langchain 0.0.175 → 0.0.177

package/dist/llms/googlevertexai/common.cjs

@@ -2,6 +2,7 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.BaseGoogleVertexAI = void 0;
 const base_js_1 = require("../base.cjs");
+const index_js_1 = require("../../schema/index.cjs");
 /**
  * Base class for Google Vertex AI LLMs.
  * Implemented subclasses must provide a GoogleVertexAILLMConnection
@@ -57,6 +58,12 @@ class BaseGoogleVertexAI extends base_js_1.BaseLLM {
             writable: true,
             value: void 0
         });
+        Object.defineProperty(this, "streamedConnection", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
         this.model = fields?.model ?? this.model;
         // Change the defaults for code models
         if (this.model.startsWith("code-gecko")) {
@@ -73,26 +80,37 @@ class BaseGoogleVertexAI extends base_js_1.BaseLLM {
     _llmType() {
         return "vertexai";
     }
+    async *_streamResponseChunks(_input, _options, _runManager) {
+        // Make the call as a streaming request
+        const instance = this.formatInstance(_input);
+        const parameters = this.formatParameters();
+        const result = await this.streamedConnection.request([instance], parameters, _options);
+        // Get the streaming parser of the response
+        const stream = result.data;
+        // Loop until the end of the stream
+        // During the loop, yield each time we get a chunk from the streaming parser
+        // that is either available or added to the queue
+        while (!stream.streamDone) {
+            const output = await stream.nextChunk();
+            const chunk = output !== null
+                ? new index_js_1.GenerationChunk(this.extractGenerationFromPrediction(output.outputs[0]))
+                : new index_js_1.GenerationChunk({
+                    text: "",
+                    generationInfo: { finishReason: "stop" },
+                });
+            yield chunk;
+        }
+    }
     async _generate(prompts, options) {
         const generations = await Promise.all(prompts.map((prompt) => this._generatePrompt(prompt, options)));
         return { generations };
     }
     async _generatePrompt(prompt, options) {
         const instance = this.formatInstance(prompt);
-        const parameters = {
-            temperature: this.temperature,
-            topK: this.topK,
-            topP: this.topP,
-            maxOutputTokens: this.maxOutputTokens,
-        };
+        const parameters = this.formatParameters();
         const result = await this.connection.request([instance], parameters, options);
         const prediction = this.extractPredictionFromResponse(result);
-        return [
-            {
-                text: prediction.content,
-                generationInfo: prediction,
-            },
-        ];
+        return [this.extractGenerationFromPrediction(prediction)];
     }
     /**
      * Formats the input instance as a text instance for the Google Vertex AI
@@ -123,13 +141,28 @@ class BaseGoogleVertexAI extends base_js_1.BaseLLM {
             ? this.formatInstanceCode(prompt)
             : this.formatInstanceText(prompt);
     }
+    formatParameters() {
+        return {
+            temperature: this.temperature,
+            topK: this.topK,
+            topP: this.topP,
+            maxOutputTokens: this.maxOutputTokens,
+        };
+    }
     /**
      * Extracts the prediction from the API response.
      * @param result The API response from which to extract the prediction.
      * @returns A TextPrediction object representing the extracted prediction.
      */
     extractPredictionFromResponse(result) {
-        return result?.data?.predictions[0];
+        return result?.data
+            ?.predictions[0];
+    }
+    extractGenerationFromPrediction(prediction) {
+        return {
+            text: prediction.content,
+            generationInfo: prediction,
+        };
     }
 }
 exports.BaseGoogleVertexAI = BaseGoogleVertexAI;

package/dist/llms/googlevertexai/common.d.ts

@@ -1,8 +1,9 @@
 import { BaseLLM } from "../base.js";
-import { Generation, LLMResult } from "../../schema/index.js";
-import { GoogleVertexAILLMConnection } from "../../util/googlevertexai-connection.js";
-import { GoogleVertexAIBaseLLMInput, GoogleVertexAIBasePrediction, GoogleVertexAILLMResponse } from "../../types/googlevertexai-types.js";
+import { Generation, GenerationChunk, LLMResult } from "../../schema/index.js";
+import { GoogleVertexAILLMConnection, GoogleVertexAILLMResponse } from "../../util/googlevertexai-connection.js";
+import { GoogleVertexAIBaseLLMInput, GoogleVertexAIBasePrediction, GoogleVertexAIModelParams } from "../../types/googlevertexai-types.js";
 import { BaseLanguageModelCallOptions } from "../../base_language/index.js";
+import { CallbackManagerForLLMRun } from "../../callbacks/index.js";
 /**
  * Interface representing the instance of text input to the Google Vertex
  * AI model.
@@ -41,9 +42,11 @@ export declare class BaseGoogleVertexAI<AuthOptions> extends BaseLLM implements
     topP: number;
     topK: number;
     protected connection: GoogleVertexAILLMConnection<BaseLanguageModelCallOptions, GoogleVertexAILLMInstance, TextPrediction, AuthOptions>;
+    protected streamedConnection: GoogleVertexAILLMConnection<BaseLanguageModelCallOptions, GoogleVertexAILLMInstance, TextPrediction, AuthOptions>;
     get lc_aliases(): Record<string, string>;
     constructor(fields?: GoogleVertexAIBaseLLMInput<AuthOptions>);
     _llmType(): string;
+    _streamResponseChunks(_input: string, _options: this["ParsedCallOptions"], _runManager?: CallbackManagerForLLMRun): AsyncGenerator<GenerationChunk>;
     _generate(prompts: string[], options: this["ParsedCallOptions"]): Promise<LLMResult>;
     _generatePrompt(prompt: string, options: this["ParsedCallOptions"]): Promise<Generation[]>;
     /**
@@ -67,11 +70,13 @@ export declare class BaseGoogleVertexAI<AuthOptions> extends BaseLLM implements
      * @returns A GoogleVertexAILLMInstance object representing the formatted instance.
      */
     formatInstance(prompt: string): GoogleVertexAILLMInstance;
+    formatParameters(): GoogleVertexAIModelParams;
     /**
      * Extracts the prediction from the API response.
      * @param result The API response from which to extract the prediction.
      * @returns A TextPrediction object representing the extracted prediction.
      */
     extractPredictionFromResponse(result: GoogleVertexAILLMResponse<TextPrediction>): TextPrediction;
+    extractGenerationFromPrediction(prediction: TextPrediction): Generation;
 }
 export {};

package/dist/llms/googlevertexai/common.js

@@ -1,4 +1,5 @@
 import { BaseLLM } from "../base.js";
+import { GenerationChunk } from "../../schema/index.js";
 /**
  * Base class for Google Vertex AI LLMs.
  * Implemented subclasses must provide a GoogleVertexAILLMConnection
@@ -54,6 +55,12 @@ export class BaseGoogleVertexAI extends BaseLLM {
             writable: true,
             value: void 0
         });
+        Object.defineProperty(this, "streamedConnection", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
         this.model = fields?.model ?? this.model;
         // Change the defaults for code models
         if (this.model.startsWith("code-gecko")) {
@@ -70,26 +77,37 @@ export class BaseGoogleVertexAI extends BaseLLM {
     _llmType() {
         return "vertexai";
     }
+    async *_streamResponseChunks(_input, _options, _runManager) {
+        // Make the call as a streaming request
+        const instance = this.formatInstance(_input);
+        const parameters = this.formatParameters();
+        const result = await this.streamedConnection.request([instance], parameters, _options);
+        // Get the streaming parser of the response
+        const stream = result.data;
+        // Loop until the end of the stream
+        // During the loop, yield each time we get a chunk from the streaming parser
+        // that is either available or added to the queue
+        while (!stream.streamDone) {
+            const output = await stream.nextChunk();
+            const chunk = output !== null
+                ? new GenerationChunk(this.extractGenerationFromPrediction(output.outputs[0]))
+                : new GenerationChunk({
+                    text: "",
+                    generationInfo: { finishReason: "stop" },
+                });
+            yield chunk;
+        }
+    }
     async _generate(prompts, options) {
         const generations = await Promise.all(prompts.map((prompt) => this._generatePrompt(prompt, options)));
         return { generations };
     }
     async _generatePrompt(prompt, options) {
         const instance = this.formatInstance(prompt);
-        const parameters = {
-            temperature: this.temperature,
-            topK: this.topK,
-            topP: this.topP,
-            maxOutputTokens: this.maxOutputTokens,
-        };
+        const parameters = this.formatParameters();
         const result = await this.connection.request([instance], parameters, options);
         const prediction = this.extractPredictionFromResponse(result);
-        return [
-            {
-                text: prediction.content,
-                generationInfo: prediction,
-            },
-        ];
+        return [this.extractGenerationFromPrediction(prediction)];
     }
     /**
      * Formats the input instance as a text instance for the Google Vertex AI
@@ -120,12 +138,27 @@ export class BaseGoogleVertexAI extends BaseLLM {
             ? this.formatInstanceCode(prompt)
             : this.formatInstanceText(prompt);
     }
+    formatParameters() {
+        return {
+            temperature: this.temperature,
+            topK: this.topK,
+            topP: this.topP,
+            maxOutputTokens: this.maxOutputTokens,
+        };
+    }
     /**
      * Extracts the prediction from the API response.
      * @param result The API response from which to extract the prediction.
      * @returns A TextPrediction object representing the extracted prediction.
      */
     extractPredictionFromResponse(result) {
-        return result?.data?.predictions[0];
+        return result?.data
+            ?.predictions[0];
+    }
+    extractGenerationFromPrediction(prediction) {
+        return {
+            text: prediction.content,
+            generationInfo: prediction,
+        };
     }
 }

package/dist/llms/googlevertexai/index.cjs

@@ -1,9 +1,9 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.GoogleVertexAI = void 0;
-const google_auth_library_1 = require("google-auth-library");
 const googlevertexai_connection_js_1 = require("../../util/googlevertexai-connection.cjs");
 const common_js_1 = require("./common.cjs");
+const googlevertexai_gauth_js_1 = require("../../util/googlevertexai-gauth.cjs");
 /**
  * Enables calls to the Google Cloud's Vertex AI API to access
  * Large Language Models.
@@ -24,11 +24,12 @@ class GoogleVertexAI extends common_js_1.BaseGoogleVertexAI {
     }
     constructor(fields) {
         super(fields);
-        const client = new google_auth_library_1.GoogleAuth({
+        const client = new googlevertexai_gauth_js_1.GAuthClient({
             scopes: "https://www.googleapis.com/auth/cloud-platform",
             ...fields?.authOptions,
         });
-        this.connection = new googlevertexai_connection_js_1.GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client);
+        this.connection = new googlevertexai_connection_js_1.GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client, false);
+        this.streamedConnection = new googlevertexai_connection_js_1.GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client, true);
     }
 }
 exports.GoogleVertexAI = GoogleVertexAI;

package/dist/llms/googlevertexai/index.js

@@ -1,6 +1,6 @@
-import { GoogleAuth } from "google-auth-library";
 import { GoogleVertexAILLMConnection } from "../../util/googlevertexai-connection.js";
 import { BaseGoogleVertexAI } from "./common.js";
+import { GAuthClient } from "../../util/googlevertexai-gauth.js";
 /**
  * Enables calls to the Google Cloud's Vertex AI API to access
  * Large Language Models.
@@ -21,10 +21,11 @@ export class GoogleVertexAI extends BaseGoogleVertexAI {
     }
     constructor(fields) {
         super(fields);
-        const client = new GoogleAuth({
+        const client = new GAuthClient({
             scopes: "https://www.googleapis.com/auth/cloud-platform",
             ...fields?.authOptions,
         });
-        this.connection = new GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client);
+        this.connection = new GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client, false);
+        this.streamedConnection = new GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client, true);
     }
 }

package/dist/llms/googlevertexai/web.cjs

@@ -25,7 +25,8 @@ class GoogleVertexAI extends common_js_1.BaseGoogleVertexAI {
     constructor(fields) {
         super(fields);
         const client = new googlevertexai_webauth_js_1.WebGoogleAuth(fields?.authOptions);
-        this.connection = new googlevertexai_connection_js_1.GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client);
+        this.connection = new googlevertexai_connection_js_1.GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client, false);
+        this.streamedConnection = new googlevertexai_connection_js_1.GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client, true);
     }
 }
 exports.GoogleVertexAI = GoogleVertexAI;

package/dist/llms/googlevertexai/web.js

@@ -22,6 +22,7 @@ export class GoogleVertexAI extends BaseGoogleVertexAI {
     constructor(fields) {
         super(fields);
         const client = new WebGoogleAuth(fields?.authOptions);
-        this.connection = new GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client);
+        this.connection = new GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client, false);
+        this.streamedConnection = new GoogleVertexAILLMConnection({ ...fields, ...this }, this.caller, client, true);
     }
 }

package/dist/llms/hf.cjs

@@ -57,6 +57,12 @@ class HuggingFaceInference extends base_js_1.LLM {
             writable: true,
             value: undefined
         });
+        Object.defineProperty(this, "endpointUrl", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: undefined
+        });
         this.model = fields?.model ?? this.model;
         this.temperature = fields?.temperature ?? this.temperature;
         this.maxTokens = fields?.maxTokens ?? this.maxTokens;
@@ -65,6 +71,7 @@ class HuggingFaceInference extends base_js_1.LLM {
         this.frequencyPenalty = fields?.frequencyPenalty ?? this.frequencyPenalty;
         this.apiKey =
             fields?.apiKey ?? (0, env_js_1.getEnvironmentVariable)("HUGGINGFACEHUB_API_KEY");
+        this.endpointUrl = fields?.endpointUrl;
         if (!this.apiKey) {
             throw new Error("Please set an API key for HuggingFace Hub in the environment variable HUGGINGFACEHUB_API_KEY or in the apiKey field of the HuggingFaceInference constructor.");
         }
@@ -75,7 +82,9 @@ class HuggingFaceInference extends base_js_1.LLM {
     /** @ignore */
     async _call(prompt, options) {
         const { HfInference } = await HuggingFaceInference.imports();
-        const hf = new HfInference(this.apiKey);
+        const hf = this.endpointUrl
+            ? new HfInference(this.apiKey).endpoint(this.endpointUrl)
+            : new HfInference(this.apiKey);
         const res = await this.caller.callWithOptions({ signal: options.signal }, hf.textGeneration.bind(hf), {
             model: this.model,
             parameters: {

package/dist/llms/hf.d.ts

@@ -6,6 +6,8 @@ import { LLM, BaseLLMParams } from "./base.js";
 export interface HFInput {
     /** Model to use */
     model: string;
+    /** Custom inference endpoint URL to use */
+    endpointUrl?: string;
     /** Sampling temperature to use */
     temperature?: number;
     /**
@@ -36,6 +38,7 @@ export declare class HuggingFaceInference extends LLM implements HFInput {
     topK: number | undefined;
     frequencyPenalty: number | undefined;
     apiKey: string | undefined;
+    endpointUrl: string | undefined;
     constructor(fields?: Partial<HFInput> & BaseLLMParams);
     _llmType(): string;
     /** @ignore */

package/dist/llms/hf.js

@@ -54,6 +54,12 @@ export class HuggingFaceInference extends LLM {
             writable: true,
             value: undefined
         });
+        Object.defineProperty(this, "endpointUrl", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: undefined
+        });
         this.model = fields?.model ?? this.model;
         this.temperature = fields?.temperature ?? this.temperature;
         this.maxTokens = fields?.maxTokens ?? this.maxTokens;
@@ -62,6 +68,7 @@ export class HuggingFaceInference extends LLM {
         this.frequencyPenalty = fields?.frequencyPenalty ?? this.frequencyPenalty;
         this.apiKey =
             fields?.apiKey ?? getEnvironmentVariable("HUGGINGFACEHUB_API_KEY");
+        this.endpointUrl = fields?.endpointUrl;
         if (!this.apiKey) {
             throw new Error("Please set an API key for HuggingFace Hub in the environment variable HUGGINGFACEHUB_API_KEY or in the apiKey field of the HuggingFaceInference constructor.");
         }
@@ -72,7 +79,9 @@ export class HuggingFaceInference extends LLM {
     /** @ignore */
     async _call(prompt, options) {
         const { HfInference } = await HuggingFaceInference.imports();
-        const hf = new HfInference(this.apiKey);
+        const hf = this.endpointUrl
+            ? new HfInference(this.apiKey).endpoint(this.endpointUrl)
+            : new HfInference(this.apiKey);
         const res = await this.caller.callWithOptions({ signal: options.signal }, hf.textGeneration.bind(hf), {
             model: this.model,
             parameters: {

package/dist/llms/llama_cpp.cjs

@@ -1,7 +1,7 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.LlamaCpp = void 0;
-const node_llama_cpp_1 = require("node-llama-cpp");
+const llama_cpp_js_1 = require("../util/llama_cpp.cjs");
 const base_js_1 = require("./base.cjs");
 /**
  *  To use this model you need to have the `node-llama-cpp` module installed.
@@ -15,73 +15,31 @@ class LlamaCpp extends base_js_1.LLM {
     }
     constructor(inputs) {
         super(inputs);
-        Object.defineProperty(this, "batchSize", {
+        Object.defineProperty(this, "maxTokens", {
             enumerable: true,
             configurable: true,
             writable: true,
             value: void 0
         });
-        Object.defineProperty(this, "contextSize", {
+        Object.defineProperty(this, "temperature", {
             enumerable: true,
             configurable: true,
             writable: true,
             value: void 0
         });
-        Object.defineProperty(this, "embedding", {
+        Object.defineProperty(this, "topK", {
             enumerable: true,
             configurable: true,
             writable: true,
             value: void 0
         });
-        Object.defineProperty(this, "f16Kv", {
+        Object.defineProperty(this, "topP", {
             enumerable: true,
             configurable: true,
             writable: true,
             value: void 0
         });
-        Object.defineProperty(this, "gpuLayers", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "logitsAll", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "lowVram", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "seed", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "useMlock", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "useMmap", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "vocabOnly", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "modelPath", {
+        Object.defineProperty(this, "trimWhitespaceSuffix", {
             enumerable: true,
             configurable: true,
             writable: true,
@@ -105,29 +63,31 @@ class LlamaCpp extends base_js_1.LLM {
             writable: true,
             value: void 0
         });
-        this.batchSize = inputs.batchSize;
-        this.contextSize = inputs.contextSize;
-        this.embedding = inputs.embedding;
-        this.f16Kv = inputs.f16Kv;
-        this.gpuLayers = inputs.gpuLayers;
-        this.logitsAll = inputs.logitsAll;
-        this.lowVram = inputs.lowVram;
-        this.modelPath = inputs.modelPath;
-        this.seed = inputs.seed;
-        this.useMlock = inputs.useMlock;
-        this.useMmap = inputs.useMmap;
-        this.vocabOnly = inputs.vocabOnly;
-        this._model = new node_llama_cpp_1.LlamaModel(inputs);
-        this._context = new node_llama_cpp_1.LlamaContext({ model: this._model });
-        this._session = new node_llama_cpp_1.LlamaChatSession({ context: this._context });
+        this.maxTokens = inputs?.maxTokens;
+        this.temperature = inputs?.temperature;
+        this.topK = inputs?.topK;
+        this.topP = inputs?.topP;
+        this.trimWhitespaceSuffix = inputs?.trimWhitespaceSuffix;
+        this._model = (0, llama_cpp_js_1.createLlamaModel)(inputs);
+        this._context = (0, llama_cpp_js_1.createLlamaContext)(this._model, inputs);
+        this._session = (0, llama_cpp_js_1.createLlamaSession)(this._context);
     }
     _llmType() {
         return "llama2_cpp";
     }
     /** @ignore */
-    async _call(prompt, options) {
+    async _call(prompt, 
+    // @ts-expect-error - TS6133: 'options' is declared but its value is never read.
+    options) {
         try {
-            const completion = await this._session.prompt(prompt, options);
+            const promptOptions = {
+                maxTokens: this?.maxTokens,
+                temperature: this?.temperature,
+                topK: this?.topK,
+                topP: this?.topP,
+                trimWhitespaceSuffix: this?.trimWhitespaceSuffix,
+            };
+            const completion = await this._session.prompt(prompt, promptOptions);
             return completion;
         }
         catch (e) {

package/dist/llms/llama_cpp.d.ts

@@ -1,40 +1,11 @@
 import { LlamaModel, LlamaContext, LlamaChatSession } from "node-llama-cpp";
+import { LlamaBaseCppInputs } from "../util/llama_cpp.js";
 import { LLM, BaseLLMCallOptions, BaseLLMParams } from "./base.js";
 /**
  * Note that the modelPath is the only required parameter. For testing you
  * can set this in the environment variable `LLAMA_PATH`.
  */
-export interface LlamaCppInputs extends BaseLLMParams {
-    /** Prompt processing batch size. */
-    batchSize?: number;
-    /** Text context size. */
-    contextSize?: number;
-    /** Embedding mode only. */
-    embedding?: boolean;
-    /** Use fp16 for KV cache. */
-    f16Kv?: boolean;
-    /** Number of layers to store in VRAM. */
-    gpuLayers?: number;
-    /** The llama_eval() call computes all logits, not just the last one. */
-    logitsAll?: boolean;
-    /** If true, reduce VRAM usage at the cost of performance. */
-    lowVram?: boolean;
-    /** Path to the model on the filesystem. */
-    modelPath: string;
-    /** If null, a random seed will be used. */
-    seed?: null | number;
-    /** The randomness of the responses, e.g. 0.1 deterministic, 1.5 creative, 0.8 balanced, 0 disables. */
-    temperature?: number;
-    /** Consider the n most likely tokens, where n is 1 to vocabulary size, 0 disables (uses full vocabulary). Note: only applies when `temperature` > 0. */
-    topK?: number;
-    /** Selects the smallest token set whose probability exceeds P, where P is between 0 - 1, 1 disables. Note: only applies when `temperature` > 0. */
-    topP?: number;
-    /** Force system to keep model in RAM. */
-    useMlock?: boolean;
-    /** Use mmap if possible. */
-    useMmap?: boolean;
-    /** Only load the vocabulary, no weights. */
-    vocabOnly?: boolean;
+export interface LlamaCppInputs extends LlamaBaseCppInputs, BaseLLMParams {
 }
 export interface LlamaCppCallOptions extends BaseLLMCallOptions {
     /** The maximum number of tokens the response should contain. */
@@ -51,18 +22,11 @@ export interface LlamaCppCallOptions extends BaseLLMCallOptions {
 export declare class LlamaCpp extends LLM<LlamaCppCallOptions> {
     CallOptions: LlamaCppCallOptions;
     static inputs: LlamaCppInputs;
-    batchSize?: number;
-    contextSize?: number;
-    embedding?: boolean;
-    f16Kv?: boolean;
-    gpuLayers?: number;
-    logitsAll?: boolean;
-    lowVram?: boolean;
-    seed?: null | number;
-    useMlock?: boolean;
-    useMmap?: boolean;
-    vocabOnly?: boolean;
-    modelPath: string;
+    maxTokens?: number;
+    temperature?: number;
+    topK?: number;
+    topP?: number;
+    trimWhitespaceSuffix?: boolean;
     _model: LlamaModel;
     _context: LlamaContext;
     _session: LlamaChatSession;