genkitx-mistral/lib/ocr.mjs

Firebase Genkit AI framework plugin for Mistral AI APIs.

import {
  __async,
  __spreadProps,
  __spreadValues
} from "./chunk-MLCSNVBT.mjs";
import { GenerationCommonConfigSchema, z } from "genkit";
import { modelRef } from "genkit/model";
const ImageURLSchema = z.object({
  url: z.string(),
  detail: z.string().nullable().optional()
});
const OCRSchema = GenerationCommonConfigSchema.extend({
  document: z.object({
    type: z.enum(["document_url", "image_url"]).optional(),
    documentUrl: z.string(),
    documentName: z.string().nullable().optional(),
    imageUrl: z.union([z.string(), ImageURLSchema]).optional()
  })
});
const ocr = modelRef({
  name: "mistral/ocr",
  info: {
    versions: ["mistral-ocr-latest"],
    label: "Mistral - OCR",
    supports: {
      media: false,
      output: ["text", "json"],
      multiturn: false,
      systemRole: false,
      tools: false
    }
  },
  configSchema: OCRSchema
});
function toOCRRequest(request) {
  var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k;
  const docType = ((_b = (_a = request.config) == null ? void 0 : _a.document) == null ? void 0 : _b.type) || "document_url";
  const options = {
    model: "mistral-ocr-latest",
    document: docType === "document_url" ? {
      type: "document_url",
      documentUrl: ((_d = (_c = request.config) == null ? void 0 : _c.document) == null ? void 0 : _d.documentUrl) || "",
      documentName: ((_f = (_e = request.config) == null ? void 0 : _e.document) == null ? void 0 : _f.documentName) || null
    } : {
      type: "image_url",
      imageUrl: typeof ((_h = (_g = request.config) == null ? void 0 : _g.document) == null ? void 0 : _h.imageUrl) === "string" ? request.config.document.imageUrl : ((_k = (_j = (_i = request.config) == null ? void 0 : _i.document) == null ? void 0 : _j.imageUrl) == null ? void 0 : _k.url) || ""
    }
  };
  Object.keys(options).forEach((k) => {
    if (options[k] === void 0) {
      delete options[k];
    }
  });
  return options;
}
function toGenerateResponse(result) {
  const candidates = result.pages.map(
    (page, index) => ({
      index,
      finishReason: "stop",
      message: {
        role: "model",
        content: [
          {
            text: page.markdown
          },
          ...page.images.map((image) => ({
            media: {
              contentType: "image/png",
              url: `data:image/png;base64,${image.imageBase64}`
            }
          }))
        ]
      }
    })
  );
  return { candidates };
}
function OCRModel(ai, client) {
  return ai.defineModel(
    __spreadProps(__spreadValues({
      name: ocr.name
    }, ocr.info), {
      configSchema: ocr.configSchema
    }),
    (request) => __async(this, null, function* () {
      const result = yield client.ocr.process(toOCRRequest(request));
      return toGenerateResponse(result);
    })
  );
}
export {
  OCRModel,
  ocr
};
//# sourceMappingURL=ocr.mjs.map