feat(community): Langchainjs oci genai integration

docs/core_docs/docs/integrations/chat/oci_genai.mdx

@@ -0,0 +1,263 @@
+---
+title: Oracle Cloud Infrastructure Generative AI
+---
+
+Oracle Cloud Infrastructure (OCI) Generative AI is a fully managed
+service that provides a set of state-of-the-art, customizable large
+language models (LLMs) that cover a wide range of use cases, and which
+is available through a single API. Using the OCI Generative AI service
+you can access ready-to-use pretrained models, or create and host your
+own fine-tuned custom models based on your own data on dedicated AI
+clusters. Detailed documentation of the service and API is available
+[here](https://docs.oracle.com/en-us/iaas/Content/generative-ai/home.htm)
+and
+[here](https://docs.oracle.com/en-us/iaas/api/#/en/generative-ai/20231130/).
+
+This notebook explains how to use OCI’s Genrative AI models with
+LangChainJS.
+
+## Prerequisites
+
+In order to use this integration you will need the following: 1. An OCI
+tenancy. If you do not already have and account, please create one
+[here](https://signup.cloud.oracle.com?sourceType=:ex:of:::::LangChainJSIntegration&SC=:ex:of:::::LangChainJSIntegration&pcode=). 2. Setup an [authentication
+method](https://docs.oracle.com/en-us/iaas/Content/API/Concepts/sdk_authentication_methods.htm)
+(Using a [configuration
+file](https://docs.oracle.com/en-us/iaas/Content/API/Concepts/sdkconfig.htm)
+with [API Key
+authentication](https://docs.oracle.com/en-us/iaas/Content/API/Concepts/apisigningkey.htm#apisigningkey_topic_How_to_Generate_an_API_Signing_Key_Console)
+is the simplest to start with). 3. Please make sure that your OCI
+tenancy is registered in one of the [supported
+regions](https://docs.oracle.com/en-us/iaas/Content/generative-ai/overview.htm#regions). 4. You will need the ID (aka OCID) of a compartment in which your OCI
+user has [access to use the Generative AI
+service](https://docs.oracle.com/en-us/iaas/Content/generative-ai/iam-policies.htm).
+You can either use the `root` compartment or [create your
+own](https://docs.oracle.com/en-us/iaas/Content/Identity/compartments/To_create_a_compartment.htm). 5. Retrieve the desired model name from the [available
+models](https://docs.oracle.com/en-us/iaas/Content/generative-ai/pretrained-models.htm)
+list (please make sure not to select a deprecated model).
+
+## Installation
+
+The integration makes use of the [OCI TypeScript
+SDK](https://docs.oracle.com/en-us/iaas/Content/API/SDKDocs/typescriptsdk.htm).
+To install the integration dependencies, execute the following:
+
+```python
+npm install oci-common oci-generativeaiinference
+```
+
+## Instantiation
+
+The OCI Generative AI service supports two groups of LLMs: 1. Cohere
+family of LLMs. 2. Generic family of LLMs which include model such as
+Llama.
+
+The following code demonstrates how to create an instance for each of
+the families. The only mandatory two parameters are: 1.
+`compartmentId` - A compartment OCID in which the user you are using for
+authentication was granted permissions to access the Generative AI
+service. 2. `onDemandModelId` or `dedicatedEndpointId` - Either a
+[pre-trained
+model](https://docs.oracle.com/en-us/iaas/Content/generative-ai/pretrained-models.htm)
+name/OCID or a dedicated endpoint OCID for an endpoint configured on a
+[dedicated AI cluster
+(DAC)](https://docs.oracle.com/en-us/iaas/Content/generative-ai/ai-cluster.htm).
+Either `onDemandModelId` or `dedicatedEndpointId` must be provided but
+not both.
+
+In this example, since no other parameters are specified, a default SDK
+client will be created with the following configuration: 1.
+Authentication will be attempted using a [configuration
+file](https://docs.oracle.com/en-us/iaas/Content/API/Concepts/sdkconfig.htm)
+which should be already setup and available under `~/.oci/config`. The
+`config` file is expected to contain a `DEFAULT` profile with the
+correct information. Please see the prerequisites for more information. 2. The retry strategy will be set to a single attempt. If the first API
+call was not successful, the request will fail. 3. The region will be
+set to `us-chicago-1`. Please make sure that your tenancy is registered
+this region.
+
+```python
+import { OciGenAiCohereChat } from "@langchain/community/chat_models/oci_genai/cohere_chat";
+import { OciGenAiGenericChat } from "@langchain/community/chat_models/oci_genai/generic_chat";
+
+const cohereLlm = new OciGenAiCohereChat({
+  compartmentId: "oci.compartment...",
+  onDemandModelId: "cohere.command-r-plus-08-2024",
+  // dedicatedEndpointId: "oci.dedicatedendpoint..."
+});
+
+const genericLlm = new OciGenAiGenericChat({
+  compartmentId: "oci.compartment...",
+  onDemandModelId: "meta.llama-3.3-70b-instruct",
+  // dedicatedEndpointId: "oci.dedicatedendpoint..."
+});
+```
+
+## SDK client options
+
+The above example used default values to create the SDK client behind
+the scenes. If you need more control in the creation of the client, here
+are additional options (the options are the same for
+`OciGenAiCohereChat` and `OciGenAiGenericChat`).
+
+The first example will create an SDK client with the following
+configuration: 1. [Instance Principal
+authentication](https://docs.oracle.com/en-us/iaas/Content/API/Concepts/sdk_authentication_methods.htm#sdk_authentication_methods_instance_principaldita).
+Please note that this authentication method requires
+[configuration](https://docs.oracle.com/en-us/iaas/Content/Identity/Tasks/callingservicesfrominstances.htm). 2. Using the Sao Paulo region. 3. Up to 3 attempts will be made in case
+API calls fail.
+
+```python
+import { MaxAttemptsTerminationStrategy, Region } from "oci-common"
+import { OciGenAiCohereChat } from "@langchain/community/chat_models/oci_genai/cohere_chat";
+import { OciGenAiNewClientAuthType } from "@langchain/community/chat_models/oci_genai/types";
+
+const cohereLlm = new OciGenAiCohereChat({
+  compartmentId: "oci.compartment...",
+  onDemandModelId: "cohere.command-r-plus-08-2024",
+  newClientParams: {
+    authType: OciGenAiNewClientAuthType.InstancePrincipal,
+    regionId: Region.SA_SAOPAULO_1.regionId,
+    clientConfiguration: {
+      retryConfiguration: {
+        terminationStrategy: new MaxAttemptsTerminationStrategy(3)
+      }
+    }
+  }
+});
+```
+
+The second example will create an SDK client with the following
+configuration: 1. Config file authentication. 1. Use the config file:
+`/my/path/config`. 1. Use the details under the
+`MY_PROFILE_IN_CONFIG_FILE` profile in the specified config file. 1. The
+retry strategy will be set to a single attempt. If the first API call
+was not successful, the request will fail. 1. The region will be set to
+`us-chicago-1`. Please make sure that your tenancy is registered this
+region.
+
+```python
+import { OciGenAiCohereChat } from "@langchain/community/chat_models/oci_genai/cohere_chat";
+import { OciGenAiNewClientAuthType } from "@langchain/community/chat_models/oci_genai/types";
+
+const cohereLlm = new OciGenAiCohereChat({
+  compartmentId: "oci.compartment...",
+  onDemandModelId: "cohere.command-r-plus-08-2024",
+  newClientParams: {
+    authType: OciGenAiNewClientAuthType.ConfigFile,
+    authParams: {
+      clientConfigFilePath: "/my/path/config",
+      clientProfile: "MY_PROFILE_IN_CONFIG_FILE"
+    }
+  }
+});
+```
+
+The third example will create an SDK client with the following
+configuration: 1. Config file authentication. 1. Use [Resource
+Principal](https://docs.oracle.com/en-us/iaas/Content/API/Concepts/sdk_authentication_methods.htm#sdk_authentication_methods_resource_principal)
+authentication. 1. The retry strategy will be set to a single attempt.
+If the first API call was not successful, the request will fail. 1. The
+region will be set to `us-chicago-1`. Please make sure that your tenancy
+is registered this region.
+
+```python
+import { OciGenAiCohereChat } from "@langchain/community/chat_models/oci_genai/cohere_chat";
+import { OciGenAiNewClientAuthType } from "@langchain/community/chat_models/oci_genai/types";
+
+const cohereLlm = new OciGenAiCohereChat({
+  compartmentId: "oci.compartment...",
+  onDemandModelId: "cohere.command-r-plus-08-2024",
+  newClientParams: {
+    authType: OciGenAiNewClientAuthType.Other,
+    authParams: {
+      authenticationDetailsProvider: await ResourcePrincipalAuthenticationDetailsProvider.builder()
+    },
+  }
+});
+```
+
+You can also instantiate the OCI Generative AI chat classes using
+`GenerativeAiInferenceClient` that you create on your own. This way you
+control the creation and configuration of the client to suit your
+specific needs:
+
+```python
+import { ConfigFileAuthenticationDetailsProvider } from "oci-common";
+import { GenerativeAiInferenceClient } from "oci-generativeaiinference";
+import { OciGenAiCohereChat } from "@langchain/community/chat_models/oci_genai/cohere_chat";
+
+const client = new GenerativeAiInferenceClient({
+  authenticationDetailsProvider: new ConfigFileAuthenticationDetailsProvider()
+});
+
+const cohereLlm = new OciGenAiCohereChat({
+  compartmentId: "oci.compartment...",
+  onDemandModelId: "cohere.command-r-plus-08-2024",
+  client
+});
+```
+
+## Invocation
+
+In this example, we make a simple call to the OCI Generative AI service
+while leveraging the power of the `cohere.command-r-plus-08-2024` model.
+Please note that you can pass additional request parameters under the
+`requestParams` key as shown in the `invoke` call below. For more
+information please see the [Cohere request
+parameters](https://docs.oracle.com/en-us/iaas/api/#/en/generative-ai-inference/20231130/datatypes/CohereChatRequest)
+(the `apiFormat`, `chatHistory`, `isStream`, `message` & `stopSequences`
+parameters are automatically generated or inferred from the call
+context) and the [Generic request
+parameters](https://docs.oracle.com/en-us/iaas/api/#/en/generative-ai-inference/20231130/datatypes/GenericChatRequest)
+(the `apiFormat`, `isStream`, `messages` & `stop` parameters are
+automatically generated or inferred from the call context).
+
+If you wish to specify the chat history for a Cohere request, the list
+of messages passed into the request will be analyzed and split into the
+current message and history messages. The last `Human` message sent in
+the list (regardless of it’s position in the list) will be considered as
+the `message` parameter for the request and the rest of the messages
+will be added to the `chatHistory` parameter. If there are more than one
+`Human` messages, the very last one will be considered as the `message`
+to be sent to the LLM in the current request and the other will be
+appended to the `chatHistory`.
+
+```python
+import { OciGenAiCohereChat } from "@langchain/community/chat_models/oci_genai/cohere_chat";
+
+const llm = new OciGenAiCohereChat({
+  compartmentId: "oci.compartment...",
+  onDemandModelId: "cohere.command-r-plus-08-2024"
+});
+
+const result = await llm.invoke("Tell me a joke about beagles", {
+  requestParams: {
+    temperature: 1,
+    maxTokens: 300
+  }
+});
+
+console.log(result);
+```
+
+AIMessage { “content”: “Why did the beagle cross the road?he was tied to
+the chicken!hope you enjoyed the joke! Would you like to hear another
+one?”, “additional_kwargs”: {}, “response_metadata”: {}, “tool_calls”:
+\[\], “invalid_tool_calls”: \[\] }
+
+## Additional information
+
+For additional information, please checkout the [OCI Generative AI
+service
+documentation](https://docs.oracle.com/en-us/iaas/Content/generative-ai/home.htm).
+
+If you are interested in the python version of this integration, you can
+find more information
+[here](https://python.langchain.com/docs/integrations/llms/oci_generative_ai/).
+
+
+## Related
+
+- Chat model [conceptual guide](/docs/concepts/#chat-models)
+- Chat model [how-to guides](/docs/how_to/#chat-models)

libs/langchain-community/.gitignore

@@ -582,6 +582,10 @@ chat_models/ollama.cjs
 chat_models/ollama.js
 chat_models/ollama.d.ts
 chat_models/ollama.d.cts
+chat_models/oci_genai.cjs
+chat_models/oci_genai.js
+chat_models/oci_genai.d.ts
+chat_models/oci_genai.d.cts
 chat_models/portkey.cjs
 chat_models/portkey.js
 chat_models/portkey.d.ts

libs/langchain-community/langchain.config.js

@@ -184,6 +184,7 @@ export const config = {
     "chat_models/moonshot": "chat_models/moonshot",
     "chat_models/novita": "chat_models/novita",
     "chat_models/ollama": "chat_models/ollama",
+    "chat_models/oci_genai": "chat_models/oci_genai/index",
     "chat_models/portkey": "chat_models/portkey",
     "chat_models/premai": "chat_models/premai",
     "chat_models/tencent_hunyuan": "chat_models/tencent_hunyuan/index",
@@ -435,6 +436,7 @@ export const config = {
     "chat_models/bedrock",
     "chat_models/bedrock/web",
     "chat_models/llama_cpp",
+    "chat_models/oci_genai",
     "chat_models/portkey",
     "chat_models/premai",
     "chat_models/tencent_hunyuan",

libs/langchain-community/package.json

@@ -193,6 +193,8 @@
     "neo4j-driver": "^5.17.0",
     "node-llama-cpp": "3.1.1",
     "notion-to-md": "^3.1.0",
+    "oci-common": "^2.102.2",
+    "oci-generativeaiinference": "^2.102.2",
     "officeparser": "^4.0.4",
     "openai": "*",
     "pdf-parse": "1.1.1",
@@ -324,6 +326,8 @@
     "mysql2": "^3.9.8",
     "neo4j-driver": "*",
     "notion-to-md": "^3.1.0",
+    "oci-common": "^2.102.2",
+    "oci-generativeaiinference": "^2.102.2",
     "officeparser": "^4.0.4",
     "openai": "*",
     "pdf-parse": "1.1.1",
@@ -646,6 +650,12 @@
     "notion-to-md": {
       "optional": true
     },
+    "oci-common": {
+      "optional": true
+    },
+    "oci-generativeaiinference": {
+      "optional": true
+    },
     "officeparser": {
       "optional": true
     },
@@ -2031,6 +2041,15 @@
       "import": "./chat_models/ollama.js",
       "require": "./chat_models/ollama.cjs"
     },
+    "./chat_models/oci_genai": {
+      "types": {
+        "import": "./chat_models/oci_genai.d.ts",
+        "require": "./chat_models/oci_genai.d.cts",
+        "default": "./chat_models/oci_genai.d.ts"
+      },
+      "import": "./chat_models/oci_genai.js",
+      "require": "./chat_models/oci_genai.cjs"
+    },
     "./chat_models/portkey": {
       "types": {
         "import": "./chat_models/portkey.d.ts",
@@ -3807,6 +3826,10 @@
     "chat_models/ollama.js",
     "chat_models/ollama.d.ts",
     "chat_models/ollama.d.cts",
+    "chat_models/oci_genai.cjs",
+    "chat_models/oci_genai.js",
+    "chat_models/oci_genai.d.ts",
+    "chat_models/oci_genai.d.cts",
     "chat_models/portkey.cjs",
     "chat_models/portkey.js",
     "chat_models/portkey.d.ts",