Gateway

Gateway Class

The main class for interacting with LLMs through the Adaline Gateway.

Constructor

import { Gateway } from "@adaline/gateway";

const gateway = new Gateway({
  cache: myCache,        // optional: pluggable cache backend
  httpClient: myClient,  // optional: custom HTTP client
  logger: myLogger,      // optional: custom logger
});

Parameters

options · GatewayOptionsType

Optional configuration for plugins.

Show GatewayOptionsType

cache · Cache<T>

A pluggable cache backend implementing the Cache interface.

httpClient · HttpClient

A custom HTTP client implementing the HttpClient interface.

logger · Logger

A custom logger implementing the Logger interface.

Methods

completeChat

Non-streaming chat completion. Sends messages to an LLM and returns the full response.

const response = await gateway.completeChat({
  model: gpt4o,
  config: Config().parse({ temperature: 0.7, maxTokens: 500 }),
  messages: [
    { role: "system", content: [{ modality: "text", value: "You are a helpful assistant." }] },
    { role: "user", content: [{ modality: "text", value: "Explain quantum computing." }] },
  ],
  tools: [],
});

request · GatewayCompleteChatRequestType · required

Show Properties

model · ChatModel · required

The model instance from a provider (e.g., openai.chatModel({ modelName: "gpt-4o" })).

config · Config · required

Model configuration parsed via Config().parse({...}). See Config Type.

messages · MessageType[] · required

Array of messages. See Message Types.

tools · ToolType[] · required

Array of tool definitions (can be empty). See Tool Types.

Returns: CompleteChatHandlerResponseType

---

streamChat

Streaming chat completion. Returns an async generator that yields response chunks.

for await (const chunk of gateway.streamChat({
  model: gpt4o,
  config: Config().parse({ temperature: 0.7, maxTokens: 500 }),
  messages: [
    { role: "user", content: [{ modality: "text", value: "Write a poem." }] },
  ],
  tools: [],
})) {
  process.stdout.write(chunk.response);
}

request · GatewayStreamChatRequestType · required

Same shape as completeChat request. See above.

Returns: AsyncGenerator<StreamChatHandlerResponseType>

---

getEmbeddings

Generate embeddings for text or other modalities.

const embeddings = await gateway.getEmbeddings({
  model: textEmbedding3Large,
  config: Config().parse({ encodingFormat: "float", dimensions: 256 }),
  embeddingRequests: {
    modality: "text",
    requests: ["Hello world", "How are you?"],
  },
});

request · GatewayGetEmbeddingsRequestType · required

Show Properties

model · EmbeddingModel · required

The embedding model instance from a provider.

config · Config · required

Embedding model configuration.

embeddingRequests · EmbeddingRequestType · required

The input data to embed.

Returns: GetEmbeddingsHandlerResponseType

---

getToolResponses

Execute tool calls returned by an LLM.

const toolResponses = await gateway.getToolResponses({
  tools: myToolDefinitions,
  toolCalls: response.toolCalls,
});

request · GatewayGetToolResponsesRequestType · required

Show Properties

tools · ToolType[] · required

The tool definitions that match the tool calls.

toolCalls · ToolCallType[] · required

Tool calls returned from a chat completion response.

Returns: GetToolResponsesHandlerResponseType

---

getChatUsageCost (static)

Calculate the cost of a chat completion based on token usage.

const cost = Gateway.getChatUsageCost({
  model: gpt4o,
  usage: response.usage,
});

Returns: GetChatUsageCostHandlerResponseType