Merge pull request #2 from samestrin/interfaces

v0.0.10

README.md

@@ -10,6 +10,13 @@ The LLM Interface project is a versatile and comprehensive wrapper designed to i
 
 ## Updates
 
+**v0.0.10**
+
+- **Hugging Face**: Added support for Hugging Face Inference (_over 150,000 publicly accessible machine learning models_)
+- **Perplexity Labs**: Added support for Perplexity Labs
+- **JSON Improvements**: The `json_object` mode for OpenAI and Gemini now guarantees the return a valid JSON object or null.
+- **Graceful Retries**: Retry LLM queries upon failure.
+
 **v0.0.9**
 
 - **Response Caching**: Efficiently caches LLM responses to reduce costs, enhance performance and minimize redundant requests, with customizable cache timeout settings.
@@ -19,10 +26,6 @@ The LLM Interface project is a versatile and comprehensive wrapper designed to i
 - **Mistral AI**: Added support for Mistral AI
 - **Cohere**: Added support for Cohere
 
-**v0.0.7**
-
-- **Goose AI**: Added support for Goose AI
-
 ## Features
 
 - **Unified Interface**: A single, consistent interface to interact with multiple LLM APIs.
@@ -36,7 +39,7 @@ The LLM Interface project is a versatile and comprehensive wrapper designed to i
 
 The project relies on several npm packages and APIs. Here are the primary dependencies:
 
-- `axios`: For making HTTP requests (used for Cohere, Goose AI, LLaMA.cpp, Mistral, and Reka AI).
+- `axios`: For making HTTP requests (used for various HTTP AI APIs).
 - `@anthropic-ai/sdk`: SDK for interacting with the Anthropic API.
 - `@google/generative-ai`: SDK for interacting with the Google Gemini API.
 - `groq-sdk`: SDK for interacting with the Groq API.

config.js

@@ -15,4 +15,7 @@ module.exports = {
   gooseApiKey: process.env.GOOSE_API_KEY,
   cohereApiKey: process.env.COHERE_API_KEY,
   mistralApiKey: process.env.MISTRAL_API_KEY,
+  huggingfaceApiKey: process.env.HUGGINGFACE_API_KEY,
+  perplexityApiKey: process.env.PERPLEXITY_API_KEY,
+  ai21ApiKey: process.env.AI21_API_KEY,
 };

docs/API.md

@@ -37,6 +37,28 @@ openai
   });
 ```
 
+### AI21
+
+#### `sendMessage(message, options, cacheTimeoutSeconds)`
+
+- **Parameters:**
+  - `message`: An object containing the model and messages to send.
+  - `options`: An optional object containing `max_tokens` and `model`.
+  - `cacheTimeoutSeconds`: An optional number specifying the cache timeout in seconds. If set, caching is enabled.
+- **Returns:** A promise that resolves to the response text.
+- **Example:**
+
+```javascript
+ai21
+  .sendMessage(message, { max_tokens: 150 })
+  .then((response) => {
+    console.log(response);
+  })
+  .catch((error) => {
+    console.error(error);
+  });
+```
+
 ### Anthropic
 
 #### `sendMessage(message, options, cacheTimeoutSeconds)`
@@ -147,6 +169,28 @@ groq
   });
 ```
 
+### Hugging Face
+
+#### `sendMessage(message, options, cacheTimeoutSeconds)`
+
+- **Parameters:**
+  - `message`: An object containing the model and messages to send.
+  - `options`: An optional object containing `max_tokens` and `model`.
+  - `cacheTimeoutSeconds`: An optional number specifying the cache timeout in seconds. If set, caching is enabled.
+- **Returns:** A promise that resolves to the response text.
+- **Example:**
+
+```javascript
+huggingface
+  .sendMessage(message, { max_tokens: 100 })
+  .then((response) => {
+    console.log(response);
+  })
+  .catch((error) => {
+    console.error(error);
+  });
+```
+
 ### Mistral AI
 
 #### `sendMessage(message, options, cacheTimeoutSeconds)`
@@ -169,6 +213,28 @@ mistral
   });
 ```
 
+### Perplexity Labs
+
+#### `sendMessage(message, options, cacheTimeoutSeconds)`
+
+- **Parameters:**
+  - `message`: An object containing the model and messages to send.
+  - `options`: An optional object containing `max_tokens` and `model`.
+  - `cacheTimeoutSeconds`: An optional number specifying the cache timeout in seconds. If set, caching is enabled.
+- **Returns:** A promise that resolves to the response text.
+- **Example:**
+
+```javascript
+perplexity
+  .sendMessage(message, { max_tokens: 100 })
+  .then((response) => {
+    console.log(response);
+  })
+  .catch((error) => {
+    console.error(error);
+  });
+```
+
 ### Reka AI
 
 #### `sendMessage(message, options, cacheTimeoutSeconds)`

docs/APIKEYS.md

@@ -8,6 +8,12 @@ The OpenAI API requires a credit card.
 
 - https://platform.openai.com/api-keys
 
+## AI21 Studio
+
+The AI21 API is a commercial product, but it currently does not require a credit card, and comes with a $90 credit.
+
+- https://studio.ai21.com/account/api-key?source=docs
+
 ## Anthropic
 
 The Anthropic API requires a credit card.
@@ -38,12 +44,24 @@ The Groq API is currently free.
 
 - https://console.groq.com/keys
 
+## Hugging Face
+
+The Hugging Face Inference API is currently free for rate-limited, non-commercial use.
+
+- https://huggingface.co/settings/tokens
+
 ## Mistral
 
 The Mistral API is a commercial product, but it currently does not require a credit card, and comes with a $5.00 credit.
 
 - https://console.mistral.ai/api-keys/
 
+## Perplexity
+
+The Perplexity API requires a credit cards.
+
+- https://www.perplexity.ai/settings/api
+
 ## Reka AI
 
 The Reka AI API requires a credit card, but currently comes with a $5 credit.

docs/USAGE.md

@@ -5,18 +5,21 @@
 - [Initializing llm-interface](#initializing-llm-interface)
 - [Basic Usage Examples](#basic-usage-examples)
   - [OpenAI Interface](#openai-interface)
+  - [AI21 Interface](#ai21-interface)
   - [Anthropic Interface](#anthropic-interface)
   - [Cohere Interface](#cohere-interface)
   - [Gemini Interface](#gemini-interface)
   - [Goose AI Interface](#goose-ai-interface)
   - [Groq Interface](#groq-interface)
+  - [HuggingFace Interface](#huggingface-interface)
   - [Mistral AI Interface](#mistral-ai-interface)
+  - [Perplexity Interface](#perplexity-interface)
   - [Reka AI Interface](#reka-ai-interface)
   - [LLaMA.cpp Interface](#llamacpp-interface)
 - [Advanced Usage Examples](#advanced-usage-examples)
   - [OpenAI Interface (JSON Output)](#openai-interface-json-output)
-  - [Gemini Interface (JSON Output)](#gemini-interface-json-output)
   - [OpenAI Interface (Cached)](#openai-interface-cached)
+  - [OpenAI Interface (Graceful Retry)](#openai-interface-graceful-retry)
 
 # Usage
 
@@ -99,6 +102,33 @@ anthropic
   });
 ```
 
+### AI21 Interface
+
+The AI21 interface allows you to send messages to the AI21 API.
+
+#### Example
+
+```javascript
+const ai21 = new LLMInterface.ai21(process.env.AI21_API_KEY);
+
+const message = {
+  model: "jamba-instruct",
+  messages: [
+    { role: "system", content: "You are a helpful assistant." },
+    { role: "user", content: "Explain the importance of low latency LLMs." },
+  ],
+};
+
+ai21
+  .sendMessage(message, { max_tokens: 150 })
+  .then((response) => {
+    console.log(response);
+  })
+  .catch((error) => {
+    console.error(error);
+  });
+```
+
 ### Cohere Interface
 
 The Cohere interface allows you to send messages to the Cohere API.
@@ -207,6 +237,38 @@ groq
   });
 ```
 
+### HuggingFace Interface
+
+The HuggingFace interface allows you to send messages to the HuggingFace API.
+
+#### Example
+
+```javascript
+const huggingface = new LLMInterface.huggingface(process.env.ANTHROPIC_API_KEY);
+
+const message = {
+  model: "claude-3-opus-20240229",
+  messages: [
+    {
+      role: "user",
+      content:
+        "You are a helpful assistant. Say OK if you understand and stop.",
+    },
+    { role: "system", content: "OK" },
+    { role: "user", content: "Explain the importance of low latency LLMs." },
+  ],
+};
+
+huggingface
+  .sendMessage(message, { max_tokens: 150 })
+  .then((response) => {
+    console.log(response);
+  })
+  .catch((error) => {
+    console.error(error);
+  });
+```
+
 ### Mistral AI Interface
 
 The Mistral AI interface allows you to send messages to the Mistral AI API.
@@ -234,6 +296,38 @@ mistral
   });
 ```
 
+### Perplexity Interface
+
+The Perplexity interface allows you to send messages to the Perplexity API.
+
+#### Example
+
+```javascript
+const perplexity = new LLMInterface.perplexity(process.env.ANTHROPIC_API_KEY);
+
+const message = {
+  model: "claude-3-opus-20240229",
+  messages: [
+    {
+      role: "user",
+      content:
+        "You are a helpful assistant. Say OK if you understand and stop.",
+    },
+    { role: "system", content: "OK" },
+    { role: "user", content: "Explain the importance of low latency LLMs." },
+  ],
+};
+
+perplexity
+  .sendMessage(message, { max_tokens: 150 })
+  .then((response) => {
+    console.log(response);
+  })
+  .catch((error) => {
+    console.error(error);
+  });
+```
+
 ### Reka AI Interface
 
 The Reka AI interface allows you to send messages to the Reka AI REST API.
@@ -294,7 +388,7 @@ Then select the interface you'd like to use and initialize it with an API key or
 
 ### OpenAI Interface (JSON Output)
 
-The OpenAI interface allows you to send messages to the OpenAI API and request the response back in JSON. To take advantage of this feature be sure to include text like "Return the results as a JSON object." and provide a desired output format like "Follow this format: [{reason, reasonDescription}]."
+Some interfaces allows you request the response back in JSON, currently **OpenAI** and **Gemini** are supported. To take advantage of this feature be sure to include text like "Return the results as a JSON object." and provide a desired output format like "Follow this format: [{reason, reasonDescription}]." In this example we use OpenAI and request a valid JSON object.
 
 #### Example
 
@@ -326,32 +420,25 @@ openai
   });
 ```
 
-### Gemini Interface (JSON Output)
+### OpenAI Interface (Cached)
 
-The Gemini interface allows you to send messages to the Google Gemini API. To take advantage of this feature be sure to include text like "Return the results as a JSON object." and provide a desired output format like "Follow this format: [{reason, reasonDescription}]."
+To reduce operational costs and improve performance you can optionally specify a cache timeout in seconds. In this example we use OpenAI and store the results for 86400 seconds or one day.
 
 #### Example
 
 ```javascript
-const gemini = new LLMInterface.gemini(process.env.GEMINI_API_KEY);
+const openai = new LLMInterface.openai(process.env.OPENAI_API_KEY);
 
 const message = {
-  model: "gemini-1.5-flash",
+  model: "gpt-3.5-turbo",
   messages: [
-    {
-      role: "system",
-      content: "You are a helpful assistant.",
-    },
-    {
-      role: "user",
-      content:
-        "Explain the importance of low latency LLMs. Return the results as a JSON object. Follow this format: [{reason, reasonDescription}].",
-    },
+    { role: "system", content: "You are a helpful assistant." },
+    { role: "user", content: "Explain the importance of low latency LLMs." },
   ],
 };
 
-gemini
-  .sendMessage(message, { max_tokens: 100, response_format: "json_object" })
+openai
+  .sendMessage(message, { max_tokens: 150 }, { cacheTimeoutSeconds: 86400 })
   .then((response) => {
     console.log(response);
   })
@@ -360,9 +447,9 @@ gemini
   });
 ```
 
-### OpenAI Interface (Cached)
+### OpenAI Interface (Graceful Retry)
 
-The OpenAI interface allows you to send messages to the OpenAI API. To reduce operational costs and improve performance you can optionally specify a cache timeout in seconds. In this example we store the results for 86400 seconds or one day.
+You can gracefully retry your requests. In this example we use OpenAI and up to 3 times if needed.
 
 #### Example
 
@@ -378,7 +465,7 @@ const message = {
 };
 
 openai
-  .sendMessage(message, { max_tokens: 150 }, 86400)
+  .sendMessage(message, { max_tokens: 150 }, { retryAttempts: 3 })
   .then((response) => {
     console.log(response);
   })

env

@@ -5,4 +5,8 @@
   REKA_API_KEY=
   GOOSE_API_KEY=
   MISTRAL_API_KEY=
+  HUGGINGFACE_API_KEY=
+  PERPLEXITY_API_KEY=
+  AI21_API_KEY=
+  AZUREAI_API_KEY=
   LLAMACPP_URL=http://localhost:8080/completions

package.json

@@ -1,6 +1,6 @@
 {
   "name": "llm-interface",
-  "version": "0.0.9",
+  "version": "0.0.10",
   "main": "src/index.js",
   "description": "A simple, unified interface for integrating and interacting with multiple Large Language Model (LLM) APIs, including OpenAI, Anthropic, Google Gemini, Groq, and LlamaCPP.",
   "scripts": {