Spaces:
Sleeping
Sleeping
Create app.js
Browse files
app.js
ADDED
|
@@ -0,0 +1,191 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
const express = require("express");
|
| 2 |
+
const app = express();
|
| 3 |
+
const port = 3000;
|
| 4 |
+
|
| 5 |
+
app.use(express.json());
|
| 6 |
+
|
| 7 |
+
app.get("//models", async (req, res) => {
|
| 8 |
+
try {
|
| 9 |
+
res.status(200).json({ object: "list", data: createFakeModelsList() });
|
| 10 |
+
} catch {}
|
| 11 |
+
});
|
| 12 |
+
|
| 13 |
+
app.post("//chat/completions", async (clientRequest, clientResponse) => {
|
| 14 |
+
try {
|
| 15 |
+
const {
|
| 16 |
+
frequency_penalty,
|
| 17 |
+
presence_penalty,
|
| 18 |
+
max_tokens,
|
| 19 |
+
stop,
|
| 20 |
+
temperature,
|
| 21 |
+
top_p,
|
| 22 |
+
} = clientRequest.body;
|
| 23 |
+
|
| 24 |
+
const apiRequestBody = {
|
| 25 |
+
model: "gpt-4",
|
| 26 |
+
prompt: convertChatMLPrompt(clientRequest.body.messages),
|
| 27 |
+
frequency_penalty,
|
| 28 |
+
presence_penalty,
|
| 29 |
+
max_tokens,
|
| 30 |
+
stop,
|
| 31 |
+
temperature,
|
| 32 |
+
top_p,
|
| 33 |
+
};
|
| 34 |
+
|
| 35 |
+
const apiResponse = await fetch(process.env.API_URL, {
|
| 36 |
+
method: "POST",
|
| 37 |
+
headers: {
|
| 38 |
+
"Content-Type": "application/json",
|
| 39 |
+
},
|
| 40 |
+
body: JSON.stringify(apiRequestBody),
|
| 41 |
+
});
|
| 42 |
+
|
| 43 |
+
if (clientRequest.body.stream) {
|
| 44 |
+
handleResponseAsStream(clientResponse, apiResponse);
|
| 45 |
+
} else {
|
| 46 |
+
handleResponseAsNonStreamable(clientResponse, apiResponse);
|
| 47 |
+
}
|
| 48 |
+
} catch {}
|
| 49 |
+
});
|
| 50 |
+
|
| 51 |
+
app.listen(port, () => {
|
| 52 |
+
console.log(`Example app listening on port ${port}`);
|
| 53 |
+
});
|
| 54 |
+
|
| 55 |
+
async function handleResponseAsNonStreamable(clientResponse, apiResponse) {
|
| 56 |
+
const apiText = await apiResponse.text();
|
| 57 |
+
const clientMessage = createClientMessage(apiText);
|
| 58 |
+
clientResponse.send(JSON.stringify(clientMessage));
|
| 59 |
+
}
|
| 60 |
+
|
| 61 |
+
async function handleResponseAsStream(clientResponse, apiResponse) {
|
| 62 |
+
const reader = apiResponse.body.getReader();
|
| 63 |
+
const nextDecoder = new TextDecoder();
|
| 64 |
+
|
| 65 |
+
clientResponse.write("data: " + JSON.stringify(createBeginChunk()) + "\n\n");
|
| 66 |
+
|
| 67 |
+
new ReadableStream({
|
| 68 |
+
start(controller) {
|
| 69 |
+
return pump();
|
| 70 |
+
function pump() {
|
| 71 |
+
return reader.read().then(({ done, value }) => {
|
| 72 |
+
const textData = nextDecoder.decode(value);
|
| 73 |
+
|
| 74 |
+
clientResponse.write(
|
| 75 |
+
"data: " + JSON.stringify(createMessageChunk(textData)) + "\n\n"
|
| 76 |
+
);
|
| 77 |
+
|
| 78 |
+
// When no more data needs to be consumed, close the stream
|
| 79 |
+
if (done) {
|
| 80 |
+
clientResponse.write(
|
| 81 |
+
"data: " + JSON.stringify(createEndChunk()) + "\n\n"
|
| 82 |
+
);
|
| 83 |
+
clientResponse.end();
|
| 84 |
+
controller.close();
|
| 85 |
+
return;
|
| 86 |
+
}
|
| 87 |
+
// Enqueue the next data chunk into our target stream
|
| 88 |
+
controller.enqueue(value);
|
| 89 |
+
return pump();
|
| 90 |
+
});
|
| 91 |
+
}
|
| 92 |
+
},
|
| 93 |
+
});
|
| 94 |
+
}
|
| 95 |
+
|
| 96 |
+
function getCurrentDate() {
|
| 97 |
+
return Math.floor(new Date().getTime());
|
| 98 |
+
}
|
| 99 |
+
|
| 100 |
+
function convertChatMLPrompt(messages) {
|
| 101 |
+
const messageStrings = [];
|
| 102 |
+
messages.forEach((m) => {
|
| 103 |
+
if (m.role === "system" && m.name === undefined) {
|
| 104 |
+
messageStrings.push("System: " + m.content);
|
| 105 |
+
} else if (m.role === "system" && m.name !== undefined) {
|
| 106 |
+
messageStrings.push(m.name + ": " + m.content);
|
| 107 |
+
} else {
|
| 108 |
+
messageStrings.push(m.role + ": " + m.content);
|
| 109 |
+
}
|
| 110 |
+
});
|
| 111 |
+
return messageStrings.join("\n") + "\nassistant:";
|
| 112 |
+
}
|
| 113 |
+
|
| 114 |
+
const createClientMessage = (text) => ({
|
| 115 |
+
id: "chatcmpl-123",
|
| 116 |
+
object: "chat.completion",
|
| 117 |
+
created: getCurrentDate(),
|
| 118 |
+
model: "gpt-4",
|
| 119 |
+
choices: [
|
| 120 |
+
{
|
| 121 |
+
index: 0,
|
| 122 |
+
message: { role: "assistant", content: text },
|
| 123 |
+
logprobs: null,
|
| 124 |
+
finish_reason: "stop",
|
| 125 |
+
},
|
| 126 |
+
],
|
| 127 |
+
});
|
| 128 |
+
|
| 129 |
+
const createBeginChunk = () => ({
|
| 130 |
+
id: "chatcmpl-123",
|
| 131 |
+
object: "chat.completion.chunk",
|
| 132 |
+
created: getCurrentDate(),
|
| 133 |
+
model: "gpt-4",
|
| 134 |
+
system_fingerprint: "",
|
| 135 |
+
choices: [
|
| 136 |
+
{
|
| 137 |
+
index: 0,
|
| 138 |
+
delta: { role: "assistant", content: "" },
|
| 139 |
+
logprobs: null,
|
| 140 |
+
finish_reason: null,
|
| 141 |
+
},
|
| 142 |
+
],
|
| 143 |
+
});
|
| 144 |
+
|
| 145 |
+
const createMessageChunk = (text) => ({
|
| 146 |
+
id: "chatcmpl-123",
|
| 147 |
+
object: "chat.completion.chunk",
|
| 148 |
+
created: getCurrentDate(),
|
| 149 |
+
model: "gpt-4",
|
| 150 |
+
system_fingerprint: "",
|
| 151 |
+
choices: [
|
| 152 |
+
{
|
| 153 |
+
index: 0,
|
| 154 |
+
delta: { content: text },
|
| 155 |
+
logprobs: null,
|
| 156 |
+
finish_reason: null,
|
| 157 |
+
},
|
| 158 |
+
],
|
| 159 |
+
});
|
| 160 |
+
|
| 161 |
+
const createEndChunk = () => ({
|
| 162 |
+
id: "chatcmpl-123",
|
| 163 |
+
object: "chat.completion.chunk",
|
| 164 |
+
created: getCurrentDate(),
|
| 165 |
+
model: "gpt-4",
|
| 166 |
+
system_fingerprint: "",
|
| 167 |
+
choices: [{ index: 0, delta: {}, logprobs: null, finish_reason: "stop" }],
|
| 168 |
+
});
|
| 169 |
+
|
| 170 |
+
function createFakeModelsList() {
|
| 171 |
+
return [
|
| 172 |
+
{
|
| 173 |
+
id: "gpt-4",
|
| 174 |
+
object: "model",
|
| 175 |
+
created: getCurrentDate(),
|
| 176 |
+
owned_by: "openai",
|
| 177 |
+
permission: [
|
| 178 |
+
{
|
| 179 |
+
id: "modelperm-gpt-4",
|
| 180 |
+
object: "model_permission",
|
| 181 |
+
created: getCurrentDate(),
|
| 182 |
+
organization: "*",
|
| 183 |
+
group: null,
|
| 184 |
+
is_blocking: false,
|
| 185 |
+
},
|
| 186 |
+
],
|
| 187 |
+
root: "gpt-4",
|
| 188 |
+
parent: null,
|
| 189 |
+
},
|
| 190 |
+
];
|
| 191 |
+
}
|