All files / api/src/ai-providers/ollama chat-complete.ts

96.12% Statements 124/129
91.66% Branches 11/12
100% Functions 3/3
96.12% Lines 124/129

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 1471x             1x 1x   1x   1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 2x 2x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x   1x 3x 3x 3x 3x 3x 3x 3x 1x 1x 1x 1x 1x   2x 2x   3x   1x             1x 1x 1x 1x 1x 1x 1x 1x 1x 1x   1x 1x   1x 1x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x   4x 1x 1x   3x 3x 3x 3x 3x 3x 3x 3x 3x 3x  
import { generateInvalidProviderResponseError } from '@api/utils/ai-provider';
import type {
  AIProviderFunctionConfig,
  ResponseChunkStreamTransformFunction,
  ResponseTransformFunction,
} from '@shared/types/ai-providers/config';
import type { ChatCompletionRequestBody } from '@shared/types/api/routes/chat-completions-api';
import { ChatCompletionMessageRole } from '@shared/types/api/routes/shared/messages';
import { AIProvider } from '@shared/types/constants';
import type { OllamaChatCompleteResponse, OllamaStreamChunk } from './types';
import { ollamaErrorResponseTransform } from './utils';
 
export const ollamaChatCompleteConfig: AIProviderFunctionConfig = {
  model: {
    param: 'model',
    required: true,
    default: 'llama3.2:latest',
  },
  messages: {
    param: 'messages',
    default: '',
    transform: (params: ChatCompletionRequestBody) => {
      return params.messages?.map((message) => {
        if (message.role === ChatCompletionMessageRole.DEVELOPER)
          return { ...message, role: ChatCompletionMessageRole.SYSTEM };
        return message;
      });
    },
  },
  frequency_penalty: {
    param: 'frequency_penalty',
    min: -2,
    max: 2,
  },
  presence_penalty: {
    param: 'presence_penalty',
    min: -2,
    max: 2,
  },
  response_format: {
    param: 'response_format',
  },
  seed: {
    param: 'seed',
  },
  stop: {
    param: 'stop',
  },
  stream: {
    param: 'stream',
    default: false,
  },
  temperature: {
    param: 'temperature',
    default: 1,
    min: 0,
    max: 2,
  },
  top_p: {
    param: 'top_p',
    default: 1,
    min: 0,
    max: 1,
  },
  max_tokens: {
    param: 'max_tokens',
    default: 100,
    min: 0,
  },
  max_completion_tokens: {
    param: 'max_tokens',
    default: 100,
    min: 0,
  },
  tools: {
    param: 'tools',
  },
};
 
export const ollamaChatCompleteResponseTransform: ResponseTransformFunction = (
  aiProviderResponseBody,
  aiProviderResponseStatus,
  _aiProviderResponseHeaders,
  _strictOpenAiCompliance,
  _raRequestData,
) => {
  if (aiProviderResponseStatus !== 200 && 'error' in aiProviderResponseBody) {
    return ollamaErrorResponseTransform(
      aiProviderResponseBody,
      aiProviderResponseStatus,
    );
  }
 
  const response =
    aiProviderResponseBody as unknown as OllamaChatCompleteResponse;
 
  if ('choices' in response) {
    // Construct usage object from Ollama's token counts if usage is not provided
    const usage = response.usage ?? {
      prompt_tokens: response.prompt_eval_count ?? 0,
      completion_tokens: response.eval_count ?? 0,
      total_tokens:
        (response.prompt_eval_count ?? 0) + (response.eval_count ?? 0),
    };
 
    return {
      id: response.id,
      object: response.object,
      created: response.created,
      model: response.model,
      provider: AIProvider.OLLAMA,
      choices: response.choices,
      usage,
    };
  }
 
  return generateInvalidProviderResponseError(response, AIProvider.OLLAMA);
};
 
export const ollamaChatCompleteStreamChunkTransform: ResponseChunkStreamTransformFunction =
  (
    responseChunk,
    _fallbackId,
    _streamState,
    _strictOpenAiCompliance,
    _raRequestData,
  ) => {
    const chunk = responseChunk
      .trim()
      .replace(/^data: /, '')
      .trim();
 
    if (chunk === '[DONE]') {
      return `data: ${chunk}\n\n`;
    }
 
    const parsedChunk: OllamaStreamChunk = JSON.parse(chunk);
    return `data: ${JSON.stringify({
      id: parsedChunk.id,
      object: parsedChunk.object,
      created: parsedChunk.created,
      model: parsedChunk.model,
      provider: AIProvider.OLLAMA,
      choices: parsedChunk.choices,
    })}\n\n`;
  };