Add Amazon Bedrock LLM endpoint integration (#1)

A contribution from Prompt Security

* Kept it as close to the Python implementation as possible
* Followed the guidelines from https://github.com/hwchase17/langchainjs/blob/main/CONTRIBUTING.md and https://github.com/hwchase17/langchainjs/blob/main/.github/contributing/INTEGRATIONS.md
* Supplied with unit test coverage
* Added documentation

docs/extras/modules/model_io/models/llms/integrations/bedrock.mdx

@@ -0,0 +1,16 @@
+# Bedrock
+
+>[Amazon Bedrock](https://aws.amazon.com/bedrock/) is a fully managed service that makes FMs from leading AI startups and Amazon available via an API, so you can choose from a wide range of FMs to find the model that is best suited for your use case.
+
+## Installation and Setup
+
+Install the underlying library from AWS
+```bash npm2yarn
+npm install aws-sigv4-fetch
+```
+
+# LLM example usage
+```
+import CodeBlock from "@theme/CodeBlock";
+import BedrockExample from "@examples/models/llm/bedrock.ts";
+```

examples/src/llms/bedrock.js

@@ -0,0 +1,8 @@
+import { Bedrock } from "langchain/llms/bedrock";
+
+async function test() {
+  const model = new Bedrock({model: "bedrock-model-name", regionName: "aws-region"});
+  const res = await model.call("Question: What would be a good company name a company that makes colorful socks?\nAnswer:");
+  console.log(res);
+}
+test();

examples/src/models/llm/bedrock.ts

@@ -0,0 +1 @@
+// TODO:: add content here

langchain/.gitignore

@@ -121,6 +121,9 @@ llms/googlepalm.d.ts
 llms/sagemaker_endpoint.cjs
 llms/sagemaker_endpoint.js
 llms/sagemaker_endpoint.d.ts
+llms/bedrock.cjs
+llms/bedrock.js
+llms/bedrock.d.ts
 prompts.cjs
 prompts.js
 prompts.d.ts

langchain/package.json

@@ -133,6 +133,9 @@
     "llms/sagemaker_endpoint.cjs",
     "llms/sagemaker_endpoint.js",
     "llms/sagemaker_endpoint.d.ts",
+    "llms/bedrock.cjs",
+    "llms/bedrock.js",
+    "llms/bedrock.d.ts",
     "prompts.cjs",
     "prompts.js",
     "prompts.d.ts",
@@ -583,6 +586,7 @@
     "@xata.io/client": "^0.25.1",
     "@zilliz/milvus2-sdk-node": ">=2.2.7",
     "apify-client": "^2.7.1",
+    "aws-sigv4-fetch": "^2.1.1",
     "axios": "^0.26.0",
     "cheerio": "^1.0.0-rc.12",
     "chromadb": "^1.5.3",
@@ -665,6 +669,7 @@
     "@xata.io/client": "^0.25.1",
     "@zilliz/milvus2-sdk-node": ">=2.2.7",
     "apify-client": "^2.7.1",
+    "aws-sigv4-fetch": "^2.1.1",
     "axios": "*",
     "cheerio": "^1.0.0-rc.12",
     "chromadb": "^1.5.3",
@@ -789,6 +794,9 @@
     "apify-client": {
       "optional": true
     },
+    "aws-sigv4-fetch": {
+      "optional": true
+    },
     "axios": {
       "optional": true
     },
@@ -1141,6 +1149,11 @@
       "import": "./llms/sagemaker_endpoint.js",
       "require": "./llms/sagemaker_endpoint.cjs"
     },
+    "./llms/bedrock": {
+      "types": "./llms/bedrock.d.ts",
+      "import": "./llms/bedrock.js",
+      "require": "./llms/bedrock.cjs"
+    },
     "./prompts": {
       "types": "./prompts.d.ts",
       "import": "./prompts.js",

langchain/scripts/create-entrypoints.js

@@ -55,6 +55,7 @@ const entrypoints = {
   "llms/googlevertexai": "llms/googlevertexai",
   "llms/googlepalm": "llms/googlepalm",
   "llms/sagemaker_endpoint": "llms/sagemaker_endpoint",
+  "llms/bedrock": "llms/bedrock",
   // prompts
   prompts: "prompts/index",
   "prompts/load": "prompts/load",
@@ -243,6 +244,7 @@ const requiresOptionalDependency = [
   "llms/hf",
   "llms/replicate",
   "llms/sagemaker_endpoint",
+  "llms/bedrock",
   "prompts/load",
   "vectorstores/analyticdb",
   "vectorstores/chroma",

langchain/src/llms/bedrock.ts

@@ -0,0 +1,154 @@
+import { getEnvironmentVariable } from "../util/env.js";
+import { LLM, BaseLLMParams } from "./base.js";
+
+type Dict = { [key: string]: any };
+
+class LLMInputOutputAdapter {
+  /** Adapter class to prepare the inputs from Langchain to a format
+  that LLM model expects. Also, provides a helper function to extract
+  the generated text from the model response. */
+
+  static prepareInput(provider: string, prompt: string): Dict {
+    const inputBody: Dict = {};
+
+    if (provider === "anthropic" || provider === "ai21") {
+      inputBody.prompt = prompt;
+    } else if (provider === "amazon") {
+      inputBody.inputText = prompt;
+      inputBody.textGenerationConfig = {};
+    } else {
+      inputBody.inputText = prompt;
+    }
+
+    if (provider === "anthropic" && !("max_tokens_to_sample" in inputBody)) {
+      inputBody.max_tokens_to_sample = 50;
+    }
+
+    return inputBody;
+  }
+
+  static prepareOutput(provider: string, responseBody: any): string {
+    if (provider === "anthropic") {
+      return responseBody.completion;
+    } else if (provider === "ai21") {
+      return responseBody.completions[0].data.text;
+    }
+    return responseBody.results[0].outputText;
+  }
+}
+
+/** Bedrock models.
+    To authenticate, the AWS client uses the following methods to automatically load credentials:
+    https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html
+    If a specific credential profile should be used, you must pass the name of the profile from the ~/.aws/credentials file that is to be used.
+    Make sure the credentials / roles used have the required policies to access the Bedrock service.
+*/
+export interface BedrockInput {
+  /** Model to use.
+      For example, "amazon.titan-tg1-large", this is equivalent to the modelId property in the list-foundation-models api.
+  */
+  model: string;
+
+  /** The AWS region e.g. `us-west-2`.
+      Fallback to AWS_DEFAULT_REGION env variable or region specified in ~/.aws/config in case it is not provided here.
+  */
+  regionName?: string;
+
+  /** Temperature */
+  temperature?: number;
+
+  /** Max tokens */
+  maxTokens?: number;
+}
+
+export class Bedrock extends LLM implements BedrockInput {
+  model = "amazon.titan-tg1-large";
+
+  regionName?: string | undefined = undefined;
+
+  temperature?: number | undefined = undefined;
+
+  maxTokens?: number | undefined = undefined;
+
+  get lc_secrets(): { [key: string]: string } | undefined {
+    return {};
+  }
+
+  _llmType() {
+    return "bedrock";
+  }
+
+  constructor(fields?: Partial<BedrockInput> & BaseLLMParams) {
+    super(fields ?? {});
+
+    this.model = fields?.model ?? this.model;
+    const allowedModels = ["ai21", "anthropic", "amazon"];
+    if (!allowedModels.includes(this.model.split(".")[0])) {
+      throw new Error(
+        `Unknown model: '${this.model}', only these are supported: ${allowedModels}`
+      );
+    }
+    this.regionName =
+      fields?.regionName ?? getEnvironmentVariable("AWS_DEFAULT_REGION");
+    this.temperature = fields?.temperature ?? this.temperature;
+    this.maxTokens = fields?.maxTokens ?? this.maxTokens;
+  }
+
+  /** Call out to Bedrock service model.
+    Arguments:
+      prompt: The prompt to pass into the model.
+
+    Returns:
+      The string generated by the model.
+
+    Example:
+      response = model.call("Tell me a joke.")
+  */
+  async _call(prompt: string): Promise<string> {
+    const { createSignedFetcher } = await Bedrock.imports();
+
+    const signedFetcher = createSignedFetcher({
+      service: "bedrock",
+      region: this.regionName,
+    });
+
+    const url = `https://bedrock.${this.regionName}.amazonaws.com/model/${this.model}/invoke`;
+    const provider = this.model.split(".")[0];
+    const inputBody = LLMInputOutputAdapter.prepareInput(provider, prompt);
+
+    const response = await this.caller.call(
+      async () =>
+        await signedFetcher(url, {
+          method: "post",
+          body: JSON.stringify(inputBody),
+          headers: {
+            "Content-Type": "application/json",
+            accept: "application/json",
+          },
+        })
+    );
+
+    if (response.status < 200 || response.status >= 300) {
+      throw Error(
+        `Failed to access underlying url '${url}': got ${response.status} ${
+          response.statusText
+        }: ${await response.text()}`
+      );
+    }
+    const responseJson = await response.json();
+    const text = LLMInputOutputAdapter.prepareOutput(provider, responseJson);
+    return text;
+  }
+
+  /** @ignore */
+  static async imports(): Promise<{ createSignedFetcher: any }> {
+    try {
+      const { createSignedFetcher } = await import("aws-sigv4-fetch");
+      return { createSignedFetcher };
+    } catch (e) {
+      throw new Error(
+        "Please install a dependency for bedrock with, e.g. `yarn add aws-sigv4-fetch`"
+      );
+    }
+  }
+}