File size: 7,359 Bytes
c2c8c8d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
da1152a
c2c8c8d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
da1152a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c2c8c8d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
import axios, { AxiosInstance, AxiosError } from 'axios';
import { GLMMessage, GLMResponse, GLMStreamResponse, GLM_DEFAULTS } from '@glmpilot/shared';
import { config } from '../config/env.js';
import { logger } from '../utils/logger.js';
import { stripCodeFences } from '../utils/code-parser.js';
import { countTokens } from '../utils/token-counter.js';

export class GLMClient {
  private client: AxiosInstance;
  private model: string;
  private concurrentRequests = 0;
  private maxConcurrent = GLM_DEFAULTS.MAX_CONCURRENT_REQUESTS;
  private waitQueue: Array<() => void> = [];

  constructor() {
    this.model = config.GLM_MODEL || GLM_DEFAULTS.MODEL;
    const baseUrl = (config.GLM_BASE_URL || GLM_DEFAULTS.BASE_URL).trim().replace(/\/?$/, '/');

    this.client = axios.create({
      baseURL: baseUrl,
      timeout: GLM_DEFAULTS.TIMEOUT_MS,
      headers: {
        'Content-Type': 'application/json',
        Authorization: `Bearer ${config.GLM_API_KEY}`,
      },
    });
  }

  async chat(
    messages: GLMMessage[],
    options: { temperature?: number; max_tokens?: number; top_p?: number } = {}
  ): Promise<GLMResponse> {
    await this.waitForSlot();
    try {
      const body = {
        model: this.model,
        messages,
        temperature: options.temperature ?? GLM_DEFAULTS.TEMPERATURE,
        max_tokens: options.max_tokens ?? GLM_DEFAULTS.MAX_TOKENS,
        top_p: options.top_p ?? GLM_DEFAULTS.TOP_P,
        stream: false,
      };

      const fullUrl = `${this.client.defaults.baseURL}chat/completions`;
      logger.info('GLM request starting', { 
        url: fullUrl,
        messageCount: messages.length, 
        tokenEstimate: countTokens(messages.map(m => m.content).join('')),
      });

      const result = await this.retryWithBackoff(async () => {
        const response = await this.client.post<GLMResponse>('chat/completions', body);
        return response.data;
      }, GLM_DEFAULTS.MAX_RETRIES);

      logger.info('GLM response', { status: 'success', usage: result.usage });
      return result;
    } finally {
      this.releaseSlot();
    }
  }

  async *chatStream(
    messages: GLMMessage[],
    options: { temperature?: number; max_tokens?: number } = {},
    signal?: AbortSignal
  ): AsyncGenerator<string> {
    await this.waitForSlot();
    try {
      const body = {
        model: this.model,
        messages,
        temperature: options.temperature ?? GLM_DEFAULTS.TEMPERATURE,
        max_tokens: options.max_tokens ?? GLM_DEFAULTS.MAX_TOKENS,
        top_p: GLM_DEFAULTS.TOP_P,
        stream: true,
      };

      const response = await this.client.post('chat/completions', body, {
        responseType: 'stream',
        signal,
      });

      let buffer = '';
      for await (const chunk of response.data) {
        buffer += chunk.toString();
        const lines = buffer.split('\n');
        buffer = lines.pop() || '';

        for (const line of lines) {
          const trimmed = line.trim();
          if (!trimmed || !trimmed.startsWith('data: ')) continue;
          const data = trimmed.slice(6);
          if (data === '[DONE]') return;

          try {
            const parsed: GLMStreamResponse = JSON.parse(data);
            const content = parsed.choices?.[0]?.delta?.content;
            if (content) yield content;
          } catch {
            // Skip malformed chunks
          }
        }
      }
    } catch (err) {
      if (signal?.aborted) return;
      const axiosErr = err as import('axios').AxiosError;
      if (axiosErr.response) {
        // For streaming requests, response.data is a Node.js Readable stream.
        // Read it to get the actual GLM API error body.
        try {
          const chunks: Buffer[] = [];
          for await (const chunk of axiosErr.response.data as AsyncIterable<Buffer>) {
            chunks.push(Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk));
          }
          const body = Buffer.concat(chunks).toString('utf-8');
          throw new Error(`GLM API ${axiosErr.response.status}: ${body}`);
        } catch (readErr) {
          if ((readErr as Error).message.startsWith('GLM API')) throw readErr;
        }
      }
      throw err;
    } finally {
      this.releaseSlot();
    }
  }

  async analyzeCode(code: string, instruction: string, systemPrompt: string): Promise<string> {
    const messages: GLMMessage[] = [
      { role: 'system', content: systemPrompt },
      { role: 'user', content: `${instruction}\n\n\`\`\`\n${code}\n\`\`\`` },
    ];

    const response = await this.chat(messages, {
      temperature: 0.2,
      max_tokens: GLM_DEFAULTS.MAX_TOKENS,
    });

    return response.choices[0]?.message?.content || '';
  }

  async completeCode(
    prefix: string,
    suffix: string,
    language: string,
    fileContext?: string
  ): Promise<string> {
    const systemPrompt = `You are an expert ${language} code completion engine. You will be given code with a cursor position marked as <CURSOR>. Output ONLY the code that should be inserted at the cursor position. Do not include any explanation, markdown formatting, or the surrounding code. Output raw code only.`;

    const userContent = fileContext
      ? `${fileContext}\n\n// Current file:\n${prefix}<CURSOR>${suffix}`
      : `${prefix}<CURSOR>${suffix}`;

    const messages: GLMMessage[] = [
      { role: 'system', content: systemPrompt },
      { role: 'user', content: userContent },
    ];

    const response = await this.chat(messages, {
      temperature: 0.3,
      max_tokens: 256,
    });

    const content = response.choices[0]?.message?.content || '';
    return stripCodeFences(content);
  }

  private async waitForSlot(): Promise<void> {
    if (this.concurrentRequests < this.maxConcurrent) {
      this.concurrentRequests++;
      return;
    }
    return new Promise((resolve) => {
      this.waitQueue.push(() => {
        this.concurrentRequests++;
        resolve();
      });
    });
  }

  private releaseSlot(): void {
    this.concurrentRequests--;
    const next = this.waitQueue.shift();
    if (next) next();
  }

  private async retryWithBackoff<T>(fn: () => Promise<T>, maxRetries: number): Promise<T> {
    const delays = GLM_DEFAULTS.RETRY_DELAYS;
    for (let attempt = 0; attempt <= maxRetries; attempt++) {
      try {
        return await fn();
      } catch (error) {
        const axiosErr = error as AxiosError;
        const status = axiosErr.response?.status;

        if (axiosErr.response) {
          logger.error('GLM API Error Response:', {
            status: axiosErr.response.status,
            data: axiosErr.response.data,
          });
        }

        if (status && status >= 400 && status < 500 && status !== 429) {
          throw error;
        }

        if (attempt === maxRetries) throw error;

        let delay = delays[attempt] || 4000;
        if (status === 429) {
          const retryAfter = axiosErr.response?.headers?.['retry-after'];
          if (retryAfter) delay = parseInt(retryAfter, 10) * 1000;
        }

        logger.warn(`GLM request failed (attempt ${attempt + 1}/${maxRetries + 1}), retrying in ${delay}ms`, {
          status,
          message: axiosErr.message,
        });

        await new Promise((r) => setTimeout(r, delay));
      }
    }
    throw new Error('Retry exhausted');
  }
}

export const glm = new GLMClient();