app.js

const express = require("express");
const app = express();
const port = 7860;

app.use(express.json());

app.get("//models", async (req, res) => {
  try {
    res.status(200).json({ object: "list", data: createFakeModelsList() });
  } catch {}
});

app.post("//chat/completions", async (clientRequest, clientResponse) => {
  try {
    const {
      frequency_penalty,
      presence_penalty,
      max_tokens,
      stop,
      temperature,
      top_p,
    } = clientRequest.body;

    const apiRequestBody = {
      model: "gpt-4",
      prompt: convertChatMLPrompt(clientRequest.body.messages),
      frequency_penalty,
      presence_penalty,
      max_tokens,
      stop,
      temperature,
      top_p,
    };

    const apiResponse = await fetch(process.env.API_URL, {
      method: "POST",
      headers: {
        "Content-Type": "application/json",
      },
      body: JSON.stringify(apiRequestBody),
    });

    if (clientRequest.body.stream) {
      handleResponseAsStream(clientResponse, apiResponse);
    } else {
      handleResponseAsNonStreamable(clientResponse, apiResponse);
    }
  } catch {}
});

app.listen(port, () => {
  console.log(`Example app listening on port ${port}`);
});

async function handleResponseAsNonStreamable(clientResponse, apiResponse) {
  const apiText = await apiResponse.text();
  const clientMessage = createClientMessage(apiText);
  clientResponse.send(JSON.stringify(clientMessage));
}

async function handleResponseAsStream(clientResponse, apiResponse) {
  const reader = apiResponse.body.getReader();
  const nextDecoder = new TextDecoder();

  clientResponse.write("data: " + JSON.stringify(createBeginChunk()) + "\n\n");

  new ReadableStream({
    start(controller) {
      return pump();
      function pump() {
        return reader.read().then(({ done, value }) => {
          const textData = nextDecoder.decode(value);

          clientResponse.write(
            "data: " + JSON.stringify(createMessageChunk(textData)) + "\n\n"
          );

          // When no more data needs to be consumed, close the stream
          if (done) {
            clientResponse.write(
              "data: " + JSON.stringify(createEndChunk()) + "\n\n"
            );
            clientResponse.end();
            controller.close();
            return;
          }
          // Enqueue the next data chunk into our target stream
          controller.enqueue(value);
          return pump();
        });
      }
    },
  });
}

function getCurrentDate() {
  return Math.floor(new Date().getTime());
}

function convertChatMLPrompt(messages) {
  const messageStrings = [];
  messages.forEach((m) => {
    if (m.role === "system" && m.name === undefined) {
      messageStrings.push("System: " + m.content);
    } else if (m.role === "system" && m.name !== undefined) {
      messageStrings.push(m.name + ": " + m.content);
    } else {
      messageStrings.push(m.role + ": " + m.content);
    }
  });
  return messageStrings.join("\n") + "\nassistant:";
}

const createClientMessage = (text) => ({
  id: "chatcmpl-123",
  object: "chat.completion",
  created: getCurrentDate(),
  model: "gpt-4",
  choices: [
    {
      index: 0,
      message: { role: "assistant", content: text },
      logprobs: null,
      finish_reason: "stop",
    },
  ],
});

const createBeginChunk = () => ({
  id: "chatcmpl-123",
  object: "chat.completion.chunk",
  created: getCurrentDate(),
  model: "gpt-4",
  system_fingerprint: "",
  choices: [
    {
      index: 0,
      delta: { role: "assistant", content: "" },
      logprobs: null,
      finish_reason: null,
    },
  ],
});

const createMessageChunk = (text) => ({
  id: "chatcmpl-123",
  object: "chat.completion.chunk",
  created: getCurrentDate(),
  model: "gpt-4",
  system_fingerprint: "",
  choices: [
    {
      index: 0,
      delta: { content: text },
      logprobs: null,
      finish_reason: null,
    },
  ],
});

const createEndChunk = () => ({
  id: "chatcmpl-123",
  object: "chat.completion.chunk",
  created: getCurrentDate(),
  model: "gpt-4",
  system_fingerprint: "",
  choices: [{ index: 0, delta: {}, logprobs: null, finish_reason: "stop" }],
});

function createFakeModelsList() {
  return [
    {
      id: "gpt-4",
      object: "model",
      created: getCurrentDate(),
      owned_by: "openai",
      permission: [
        {
          id: "modelperm-gpt-4",
          object: "model_permission",
          created: getCurrentDate(),
          organization: "*",
          group: null,
          is_blocking: false,
        },
      ],
      root: "gpt-4",
      parent: null,
    },
  ];
}