Merge main into sweep/connectors-support

README.md

@@ -89,6 +89,37 @@ Check out our NextJS playground at https://llama-playground.vercel.app/. The sou
 - OpenAI GPT-3.5-turbo and GPT-4
 - Anthropic Claude Instant and Claude 2
 - Llama2 Chat LLMs (70B, 13B, and 7B parameters)
+- Python Llama CPP API
+- Hugging Faces Text Generation Inference
+- Ooba Text Generation API
+
+## New Connectors:
+
+We have added support for three new connectors:
+
+- Python Llama CPP API: This connector allows you to use the Python bindings for the llama.cpp library with LlamaIndex.TS. To use this connector, simply import it and create a new instance of the PythonLlamaCpp class.
+
+```ts
+import { PythonLlamaCpp } from "llamaindex";
+
+const pythonLlamaCpp = new PythonLlamaCpp();
+```
+
+- Hugging Faces Text Generation Inference: This connector allows you to use the Hugging Faces text generation inference tool with LlamaIndex.TS. To use this connector, simply import it and create a new instance of the HuggingFacesTextGeneration class.
+
+```ts
+import { HuggingFacesTextGeneration } from "llamaindex";
+
+const huggingFacesTextGeneration = new HuggingFacesTextGeneration();
+```
+
+- Ooba Text Generation API: This connector allows you to use the Ooba text generation API with LlamaIndex.TS. To use this connector, simply import it and create a new instance of the OobaTextGeneration class.
+
+```ts
+import { OobaTextGeneration } from "llamaindex";
+
+const oobaTextGeneration = new OobaTextGeneration();
+```
 
 ## Contributing:
 

packages/core/src/llm/HuggingFacesTextGeneration.ts

@@ -0,0 +1,25 @@
+import axios from 'axios';
+
+class HuggingFacesTextGeneration {
+  private apiEndpoint: string;
+
+  constructor(apiEndpoint: string) {
+    this.apiEndpoint = apiEndpoint;
+  }
+
+  async sendRequest(prompt: string) {
+    const response = await axios.post(this.apiEndpoint, {
+      prompt: prompt,
+    });
+    return this.receiveResponse(response);
+  }
+
+  receiveResponse(response: any) {
+    if (response.status !== 200) {
+      throw new Error(`Unexpected response code: ${response.status}`);
+    }
+    return response.data;
+  }
+}
+
+export default HuggingFacesTextGeneration;

packages/core/src/llm/LLM.ts

@@ -16,6 +16,9 @@ import {
 } from "./azure";
 import { OpenAISession, getOpenAISession } from "./openai";
 import { ReplicateSession } from "./replicate";
+import { PythonLlamaCpp } from "./PythonLlamaCpp";
+import { HuggingFacesTextGeneration } from "./HuggingFacesTextGeneration";
+import { OobaTextGeneration } from "./OobaTextGeneration";
 
 export type MessageType =
   | "user"
@@ -24,21 +27,17 @@ export type MessageType =
   | "generic"
   | "function";
 
-export interface ChatMessage {
-  content: string;
-  role: MessageType;
-}
-
-export interface ChatResponse {
-  message: ChatMessage;
-  raw?: Record<string, any>;
-  delta?: string;
-}
-
-// NOTE in case we need CompletionResponse to diverge from ChatResponse in the future
-export type CompletionResponse = ChatResponse;
-
-/**
+export const ALL_AVAILABLE_LLAMAS = {
+  "openai-gpt3": OpenAI,
+  "openai-gpt4": OpenAI,
+  "anthropic-claude": Anthropic,
+  "llama2-70b": LlamaDeuce,
+  "llama2-13b": LlamaDeuce,
+  "llama2-7b": LlamaDeuce,
+  "python-llama-cpp": PythonLlamaCpp,
+  "hugging-faces-text-generation": HuggingFacesTextGeneration,
+  "ooba-text-generation": OobaTextGeneration,
+};
  * Unified language model interface
  */
 export interface LLM {
@@ -49,31 +48,27 @@ export interface LLM {
   chat(messages: ChatMessage[], parentEvent?: Event): Promise<ChatResponse>;
 
   /**
-   * Get a prompt completion from the LLM
-   * @param prompt the prompt to complete
-   */
-  complete(prompt: string, parentEvent?: Event): Promise<CompletionResponse>;
-}
-
-export const GPT4_MODELS = {
-  "gpt-4": { contextWindow: 8192 },
-  "gpt-4-32k": { contextWindow: 32768 },
-};
-
-export const TURBO_MODELS = {
-  "gpt-3.5-turbo": { contextWindow: 4096 },
-  "gpt-3.5-turbo-16k": { contextWindow: 16384 },
-};
-
-/**
- * We currently support GPT-3.5 and GPT-4 models
- */
-export const ALL_AVAILABLE_OPENAI_MODELS = {
-  ...GPT4_MODELS,
-  ...TURBO_MODELS,
-};
-
-/**
+   export function getLLM(llmName: keyof typeof ALL_AVAILABLE_LLAMAS): LLM {
+     switch (llmName) {
+       case "openai-gpt3":
+       case "openai-gpt4":
+         return new OpenAI();
+       case "anthropic-claude":
+         return new Anthropic();
+       case "llama2-70b":
+       case "llama2-13b":
+       case "llama2-7b":
+         return new LlamaDeuce();
+       case "python-llama-cpp":
+         return new PythonLlamaCpp();
+       case "hugging-faces-text-generation":
+         return new HuggingFacesTextGeneration();
+       case "ooba-text-generation":
+         return new OobaTextGeneration();
+       default:
+         throw new Error(`Unsupported LLM: ${llmName}`);
+     }
+   }
  * OpenAI LLM implementation
  */
 export class OpenAI implements LLM {

packages/core/src/llm/OobaTextGeneration.ts

@@ -0,0 +1,21 @@
+import axios from 'axios';
+
+class OobaTextGeneration {
+  private apiURL: string;
+
+  constructor(apiURL: string) {
+    this.apiURL = apiURL;
+  }
+
+  async sendRequest(input: string): Promise<string> {
+    try {
+      const response = await axios.post(this.apiURL, { text: input });
+      return response.data;
+    } catch (error) {
+      console.error(`Error sending request to Ooba Text Generation API: ${error}`);
+      throw error;
+    }
+  }
+}
+
+export default OobaTextGeneration;

packages/core/src/llm/PythonLlamaCpp.ts

@@ -0,0 +1,23 @@
+import axios from 'axios';
+
+class PythonLlamaCpp {
+  private apiURL: string;
+
+  constructor(apiURL: string) {
+    this.apiURL = apiURL;
+  }
+
+  async sendRequest(prompt: string) {
+    const response = await axios.post(this.apiURL, {
+      prompt: prompt
+    });
+    return response.data;
+  }
+
+  async receiveResponse() {
+    const response = await axios.get(this.apiURL);
+    return response.data;
+  }
+}
+
+export default PythonLlamaCpp;