;
         };
-        texts: string[];
+        textGenerationResults: {
+            text: string;
+            finishReason: "length" | "stop" | "unknown";
+        }[];
         usage: {
             promptTokens: number;
             completionTokens: number;
@@ -153,11 +156,11 @@ declare const llamaCppTextGenerationResponseSchema: z.ZodObject<{
         top_p: z.ZodNumber;
         typical_p: z.ZodNumber;
     }, "strip", z.ZodTypeAny, {
+        stop: string[];
         model: string;
         stream: boolean;
-        mirostat: number;
         seed: number;
-        stop: string[];
+        mirostat: number;
         frequency_penalty: number;
         ignore_eos: boolean;
         logit_bias: number[];
@@ -177,11 +180,11 @@ declare const llamaCppTextGenerationResponseSchema: z.ZodObject<{
         top_p: number;
         typical_p: number;
     }, {
+        stop: string[];
         model: string;
         stream: boolean;
-        mirostat: number;
         seed: number;
-        stop: string[];
+        mirostat: number;
         frequency_penalty: number;
         ignore_eos: boolean;
         logit_bias: number[];
@@ -240,16 +243,16 @@ declare const llamaCppTextGenerationResponseSchema: z.ZodObject<{
     tokens_predicted: z.ZodNumber;
     truncated: z.ZodBoolean;
 }, "strip", z.ZodTypeAny, {
+    stop: true;
     model: string;
     prompt: string;
     content: string;
-    stop: true;
     generation_settings: {
+        stop: string[];
         model: string;
         stream: boolean;
-        mirostat: number;
         seed: number;
-        stop: string[];
+        mirostat: number;
         frequency_penalty: number;
         ignore_eos: boolean;
         logit_bias: number[];
@@ -288,16 +291,16 @@ declare const llamaCppTextGenerationResponseSchema: z.ZodObject<{
     tokens_predicted: number;
     truncated: boolean;
 }, {
+    stop: true;
     model: string;
     prompt: string;
     content: string;
-    stop: true;
     generation_settings: {
+        stop: string[];
         model: string;
         stream: boolean;
-        mirostat: number;
         seed: number;
-        stop: string[];
+        mirostat: number;
         frequency_penalty: number;
         ignore_eos: boolean;
         logit_bias: number[];
@@ -353,16 +356,16 @@ export declare const LlamaCppTextGenerationResponseFormat: {
     json: {
         stream: false;
         handler: ResponseHandler<{
+            stop: true;
             model: string;
             prompt: string;
             content: string;
-            stop: true;
             generation_settings: {
+                stop: string[];
                 model: string;
                 stream: boolean;
-                mirostat: number;
                 seed: number;
-                stop: string[];
+                mirostat: number;
                 frequency_penalty: number;
                 ignore_eos: boolean;
                 logit_bias: number[];