aidio-description/dist/providers/vision/ollamaVisionProvider.js

"use strict";
var __importDefault = (this && this.__importDefault) || function (mod) {
    return (mod && mod.__esModule) ? mod : { "default": mod };
};
Object.defineProperty(exports, "__esModule", { value: true });
exports.OllamaVisionProvider = void 0;
const fs_1 = __importDefault(require("fs"));
const axios_1 = __importDefault(require("axios"));
/**
 * Ollama Vision Provider Implementation
 * See: https://github.com/ollama/ollama/blob/main/docs/api.md
 */
class OllamaVisionProvider {
    constructor(config) {
        this.config = config;
        this.axiosInstance = axios_1.default.create({
            baseURL: config.baseUrl || "http://localhost:11434",
            headers: { "Content-Type": "application/json" }
        });
    }
    /**
     * Describe a single image
     * @param imagePath - Path to the image file
     * @param prompt - Prompt for the AI
     * @returns Description and usage stats
     */
    async describeImage(imagePath, prompt) {
        try {
            const imageData = fs_1.default.readFileSync(imagePath);
            const base64Image = imageData.toString('base64');
            const response = await this.axiosInstance.post('/api/generate', {
                model: this.config.model,
                prompt: prompt,
                images: [base64Image],
                stream: false,
                options: {
                    max_tokens: this.config.maxTokens || 300,
                    temperature: 0.1
                }
            });
            const combinedText = response.data.response || "";
            return {
                description: combinedText.trim(),
                usage: {
                    inputTokens: 0,
                    outputTokens: 0,
                    totalTokens: 0
                }
            };
        }
        catch (error) {
            console.error("Ollama describeImage error:", error);
            return {
                description: "Unable to describe this image.",
                usage: { inputTokens: 0, outputTokens: 0, totalTokens: 0 }
            };
        }
    }
    /**
     * Compare two images and describe differences
     * @param image1Path - Path to the first image
     * @param image2Path - Path to the second image
     * @param prompt - Prompt for the AI
     * @returns Description and usage stats
     */
    async compareImages(image1Path, image2Path, prompt) {
        try {
            const image1Data = fs_1.default.readFileSync(image1Path).toString('base64');
            const image2Data = fs_1.default.readFileSync(image2Path).toString('base64');
            const response = await this.axiosInstance.post('/api/generate', {
                model: this.config.model,
                prompt: prompt,
                images: [image1Data, image2Data],
                stream: false,
                options: {
                    max_tokens: this.config.maxTokens || 300,
                    temperature: 0.2
                }
            });
            const combinedText = response.data.response || "";
            return {
                description: combinedText.trim(),
                usage: { inputTokens: 0, outputTokens: 0, totalTokens: 0 }
            };
        }
        catch (error) {
            console.error("Ollama compareImages error:", error);
            return {
                description: "Unable to describe the differences.",
                usage: { inputTokens: 0, outputTokens: 0, totalTokens: 0 }
            };
        }
    }
    /**
     * Describe a batch of images
     * @param imagePaths - Array of paths to the images
     * @param lastBatchContext - Context from the previous batch (optional)
     * @param prompt - Prompt for the AI
     * @returns Description and usage stats
     */
    async describeBatch(imagePaths, lastBatchContext, prompt) {
        try {
            let userPrompt = prompt;
            // If there's context, prepend it. This helps maintain a storyline across batches.
            if (lastBatchContext && lastBatchContext.lastDescription) {
                userPrompt = `Previous batch summary: ${lastBatchContext.lastDescription}\n\n${prompt}`;
            }
            // Convert images to base64
            const imagesBase64 = imagePaths.map(fp => {
                const imageData = fs_1.default.readFileSync(fp);
                return imageData.toString('base64');
            });
            const response = await this.axiosInstance.post('/api/generate', {
                model: this.config.model,
                prompt: userPrompt,
                images: imagesBase64,
                stream: false,
                options: {
                    max_tokens: this.config.maxTokens || 300,
                    temperature: 0.2
                }
            }, {
                timeout: 120000 // Timeout in milliseconds, e.g., 5000 ms = 5 seconds
            });
            const combinedText = response.data.response || "";
            return {
                description: combinedText.trim(),
                usage: { inputTokens: 0, outputTokens: 0, totalTokens: 0 }
            };
        }
        catch (error) {
            console.error("Ollama describeBatch error:", error);
            return {
                description: "Unable to describe this batch of images.",
                usage: { inputTokens: 0, outputTokens: 0, totalTokens: 0 }
            };
        }
    }
}
exports.OllamaVisionProvider = OllamaVisionProvider;
//# sourceMappingURL=ollamaVisionProvider.js.map