All files ollama-client.ts

87.65% Statements 71/81
82.05% Branches 32/39
71.42% Functions 5/7
88.46% Lines 69/78

Press n or j to go to the next uncovered block, b, p or k for the previous block.

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                                          2x     28x     28x     28x 28x 28x               10x   10x         10x 10x                                           10x 1x     9x 1x       8x 8x 1x     7x   7x 7x   7x 7x 7x 7x 7x 7x 7x 7x 12x 12x 7x     7x 7x 7x 7x 61x 61x 61x 9x   9x 1x   8x                 53x   1x 1x 1x       52x 52x       52x 1x         51x         2x 2x     2x   7x 7x                     5x 5x       5x                           5x   5x 1x     4x   4x 4x 4x             21x 3x 3x        
import { OllamaMessage, OllamaTool, ToolCall } from './types';
 
interface FetchResponse {
  ok: boolean;
  status: number;
  headers?: {
    get: (name: string) => string | null;
  };
  body?: {
    getReader: () => ReadableStreamDefaultReader<Uint8Array>;
  } | null;
  json?: () => Promise<any>;
}
 
interface FetchOptions {
  method: string;
  headers: Record<string, string>;
  body: string;
  signal?: AbortSignal;
}
 
export class OllamaClient {
  private url: string;
  private model: string;
  private abortController: AbortController | null = null;
 
  // Mock fetch function for testing
  private fetchFn: typeof fetch = fetch;
 
  constructor(url: string, model: string, fetchFn?: typeof fetch) {
    this.url = url;
    this.model = model;
    if (fetchFn) this.fetchFn = fetchFn;
  }
 
  async streamChat(
    messages: OllamaMessage[],
    tools: OllamaTool[],
    timeoutMs: number = 60000
  ): Promise<AsyncIterable<{ content: string; tool_calls?: ToolCall[] }>> {
    this.abortController = new AbortController();
 
    const timeoutId = setTimeout(() => {
      this.abortController?.abort();
    }, timeoutMs);
 
    let response: FetchResponse;
    try {
      response = await this.fetchFn(`${this.url}/api/chat`, {
        method: 'POST',
        headers: {
          'Content-Type': 'application/json',
        },
        body: JSON.stringify({
          model: this.model,
          messages,
          tools,
          stream: true,
        }),
        signal: this.abortController.signal,
      } as FetchOptions);
    } catch (fetchError: any) {
      clearTimeout(timeoutId);
      this.abortController = null;
      Iif (fetchError.name === 'AbortError' || fetchError.code === 'ABORT_ERR') {
        throw new Error('Request timeout while connecting to Ollama');
      }
      throw fetchError;
    }
 
    if (!response.ok) {
      throw new Error(`Ollama API error: ${response.status}`);
    }
 
    if (!response.body) {
      throw new Error('No response body');
    }
 
    // Validate response structure
    const contentType = response.headers?.get('content-type');
    if (!contentType?.match(/application\/(x-ndjson|json)/)) {
      throw new Error('Invalid response format');
    }
 
    const reader = response.body.getReader();
 
    const self = this;
    return {
      [Symbol.asyncIterator]: async function* () {
        const decoder = new TextDecoder();
        let buffer = '';
        let chunkCount = 0;
        let skippedChunks = 0;
        const maxChunks = 1000; // Safety limit
        const maxSkipped = 50; // Fail if too many chunks are malformed
        try {
          while (true) {
            const { done, value } = await reader.read();
            if (done) break;
            Iif (++chunkCount > maxChunks) {
              throw new Error('Response too long, stopped streaming');
            }
            buffer += decoder.decode(value, { stream: true });
            const lines = buffer.split('\n');
            buffer = lines.pop() || '';
            for (const line of lines) {
              Iif (line.trim() === '') continue;
              try {
                const data = JSON.parse(line);
                if (data.message && typeof data.message === 'object') {
                  // Validate message structure
                  if (data.message.error && typeof data.message.error === 'string') {
                    throw new Error(`Ollama error: ${data.message.error}`);
                  }
                  yield {
                    content: data.message.content || '',
                    tool_calls: Array.isArray(data.message.tool_calls)
                      ? data.message.tool_calls
                      : [],
                  };
                }
              } catch (parseError) {
                // Check if this is an Ollama error (thrown intentionally) vs a parse error
                if (parseError instanceof Error && parseError.message.startsWith('Ollama error:')) {
                  // This is an intentional Ollama error, re-throw it
                  self.abortController = null;
                  reader.releaseLock();
                  throw parseError;
                }
 
                // This is a parse error, skip the malformed chunk
                skippedChunks++;
                console.warn(
                  `[OllamaClient] Skipped malformed chunk ${skippedChunks}/${maxSkipped}:`,
                  parseError instanceof Error ? parseError.message : String(parseError)
                );
                if (skippedChunks > maxSkipped) {
                  throw new Error(
                    `Too many malformed response chunks (${skippedChunks}). Connection may be degraded.`
                  );
                }
                // Skip invalid chunks but continue streaming
                continue;
              }
            }
          }
        } catch (streamError) {
          self.abortController = null;
          Iif (streamError instanceof Error && streamError.name === 'AbortError') {
            throw new Error('Streaming request was cancelled');
          }
          throw streamError;
        } finally {
          reader.releaseLock();
          self.abortController = null;
        }
      },
    };
  }
 
  async chat(
    messages: OllamaMessage[],
    tools: OllamaTool[],
    timeoutMs: number = 30000
  ): Promise<{ content: string; tool_calls?: ToolCall[] }> {
    const abortController = new AbortController();
    const timeoutId = setTimeout(() => {
      abortController.abort();
    }, timeoutMs);
 
    const response = await this.fetchFn(`${this.url}/api/chat`, {
      method: 'POST',
      headers: {
        'Content-Type': 'application/json',
      },
      body: JSON.stringify({
        model: this.model,
        messages,
        tools,
        stream: false,
      }),
      signal: abortController.signal,
    } as FetchOptions);
 
    clearTimeout(timeoutId);
 
    if (!response.ok) {
      throw new Error(`Ollama API error: ${response.status}`);
    }
 
    const responseData = await response.json();
 
    const data = responseData;
    const messageData = data.message;
    return {
      content: messageData?.content || '',
      tool_calls: messageData?.tool_calls || [],
    };
  }
 
  cancelStream(): void {
    if (this.abortController) {
      this.abortController.abort();
      this.abortController = null;
    }
  }
}