Support other/local LLMs

[olmohake]

I am very exited by the idea of text to gql and would love to implement it for my organization. The context to send a long is pretty big though. This I’d love the option to use a local llama instance to reduce costs. Any chances to support other/local llms in the future?

[danstarns]

@olmohake I shipped the adapter architecture, now you can extend the base adapter to use your own LLMs, please share more info on what other LLMs and patterns you wish to use.

1. https://github.com/rocket-connect/gqlpt/tree/main/packages/adapter-base - Extend Adapter
2. OpenAI adapter https://github.com/rocket-connect/gqlpt/tree/main/packages/adapter-openai

- import { AdapterOpenAI } from "@gqlpt/adapter-openai";
+ import { MyAdapter } from "my-adapter";
import { GQLPTClient } from "gqlpt";

const typeDefs = /* GraphQL */ `
  type User {
    id: ID!
    name: String!
  }

  type Query {
    user(id: ID!): User
  }
`;

const client = new GQLPTClient({
  typeDefs,
-  adapter: new AdapterOpenAI({
    apiKey: process.env.OPENAI_API_KEY,
  }),
+ adapter: new MyAdapter()
});

[danstarns]

See:

1. Adapter OpenAI - https://github.com/rocket-connect/gqlpt/tree/main/packages/adapter-openai
2. Adapter Anthropic - https://github.com/rocket-connect/gqlpt/tree/main/packages/adapter-anthropic