Skip to main content

Language Models

The SocialAGI library is written primarily to interface with GPT3.5 and GPT4 from OpenAI using the OpenAI chat interface. However, the library decouples the interfaces directly from the chat models with the LanguageModelProgramExecutor interface.

Interfaces​

So long as the following interfaces are implemented, there is no explicit OpenAI dependency.

interface LanguageModelProgramExecutor {
  execute<FunctionCallReturnType = undefined>(
    records: ChatMessage[],
    chatCompletionParams?: LanguageModelProgramExecutorExecuteOptions,
    functions?: FunctionSpecification[],
    requestOptions?: RequestOptions,
  ): Promise<ExecutorResponse<FunctionCallReturnType>>;

  experimentalStreamingExecute?<FunctionCallReturnType = undefined>(
    messages: ChatMessage[],
    completionParams?: LanguageModelProgramExecutorExecuteOptions,
    functions?: FunctionSpecification[],
    requestOptions?: RequestOptions,
  ): Promise<{
    response: Promise<ExecutorResponse<FunctionCallReturnType>>,
    stream: AsyncIterable<string>,
  }>
}

Alternative API calling​

We also include a FunctionlessLLM LanguageModelProgramExecutor implementation that supports APIs with OpenAI chat API support (but does not require function calling support). This executor also supports calling APIs that do not allow more than one system message:

let step = new CortexStep("Jonathan", {
  processor: new FunctionlessLLM({
    baseURL: "http://localhost:1234/v1",
    compressSystemMessages: true,
  })
});

Many Mistral and Together AI models work with this same interface, though we recommend only using their most intelligent models, e.g.

let step = new CortexStep("Jonathan", {
    processor: new FunctionlessLLM({
    baseURL: "https://api.mistral.ai/v1/",
    singleSystemMessage: true,
    apiKey: "your_xyz_MISTRAL_API_KEY",
  }, {
    model: "mistral-medium",
    temperature: 0.8,
    max_tokens: 300,
  })
})

or

let step = new CortexStep("Jonathan", {
  processor: new FunctionlessLLM({
    baseURL: "https://api.together.xyz/v1",
    singleSystemMessage: true,
    apiKey: "your_xyz_TOGETHER_API_KEY",
  }, {
    model: "NousResearch/Nous-Hermes-2-Yi-34B",
    temperature: 0.7,
    max_tokens: 300,
  })
})

Example OpenAI completion engines​

We provide methods to create an OpenAI version and specify the model:

const executor = new OpenAILanguageProgramProcessor(
  {},
  {
    model: "gpt-3.5-turbo",
  },
);