langchain 0.0.176 → 0.0.177

package/dist/chat_models/bedrock.cjs

@@ -133,6 +133,12 @@ class ChatBedrock extends base_js_1.SimpleChatModel {
             writable: true,
             value: new eventstream_codec_1.EventStreamCodec(util_utf8_1.toUtf8, util_utf8_1.fromUtf8)
         });
+        Object.defineProperty(this, "streaming", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: false
+        });
         this.model = fields?.model ?? this.model;
         const allowedModels = ["ai21", "anthropic", "amazon"];
         if (!allowedModels.includes(this.model.split(".")[0])) {
@@ -150,6 +156,7 @@ class ChatBedrock extends base_js_1.SimpleChatModel {
         this.endpointHost = fields?.endpointHost ?? fields?.endpointUrl;
         this.stopSequences = fields?.stopSequences;
         this.modelKwargs = fields?.modelKwargs;
+        this.streaming = fields?.streaming ?? this.streaming;
     }
     /** Call out to Bedrock service model.
       Arguments:
@@ -161,10 +168,23 @@ class ChatBedrock extends base_js_1.SimpleChatModel {
       Example:
         response = model.call("Tell me a joke.")
     */
-    async _call(messages, options) {
+    async _call(messages, options, runManager) {
         const service = "bedrock-runtime";
         const endpointHost = this.endpointHost ?? `${service}.${this.region}.amazonaws.com`;
         const provider = this.model.split(".")[0];
+        if (this.streaming) {
+            const stream = this._streamResponseChunks(messages, options, runManager);
+            let finalResult;
+            for await (const chunk of stream) {
+                if (finalResult === undefined) {
+                    finalResult = chunk;
+                }
+                else {
+                    finalResult = finalResult.concat(chunk);
+                }
+            }
+            return finalResult?.message.content ?? "";
+        }
         const response = await this._signedFetch(messages, options, {
             bedrockMethod: "invoke",
             endpointHost,
@@ -233,7 +253,6 @@ class ChatBedrock extends base_js_1.SimpleChatModel {
                     event.headers[":content-type"].value !== "application/json") {
                     throw Error(`Failed to get event chunk: got ${chunk}`);
                 }
-                // console.log(decoder.decode(event.body));
                 const body = JSON.parse(decoder.decode(event.body));
                 if (body.message) {
                     throw new Error(body.message);
@@ -245,7 +264,8 @@ class ChatBedrock extends base_js_1.SimpleChatModel {
                         text,
                         message: new index_js_1.AIMessageChunk({ content: text }),
                     });
-                    await runManager?.handleLLMNewToken(text);
+                    // eslint-disable-next-line no-void
+                    void runManager?.handleLLMNewToken(text);
                 }
             }
         }
@@ -256,7 +276,8 @@ class ChatBedrock extends base_js_1.SimpleChatModel {
                 text,
                 message: new index_js_1.AIMessageChunk({ content: text }),
             });
-            await runManager?.handleLLMNewToken(text);
+            // eslint-disable-next-line no-void
+            void runManager?.handleLLMNewToken(text);
         }
     }
     // eslint-disable-next-line @typescript-eslint/no-explicit-any

package/dist/chat_models/bedrock.d.ts

@@ -33,6 +33,7 @@ export declare class ChatBedrock extends SimpleChatModel implements BaseBedrockI
     stopSequences?: string[];
     modelKwargs?: Record<string, unknown>;
     codec: EventStreamCodec;
+    streaming: boolean;
     get lc_secrets(): {
         [key: string]: string;
     } | undefined;
@@ -49,7 +50,7 @@ export declare class ChatBedrock extends SimpleChatModel implements BaseBedrockI
       Example:
         response = model.call("Tell me a joke.")
     */
-    _call(messages: BaseMessage[], options: this["ParsedCallOptions"]): Promise<string>;
+    _call(messages: BaseMessage[], options: this["ParsedCallOptions"], runManager?: CallbackManagerForLLMRun): Promise<string>;
     _signedFetch(messages: BaseMessage[], options: this["ParsedCallOptions"], fields: {
         bedrockMethod: "invoke" | "invoke-with-response-stream";
         endpointHost: string;

package/dist/chat_models/bedrock.js

@@ -128,6 +128,12 @@ export class ChatBedrock extends SimpleChatModel {
             writable: true,
             value: new EventStreamCodec(toUtf8, fromUtf8)
         });
+        Object.defineProperty(this, "streaming", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: false
+        });
         this.model = fields?.model ?? this.model;
         const allowedModels = ["ai21", "anthropic", "amazon"];
         if (!allowedModels.includes(this.model.split(".")[0])) {
@@ -145,6 +151,7 @@ export class ChatBedrock extends SimpleChatModel {
         this.endpointHost = fields?.endpointHost ?? fields?.endpointUrl;
         this.stopSequences = fields?.stopSequences;
         this.modelKwargs = fields?.modelKwargs;
+        this.streaming = fields?.streaming ?? this.streaming;
     }
     /** Call out to Bedrock service model.
       Arguments:
@@ -156,10 +163,23 @@ export class ChatBedrock extends SimpleChatModel {
       Example:
         response = model.call("Tell me a joke.")
     */
-    async _call(messages, options) {
+    async _call(messages, options, runManager) {
         const service = "bedrock-runtime";
         const endpointHost = this.endpointHost ?? `${service}.${this.region}.amazonaws.com`;
         const provider = this.model.split(".")[0];
+        if (this.streaming) {
+            const stream = this._streamResponseChunks(messages, options, runManager);
+            let finalResult;
+            for await (const chunk of stream) {
+                if (finalResult === undefined) {
+                    finalResult = chunk;
+                }
+                else {
+                    finalResult = finalResult.concat(chunk);
+                }
+            }
+            return finalResult?.message.content ?? "";
+        }
         const response = await this._signedFetch(messages, options, {
             bedrockMethod: "invoke",
             endpointHost,
@@ -228,7 +248,6 @@ export class ChatBedrock extends SimpleChatModel {
                     event.headers[":content-type"].value !== "application/json") {
                     throw Error(`Failed to get event chunk: got ${chunk}`);
                 }
-                // console.log(decoder.decode(event.body));
                 const body = JSON.parse(decoder.decode(event.body));
                 if (body.message) {
                     throw new Error(body.message);
@@ -240,7 +259,8 @@ export class ChatBedrock extends SimpleChatModel {
                         text,
                         message: new AIMessageChunk({ content: text }),
                     });
-                    await runManager?.handleLLMNewToken(text);
+                    // eslint-disable-next-line no-void
+                    void runManager?.handleLLMNewToken(text);
                 }
             }
         }
@@ -251,7 +271,8 @@ export class ChatBedrock extends SimpleChatModel {
                 text,
                 message: new AIMessageChunk({ content: text }),
             });
-            await runManager?.handleLLMNewToken(text);
+            // eslint-disable-next-line no-void
+            void runManager?.handleLLMNewToken(text);
         }
     }
     // eslint-disable-next-line @typescript-eslint/no-explicit-any

package/dist/chat_models/llama_cpp.cjs

@@ -3,6 +3,7 @@ Object.defineProperty(exports, "__esModule", { value: true });
 exports.ChatLlamaCpp = void 0;
 const node_llama_cpp_1 = require("node-llama-cpp");
 const base_js_1 = require("./base.cjs");
+const llama_cpp_js_1 = require("../util/llama_cpp.cjs");
 /**
  *  To use this model you need to have the `node-llama-cpp` module installed.
  *  This can be installed using `npm install -S node-llama-cpp` and the minimum
@@ -15,73 +16,31 @@ class ChatLlamaCpp extends base_js_1.SimpleChatModel {
     }
     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,
@@ -105,47 +64,33 @@ class ChatLlamaCpp extends base_js_1.SimpleChatModel {
             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.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 = null;
     }
     _llmType() {
         return "llama2_cpp";
     }
-    invocationParams() {
-        return {
-            batchSize: this.batchSize,
-            contextSize: this.contextSize,
-            embedding: this.embedding,
-            f16Kv: this.f16Kv,
-            gpuLayers: this.gpuLayers,
-            logitsAll: this.logitsAll,
-            lowVram: this.lowVram,
-            modelPath: this.modelPath,
-            seed: this.seed,
-            useMlock: this.useMlock,
-            useMmap: this.useMmap,
-            vocabOnly: this.vocabOnly,
-        };
-    }
     /** @ignore */
     _combineLLMOutput() {
         return {};
     }
+    invocationParams() {
+        return {
+            maxTokens: this.maxTokens,
+            temperature: this.temperature,
+            topK: this.topK,
+            topP: this.topP,
+            trimWhitespaceSuffix: this.trimWhitespaceSuffix,
+        };
+    }
     /** @ignore */
-    async _call(messages, options) {
+    async _call(messages, _options) {
         let prompt = "";
         if (messages.length > 1) {
             // We need to build a new _session
@@ -159,8 +104,15 @@ class ChatLlamaCpp extends base_js_1.SimpleChatModel {
             prompt = messages[0].content;
         }
         try {
+            const promptOptions = {
+                maxTokens: this?.maxTokens,
+                temperature: this?.temperature,
+                topK: this?.topK,
+                topP: this?.topP,
+                trimWhitespaceSuffix: this?.trimWhitespaceSuffix,
+            };
             // @ts-expect-error - TS2531: Object is possibly 'null'.
-            const completion = await this._session.prompt(prompt, options);
+            const completion = await this._session.prompt(prompt, promptOptions);
             return completion;
         }
         catch (e) {

package/dist/chat_models/llama_cpp.d.ts

@@ -1,42 +1,13 @@
 import { LlamaModel, LlamaContext, LlamaChatSession, type ConversationInteraction } from "node-llama-cpp";
 import { SimpleChatModel, BaseChatModelParams } from "./base.js";
+import { LlamaBaseCppInputs } from "../util/llama_cpp.js";
 import { BaseLanguageModelCallOptions } from "../base_language/index.js";
 import type { BaseMessage } from "../schema/index.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 BaseChatModelParams {
-    /** 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, BaseChatModelParams {
 }
 export interface LlamaCppCallOptions extends BaseLanguageModelCallOptions {
     /** The maximum number of tokens the response should contain. */
@@ -53,42 +24,28 @@ export interface LlamaCppCallOptions extends BaseLanguageModelCallOptions {
 export declare class ChatLlamaCpp extends SimpleChatModel<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 | null;
     static lc_name(): string;
     constructor(inputs: LlamaCppInputs);
     _llmType(): string;
-    invocationParams(): {
-        batchSize: number | undefined;
-        contextSize: number | undefined;
-        embedding: boolean | undefined;
-        f16Kv: boolean | undefined;
-        gpuLayers: number | undefined;
-        logitsAll: boolean | undefined;
-        lowVram: boolean | undefined;
-        modelPath: string;
-        seed: number | null | undefined;
-        useMlock: boolean | undefined;
-        useMmap: boolean | undefined;
-        vocabOnly: boolean | undefined;
-    };
     /** @ignore */
     _combineLLMOutput(): {};
+    invocationParams(): {
+        maxTokens: number | undefined;
+        temperature: number | undefined;
+        topK: number | undefined;
+        topP: number | undefined;
+        trimWhitespaceSuffix: boolean | undefined;
+    };
     /** @ignore */
-    _call(messages: BaseMessage[], options: this["ParsedCallOptions"]): Promise<string>;
+    _call(messages: BaseMessage[], _options: this["ParsedCallOptions"]): Promise<string>;
     protected _buildSession(messages: BaseMessage[]): string;
     protected _convertMessagesToInteractions(messages: BaseMessage[]): ConversationInteraction[];
 }

package/dist/chat_models/llama_cpp.js

@@ -1,5 +1,6 @@
-import { LlamaModel, LlamaContext, LlamaChatSession, } from "node-llama-cpp";
+import { LlamaChatSession, } from "node-llama-cpp";
 import { SimpleChatModel } from "./base.js";
+import { createLlamaModel, createLlamaContext, } from "../util/llama_cpp.js";
 /**
  *  To use this model you need to have the `node-llama-cpp` module installed.
  *  This can be installed using `npm install -S node-llama-cpp` and the minimum
@@ -12,73 +13,31 @@ export class ChatLlamaCpp extends SimpleChatModel {
     }
     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,
@@ -102,47 +61,33 @@ export class ChatLlamaCpp extends SimpleChatModel {
             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 LlamaModel(inputs);
-        this._context = new LlamaContext({ model: this._model });
+        this.maxTokens = inputs?.maxTokens;
+        this.temperature = inputs?.temperature;
+        this.topK = inputs?.topK;
+        this.topP = inputs?.topP;
+        this.trimWhitespaceSuffix = inputs?.trimWhitespaceSuffix;
+        this._model = createLlamaModel(inputs);
+        this._context = createLlamaContext(this._model, inputs);
         this._session = null;
     }
     _llmType() {
         return "llama2_cpp";
     }
-    invocationParams() {
-        return {
-            batchSize: this.batchSize,
-            contextSize: this.contextSize,
-            embedding: this.embedding,
-            f16Kv: this.f16Kv,
-            gpuLayers: this.gpuLayers,
-            logitsAll: this.logitsAll,
-            lowVram: this.lowVram,
-            modelPath: this.modelPath,
-            seed: this.seed,
-            useMlock: this.useMlock,
-            useMmap: this.useMmap,
-            vocabOnly: this.vocabOnly,
-        };
-    }
     /** @ignore */
     _combineLLMOutput() {
         return {};
     }
+    invocationParams() {
+        return {
+            maxTokens: this.maxTokens,
+            temperature: this.temperature,
+            topK: this.topK,
+            topP: this.topP,
+            trimWhitespaceSuffix: this.trimWhitespaceSuffix,
+        };
+    }
     /** @ignore */
-    async _call(messages, options) {
+    async _call(messages, _options) {
         let prompt = "";
         if (messages.length > 1) {
             // We need to build a new _session
@@ -156,8 +101,15 @@ export class ChatLlamaCpp extends SimpleChatModel {
             prompt = messages[0].content;
         }
         try {
+            const promptOptions = {
+                maxTokens: this?.maxTokens,
+                temperature: this?.temperature,
+                topK: this?.topK,
+                topP: this?.topP,
+                trimWhitespaceSuffix: this?.trimWhitespaceSuffix,
+            };
             // @ts-expect-error - TS2531: Object is possibly 'null'.
-            const completion = await this._session.prompt(prompt, options);
+            const completion = await this._session.prompt(prompt, promptOptions);
             return completion;
         }
         catch (e) {