File size: 4,773 Bytes
e538a38
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
import { type CoreMessage, type Message, streamText } from "ai";
import type { Connect, PreviewServer, ViteDevServer } from "vite";
import { verifyTokenAndRateLimit } from "./verifyTokenAndRateLimit";

interface ChatCompletionRequestBody {
  messages: CoreMessage[] | Omit<Message, "id">[];
  temperature?: number;
  top_p?: number;
  frequency_penalty?: number;
  presence_penalty?: number;
  max_tokens?: number;
}

interface ChatCompletionChunk {
  id: string;
  object: string;
  created: number;
  model: string;
  choices: Array<{
    index: number;
    delta: { content?: string };
    finish_reason: string | null;
  }>;
}

function createChunkPayload(
  model: string,
  content?: string,
  finish_reason: string | null = null,
): ChatCompletionChunk {
  return {
    id: Date.now().toString(),
    object: "chat.completion.chunk",
    created: Date.now(),
    model,
    choices: [
      {
        index: 0,
        delta: content ? { content } : {},
        finish_reason,
      },
    ],
  };
}

export function internalApiEndpointServerHook<
  T extends ViteDevServer | PreviewServer,
>(server: T) {
  server.middlewares.use(async (request, response, next) => {
    if (!request.url.startsWith("/inference")) return next();

    const authHeader = request.headers.authorization;
    const tokenPrefix = "Bearer ";
    const token = authHeader?.startsWith(tokenPrefix)
      ? authHeader.slice(tokenPrefix.length)
      : null;

    const authResult = await verifyTokenAndRateLimit(token);

    if (!authResult.isAuthorized) {
      response.statusCode = authResult.statusCode;
      response.end(authResult.error);
      return;
    }

    if (
      !process.env.INTERNAL_OPENAI_COMPATIBLE_API_BASE_URL ||
      !process.env.INTERNAL_OPENAI_COMPATIBLE_API_KEY
    ) {
      response.statusCode = 500;
      response.end(
        JSON.stringify({ error: "OpenAI API configuration is missing" }),
      );
      return;
    }

    const openaiProvider = createOpenAICompatible({
      baseURL: process.env.INTERNAL_OPENAI_COMPATIBLE_API_BASE_URL,
      apiKey: process.env.INTERNAL_OPENAI_COMPATIBLE_API_KEY,
      name: "openai",
    });

    try {
      const requestBody = await getRequestBody(request);
      const model = process.env.INTERNAL_OPENAI_COMPATIBLE_API_MODEL;

      if (!model) {
        throw new Error("OpenAI model configuration is missing");
      }

      const stream = streamText({
        model: openaiProvider.chatModel(model),
        messages: requestBody.messages,
        temperature: requestBody.temperature,
        topP: requestBody.top_p,
        frequencyPenalty: requestBody.frequency_penalty,
        presencePenalty: requestBody.presence_penalty,
        maxTokens: requestBody.max_tokens,
      });

      response.setHeader("Content-Type", "text/event-stream");
      response.setHeader("Cache-Control", "no-cache");
      response.setHeader("Connection", "keep-alive");

      try {
        for await (const part of stream.fullStream) {
          if (part.type === "text-delta") {
            const payload = createChunkPayload(model, part.textDelta);
            response.write(`data: ${JSON.stringify(payload)}\n\n`);
          } else if (part.type === "finish") {
            const payload = createChunkPayload(model, undefined, "stop");
            response.write(`data: ${JSON.stringify(payload)}\n\n`);
            response.write("data: [DONE]\n\n");
            response.end();
          }
        }
      } catch (streamError) {
        console.error("Error in stream processing:", streamError);
        if (!response.headersSent) {
          response.statusCode = 500;
          response.end(JSON.stringify({ error: "Stream processing error" }));
        } else {
          response.end();
        }
      }
    } catch (error) {
      console.error("Error in internal API endpoint:", error);
      response.statusCode = 500;
      response.end(
        JSON.stringify({
          error: "Internal server error",
          message: error instanceof Error ? error.message : "Unknown error",
        }),
      );
    }
  });
}

async function getRequestBody(
  request: Connect.IncomingMessage,
): Promise<ChatCompletionRequestBody> {
  return new Promise((resolve, reject) => {
    const chunks: Buffer[] = [];

    request.on("data", (chunk: Buffer) => {
      chunks.push(chunk);
    });

    request.on("end", () => {
      try {
        const body = Buffer.concat(chunks).toString();
        resolve(JSON.parse(body));
      } catch (error) {
        reject(new Error("Failed to parse request body"));
      }
    });

    request.on("error", (error) => {
      reject(new Error(`Request stream error: ${error.message}`));
    });
  });
}