@vertesia/workflow 1.0.0-dev.20260203.130115Z → 1.0.0-dev.20260225.024852Z

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vertesia/workflow",
-  "version": "1.0.0-dev.20260203.130115Z",
+  "version": "1.0.0-dev.20260225.024852Z",
   "type": "module",
   "description": "Vertesia workflow DSL",
   "main": "./lib/esm/index.js",
@@ -30,7 +30,6 @@
     "@aws-sdk/credential-providers": "^3.948.0",
     "@temporalio/activity": "^1.11.5",
     "@temporalio/workflow": "^1.11.5",
-    "@types/json-schema": "^7.0.15",
     "fast-deep-equal": "^3.1.3",
     "jsonwebtoken": "^9.0.3",
     "mime": "^4.0.0",
@@ -44,11 +43,11 @@
     "tmp": "^0.2.4",
     "tmp-promise": "^3.0.3",
     "yaml": "^2.6.0",
-    "@llumiverse/common": "1.0.0-dev.20260202.145450Z",
-    "@vertesia/api-fetch-client": "1.0.0-dev.20260203.130115Z",
-    "@vertesia/client": "1.0.0-dev.20260203.130115Z",
-    "@vertesia/common": "1.0.0-dev.20260203.130115Z",
-    "@vertesia/memory": "1.0.0-dev.20260203.130115Z"
+    "@vertesia/api-fetch-client": "1.0.0-dev.20260225.024852Z",
+    "@llumiverse/common": "1.0.0-dev.20260224.234313Z",
+    "@vertesia/common": "1.0.0-dev.20260225.024852Z",
+    "@vertesia/client": "1.0.0-dev.20260225.024852Z",
+    "@vertesia/memory": "1.0.0-dev.20260225.024852Z"
   },
   "ts_dual_module": {
     "outDir": "lib",

package/src/activities/advanced/createOrUpdateDocumentFromInteractionRun.ts

@@ -90,7 +90,7 @@ export async function createOrUpdateDocumentFromInteractionRun(payload: DSLActiv
         generation_run_info: {
             id: run.id,
             date: new Date().toISOString(),
-            model: run.modelId,
+            model: run.modelId ?? "",
             target: jsonResult ? 'properties' : 'text'
         }
     };

package/src/activities/chunkDocument.ts

@@ -55,7 +55,9 @@ export async function chunkDocument(payload: DSLActivityExecutionPayload<ChunkDo
 
     const document = await client.objects.retrieve(objectId, "+text");
 
-    const type = document.type ? await client.types.retrieve(document.type.id) : undefined;
+    const type = document.type
+        ? await client.types.catalog.resolve(document.type)
+        : undefined;
 
     if (!type?.is_chunkable) {
         log.warn('Type is not chunkable for object ID: ' + objectId);

package/src/activities/extractDocumentText.ts

@@ -4,6 +4,7 @@ import {
     CreateContentObjectPayload,
     DSLActivityExecutionPayload,
     DSLActivitySpec,
+    WorkflowInputFile,
 } from "@vertesia/common";
 import { markdownWithMarkitdown } from "../conversion/markitdown.js";
 import { mutoolPdfToText } from "../conversion/mutool.js";
@@ -12,6 +13,10 @@ import { setupActivity } from "../dsl/setup/ActivityContext.js";
 import { DocumentNotFoundError } from "../errors.js";
 import { TextExtractionResult, TextExtractionStatus } from "../result-types.js";
 import { fetchBlobAsBuffer, md5 } from "../utils/blobs.js";
+import {
+    createFileSourceResult,
+    uploadTextPreviewToStorage
+} from "../utils/text-preview-utils.js";
 import { countTokens } from "../utils/tokens.js";
 
 //@ts-ignore
@@ -19,8 +24,9 @@ const JSON: DSLActivitySpec = {
     name: "extractDocumentText",
 };
 
-// doesn't have any own param
-export interface ExtractDocumentTextParams {}
+export interface ExtractDocumentTextParams {
+    output_storage_path?: string;
+}
 export interface ExtractDocumentText extends DSLActivitySpec<ExtractDocumentTextParams> {
     name: "extractDocumentText";
     projection?: never;
@@ -29,8 +35,27 @@ export interface ExtractDocumentText extends DSLActivitySpec<ExtractDocumentText
 export async function extractDocumentText(
     payload: DSLActivityExecutionPayload<ExtractDocumentTextParams>,
 ): Promise<TextExtractionResult> {
-    const { client, objectId } = await setupActivity(payload);
+    const context = await setupActivity(payload);
+    const { client, inputType, params } = context;
+    const { output_storage_path } = params;
+
+    if (inputType === 'files') {
+        // File mode: extract from file source
+        if (!output_storage_path) {
+            throw new Error('output_storage_path is required when extracting text from file sources');
+        }
+        return extractFromFileSource(client, context.file, output_storage_path);
+    } else {
+        // Object mode: fetch from object store
+        return extractFromObject(client, context.objectId, context.objectIds || []);
+    }
+}
 
+async function extractFromObject(
+    client: any,
+    objectId: string,
+    objectIds: string[],
+): Promise<TextExtractionResult> {
     const r = await client.objects.find({
         query: { _id: objectId },
         limit: 1,
@@ -39,7 +64,7 @@ export async function extractDocumentText(
     const doc = r[0] as ContentObject;
     if (!doc) {
         log.error(`Document ${objectId} not found`);
-        throw new DocumentNotFoundError(`Document ${objectId} not found`, payload.objectIds);
+        throw new DocumentNotFoundError(`Document ${objectId} not found`, objectIds);
     }
 
     log.info(`Extracting text for object ${doc.id}`);
@@ -65,10 +90,60 @@ export async function extractDocumentText(
         return createResponse(doc, "", TextExtractionStatus.error, e.message);
     }
 
-    let txt: string;
+    const txt = await extractTextFromBuffer(fileBuffer, doc.content.type);
+    if (!txt) {
+        return createResponse(
+            doc,
+            doc.text ?? "",
+            TextExtractionStatus.skipped,
+            `Unsupported mime type: ${doc.content.type}`,
+        );
+    }
+
+    const tokensData = countTokens(txt);
+    const etag = doc.content.etag ?? md5(txt);
+
+    const updateData: CreateContentObjectPayload = {
+        text: txt,
+        text_etag: etag,
+        tokens: {
+            ...tokensData,
+            etag: etag,
+        },
+    };
+
+    await client.objects.update(doc.id, updateData);
+
+    return createResponse(doc, txt, TextExtractionStatus.success);
+}
+
+async function extractFromFileSource(
+    client: any,
+    input_file: WorkflowInputFile,
+    output_storage_path: string
+): Promise<TextExtractionResult> {
+    log.info(`Extracting text from ${input_file}`);
 
-    // Normalize mime type by stripping charset and other parameters (e.g., "text/html; charset=UTF-8" -> "text/html")
-    const mimeType = doc.content.type.split(';')[0].trim();
+    let fileBuffer: Buffer;
+    try {
+        fileBuffer = await fetchBlobAsBuffer(client, input_file.url);
+    } catch (e: any) {
+        log.error(`Error reading file: ${e}`);
+        return createFileSourceResult(input_file.url, output_storage_path, null);
+    }
+
+    const txt = await extractTextFromBuffer(fileBuffer, input_file.mimetype);
+
+    // Upload extracted text to storage
+    if (txt && output_storage_path) {
+        await uploadTextPreviewToStorage(client, txt, output_storage_path, "Document");
+    }
+
+    return createFileSourceResult(input_file.url, output_storage_path, txt);
+}
+
+async function extractTextFromBuffer(fileBuffer: Buffer, mimeType: string): Promise<string | null> {
+    let txt: string;
 
     switch (mimeType) {
         case "application/pdf":
@@ -134,29 +209,10 @@ export async function extractDocumentText(
                 txt = fileBuffer.toString("utf8"); //TODO: add charset detection
                 break;
             }
-            return createResponse(
-                doc,
-                doc.text ?? "",
-                TextExtractionStatus.skipped,
-                `Unsupported mime type: ${doc.content.type}`,
-            );
+            return null;
     }
 
-    const tokensData = countTokens(txt);
-    const etag = doc.content.etag ?? md5(txt);
-
-    const updateData: CreateContentObjectPayload = {
-        text: txt,
-        text_etag: etag,
-        tokens: {
-            ...tokensData,
-            etag: etag,
-        },
-    };
-
-    await client.objects.update(doc.id, updateData);
-
-    return createResponse(doc, txt, TextExtractionStatus.success);
+    return txt;
 }
 
 function createResponse(

package/src/activities/generateDocumentProperties.ts

@@ -30,7 +30,9 @@ export async function generateDocumentProperties(
     const project = await context.fetchProject();
 
     const doc = await client.objects.retrieve(objectId, "+text");
-    const type = doc.type ? await client.types.retrieve(doc.type.id) : undefined;
+    const type = doc.type
+        ? await client.types.catalog.resolve(doc.type)
+        : undefined;
 
     if (!doc?.text && !params.use_vision && !doc?.content?.type?.startsWith("image/")) {
         log.warn(`Object ${objectId} not found or text is empty`);
@@ -112,7 +114,7 @@ export async function generateDocumentProperties(
         generation_run_info: {
             id: infoRes.id,
             date: new Date().toISOString(),
-            model: infoRes.modelId,
+            model: infoRes.modelId ?? "",
         },
     }, { suppressWorkflows: true });
 

package/src/activities/generateEmbeddings.ts

@@ -72,7 +72,7 @@ export async function generateEmbeddings(
     }
 
     if (!projectData?.configuration.embeddings[type]?.enabled) {
-        log.info(
+        log.debug(
             `Embeddings generation disabled for type ${type} on project: ${projectData.name} (${projectData.namespace})`,
             { config },
         );
@@ -83,7 +83,7 @@ export async function generateEmbeddings(
         };
     }
 
-    log.info(`${type} embedding generation starting for object ${objectId}`, {
+    log.debug(`${type} embedding generation starting for object ${objectId}`, {
         force,
         config,
     });
@@ -203,7 +203,7 @@ async function generateTextEmbeddings(
     // Skip if embeddings already exist with matching etag (unless force=true)
     const existingEmbedding = document.embeddings?.[type];
     if (!force && existingEmbedding?.etag && textEtag && existingEmbedding.etag === textEtag) {
-        log.info(`Skipping ${type} embeddings for document ${document.id} - etag unchanged`);
+        log.debug(`Skipping ${type} embeddings for document ${document.id} - etag unchanged`);
         return {
             id: document.id,
             type,
@@ -221,7 +221,7 @@ async function generateTextEmbeddings(
     const maxTokens = config.max_tokens ?? 8000;
 
     //generate embeddings for the main doc if document isn't too large
-    log.info(`Generating ${type} embeddings for document ${document.id}`);
+    log.debug(`Generating ${type} embeddings for document ${document.id}`);
     if (
         type === SupportedEmbeddingTypes.text &&
         tokenCount !== undefined &&
@@ -237,7 +237,7 @@ async function generateTextEmbeddings(
             message: `${type} embeddings generation, skipped for large document (${tokenCount} tokens)`,
         }
     } else {
-        log.info(`Generating ${type} embeddings for document`);
+        log.debug(`Generating ${type} embeddings for document`);
 
         const res = await generateEmbeddingsFromStudio(
             JSON.stringify(document[type]),
@@ -252,7 +252,7 @@ async function generateTextEmbeddings(
             };
         }
 
-        log.info(`${type} embeddings generated for document ${document.id}`, {
+        log.debug(`${type} embeddings generated for document ${document.id}`, {
             len: res.values.length,
         });
         await client.objects.setEmbedding(document.id, type, {
@@ -277,7 +277,7 @@ async function generateImageEmbeddings({
     config,
     force,
 }: ExecuteGenerateEmbeddingsParams) {
-    log.info("Generating image embeddings for document " + document.id, {
+    log.debug("Generating image embeddings for document " + document.id, {
         content: document.content,
     });
     if (
@@ -298,7 +298,7 @@ async function generateImageEmbeddings({
     // Skip if embeddings already exist with matching etag (unless force=true)
     const existingEmbedding = document.embeddings?.[type];
     if (!force && existingEmbedding?.etag && contentEtag && existingEmbedding.etag === contentEtag) {
-        log.info(`Skipping ${type} embeddings for document ${document.id} - content etag unchanged`);
+        log.debug(`Skipping ${type} embeddings for document ${document.id} - content etag unchanged`);
         return {
             id: document.id,
             type,
@@ -378,7 +378,7 @@ async function generateEmbeddingsFromStudio(
     client: VertesiaClient,
     model?: string,
 ): Promise<EmbeddingsResult> {
-    log.info(
+    log.debug(
         `Generating embeddings for text of ${text.length} chars with environment ${env}`,
     );
 

package/src/activities/generateOrAssignContentType.ts

@@ -52,7 +52,7 @@ export async function generateOrAssignContentType(
   const interactionName =
     params.interactionNames?.selectDocumentType ?? INT_SELECT_DOCUMENT_TYPE;
 
-  log.info("SelectDocumentType for object: " + objectId, { payload });
+  log.debug("SelectDocumentType for object: " + objectId, { payload });
 
   const object = await client.objects.retrieve(objectId, "+text");
 
@@ -80,7 +80,7 @@ export async function generateOrAssignContentType(
     return { status: "failed", error: "no-text" };
   }
 
-  const types = await client.types.list(undefined, {
+  const types = await client.types.catalog.list({
     schema: true,
   });
 

package/src/activities/index-dsl.ts

@@ -3,6 +3,7 @@ export { createDocumentTypeFromInteractionRun } from "./advanced/createDocumentT
 export { createOrUpdateDocumentFromInteractionRun } from "./advanced/createOrUpdateDocumentFromInteractionRun.js";
 export { updateDocumentFromInteractionRun } from "./advanced/updateDocumentFromInteractionRun.js";
 export { chunkDocument } from "./chunkDocument.js";
+export { copyParentArtifacts } from "./copyParentArtifacts.js";
 export { createPdfDocumentFromSource } from "./createDocumentFromOther.js";
 export { executeInteraction } from "./executeInteraction.js";
 export { extractDocumentText } from "./extractDocumentText.js";
@@ -11,16 +12,16 @@ export { generateEmbeddings } from "./generateEmbeddings.js";
 export { generateOrAssignContentType } from "./generateOrAssignContentType.js";
 export { getObjectFromStore } from "./getObjectFromStore.js";
 export { handleDslError } from "./handleError.js";
-export { prepareVideo } from "./media/prepareVideo.js";
 export { prepareAudio } from "./media/prepareAudio.js";
+export { prepareVideo } from "./media/prepareVideo.js";
 export { convertPdfToStructuredText } from "./media/processPdfWithTextract.js";
 export { saveGladiaTranscription } from "./media/saveGladiaTranscription.js";
 export { transcribeMedia } from "./media/transcribeMediaWithGladia.js";
 export type { TranscriptMediaResult } from "./media/transcribeMediaWithGladia.js";
+export { mergeChildArtifacts } from "./mergeChildArtifacts.js";
 export { notifyWebhook } from "./notifyWebhook.js";
 export { checkRateLimit } from "./rateLimiter.js";
-export { copyParentArtifacts } from "./copyParentArtifacts.js";
-export { mergeChildArtifacts } from "./mergeChildArtifacts.js";
 export { generateImageRendition } from "./renditions/generateImageRendition.js";
 export { generateVideoRendition } from "./renditions/generateVideoRendition.js";
 export { setDocumentStatus } from "./setDocumentStatus.js";
+