naisys/dist/features/media/genImg.js

NAISYS - Autonomous AI agent runner with built-in context management and cost tracking

import OpenAI from "openai";
import path from "path";
import sharp from "sharp";
import stringArgv from "string-argv";
import { genImgCmd } from "../../command/commandDefs.js";
import * as pathService from "../../services/runtime/pathService.js";
export function createGenImg({ globalConfig }, { agentConfig }, costTracker, output, getImageModel) {
    /** genimg "<description>" <filepath>: Generate an image with the description and save it to the file path */
    async function handleCommand(args) {
        // genimg sholdn't even be presented as an available command unless it is defined in the config
        if (!agentConfig().imageModel) {
            throw "Agent config: Error, 'imageModel' is not defined";
        }
        const argv = stringArgv(args);
        // args already has the command name stripped, so argv[0] is the description
        const description = argv[0];
        const filepath = argv[1] || "";
        if (!description) {
            throw "Invalid parameters: Description in quotes and fully qualified filepath with desired image extension are required";
        }
        if (!filepath) {
            throw "Error: Filepath is required";
        }
        // Validate path is fully qualified (Unix or Windows)
        if (!path.isAbsolute(filepath)) {
            throw "Error: Filepath must be fully qualified";
        }
        pathService.ensureFileDirExists(filepath);
        const imageModelName = agentConfig().imageModel;
        if (!imageModelName) {
            throw "Error: imageModel is not defined in agent config";
        }
        output.commentAndLog(`Generating image with ${imageModelName}...`);
        const model = getImageModel(imageModelName);
        const apiKey = model.apiKeyVar
            ? globalConfig().variableMap[model.apiKeyVar]
            : undefined;
        if (!apiKey) {
            throw `Error, set ${model.apiKeyVar} variable`;
        }
        const openai = new OpenAI({
            apiKey,
            baseURL: model.baseUrl,
        });
        const isDalle = model.versionName.startsWith("dall-e");
        const response = await openai.images.generate({
            prompt: description,
            model: model.versionName,
            size: model.size,
            quality: model.quality,
            ...(isDalle ? { response_format: "b64_json" } : {}),
        });
        // save to filepath
        if (!response.data || response.data.length === 0) {
            throw "Error: No image data returned from OpenAI";
        }
        const base64Image = response.data[0].b64_json;
        if (!base64Image) {
            throw 'Error: "b64_json" not found in response';
        }
        // Convert the base64 string to a buffer
        const imageBuffer = Buffer.from(base64Image, "base64");
        // Use sharp to convert the buffer and save it as a JPG file
        const fileExtension = path.extname(filepath).substring(1);
        await sharp(imageBuffer)
            /*.resize(512, 512, {
            fit: "inside",
          })*/
            .toFormat(fileExtension)
            .toFile(filepath);
        // Record the cost
        costTracker.recordCost(model.cost, "genimg", model.key);
        return "1024x1024 Image generated and saved to " + filepath;
    }
    const registrableCommand = {
        command: genImgCmd,
        handleCommand,
    };
    return {
        ...registrableCommand,
    };
}
//# sourceMappingURL=genImg.js.map