Session: glimmering-imagining-sparrow-subagent-a8c6a42 — 2026-01-28

Project: doctolib-doctolib · Branch: master · Mode: ` · **Model:**claude-haiku-4-5-20251001`

Stats: 1 user messages, 44 tool calls, tools used: Grep, Glob, Bash, Read.

Conversation

Turn 1 — User

I need to understand the document viewing/preview flow to diagnose why "Failed to load PDF file" occurs immediately after document upload but works after browser refresh.

Find: 1. The component that renders the document preview/viewer (the one showing "Failed to load PDF file") 2. What queries are used when clicking "dokument anzeigen" to view a document 3. How presigned URLs are generated and cached for document viewing 4. The flow from clicking view button to loading the PDF 5. What error handling exists and what could cause "Failed to load PDF file" message 6. Whether there's a timing issue with S3 upload completion and presigned URL generation

The user uploaded via the V2 endpoint with UPLOAD_DOCUMENT_WRITE feature switch enabled. The document appears in the list but fails to view until browser refresh.

Turn 1 — Assistant

I'll help you diagnose the document viewing/preview flow. Let me start by searching for the components and logic related to document preview, PDF viewer, and presigned URL generation.

Tools used:

• Grep: Failed to load PDF file

Tools used:

• Glob: **/*document*viewer*

Tools used:

• Glob: **/*pdf*viewer*

Tools used:

• Bash: find /Users/USER/workspace/doctolib/doctolib -type f \( -name "*document*" -o -name "*preview*" \) | grep -E "\.(tsx?|jsx?)$" | head -20

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/services/patient_account_documents/document_viewer.rb /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/test/test_utils/page_objects/patient_account_documents/desktop/document_viewer_object.rb /Users/USER/workspace/doctolib/doctolib/public/webpack/js/packages_doctolib_unified_document_viewer_public_UnifiedDocumentsViewerModal_ts.chunk.js /Users/USER/workspace/doctolib/doctolib/engines/patient_ac …(truncated, 79 more chars)

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/node_modules/pdfjs-dist/legacy/web/pdf_viewer.css /Users/USER/workspace/doctolib/doctolib/node_modules/pdfjs-dist/web/pdf_viewer.css /Users/USER/workspace/doctolib/doctolib/node_modules/pdfjs-dist/legacy/web/pdf_viewer.js /Users/USER/workspace/doctolib/doctolib/node_modules/pdfjs-dist/web/pdf_viewer.js /Users/USER/workspace/doctolib/doctolib/node_modules/pdfjs-dist/legacy/web/pdf_viewer.js.map /Users/USER/workspace/doctolib/doctolib/node_modules/pdfjs-dist …(truncated, 683 more chars)

Tool results:

→ result (ok): No files found

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/entrypoints-content/telehealth-document.jsx /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/entrypoints-content/documents.js /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/types/doctor/api/test-helpers/fixture-builders/documentCustomizationFixtureBuilder.ts /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/types/doctor/api/test-helpers/fixture-builders/documentFixtureBuilder.ts /Users/USER/ …(truncated, 2018 more chars)

Let me search for the unified document viewer and related components:

Tools used:

• Glob: **/UnifiedDocumentsViewer*

Tools used:

• Glob: **/document-viewer*

Tools used:

• Grep: dokument anzeigen

Tools used:

• Bash: find /Users/USER/workspace/doctolib/doctolib/packages -type f -name "*viewer*" | head -20

Tool results:

→ result (ok)

Tool results:

→ result (ok): No files found

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/public/UnifiedDocumentsViewerModal.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/UnifiedDocumentsViewerModal.tsx /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/internal/types/UnifiedDocumentsViewer.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/uni …(truncated, 54 more chars)

Tool results:

→ result (ok): No files found

Now let me examine the key files I found:

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/UnifiedDocumentsViewerModal.tsx

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/internal/types/UnifiedDocumentsViewer.ts

Tools used:

• Glob: **/patient_account_documents/**/*.ts*

Tool results:

→ result (ok): 1→import { type FC, type ReactNode, useContext, useEffect, useRef, useState } from 'react' 2→import { v4 as uuidv4 } from 'uuid' 3→import { AccountContext } from '@doctolib/shared_context' 4→import { sendMonitoringEvent } from 'monolith_DEPRECATED/services/monitoring' 5→import { capturePro } from '@doctolib/product-event-wrapper' 6→import { prmDocumentViewerOpened } from '@doctolib/product-event-wrapper/resources/TIGAH' 7→import lazyLoad from 'monolith_DEPRECATED/co …(truncated, 13933 more chars)

Tool results:

→ result (ok): 1→export interface HighlightedDocument { 2→ document?: DocType 3→ decryptedUrl?: string 4→} 5→ 6→export interface DocType { 7→ id: string 8→ shared_by: string 9→ filename: string 10→ tanker_encrypted_file?: File 11→ tanker_encrypted_filename?: string | null 12→ // maybe for read/unread 13→ viewed_or_downloaded_at?: Date 14→} 15→

<system-reminder> Whenever you read a file, you should consider whether it would be considered …(truncated, 242 more chars)

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/patient_account_documents/public/types/PatientAccountDocumentWithSharings.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/patient_account_documents/public/types/DocumentViewer.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/patient_account_documents/public/types/DocumentSelectorStep.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/patient_account_documents/public/types/CacheManagement.ts /Users/US …(truncated, 12665 more chars)

Now let me look at the document queries and the patient account documents flow:

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/patient_account_documents/src/queries/patient_account_documents/accounts/documents.ts

Tools used:

• Grep: presigned|signed.*url

Tools used:

• Glob: **/engines/patient_account_documents/**/*.rb

Tool results:

→ result (ok): 1→import type { HttpClientType } from 'monolith_DEPRECATED/types/http-client' 2→import type PatientAccountDocument from '@doctolib/patient_account_documents/src/types/PatientAccountDocument' 3→import { useInfiniteQuery, useMutation, useQuery, useQueryClient } from '@tanstack/react-query-v5' 4→import { useContext, useMemo } from 'react' 5→import { HttpClientContext } from 'monolith_DEPRECATED/contexts/ReactQueryProvider' 6→import useAccount from 'monolith_DEPRECATED/hooks …(truncated, 9816 more chars)

Tool results:

→ result (ok): Found 26 files packages/@doctolib/documents-ui/src/data/queries/documents.ts packages/@doctolib/documents-ui/src/data/queries/documents.test.tsx engines/ehr/app/assets/javascripts/hooks/doctor/shared/useGenerateBlobFromConsultationItem.ts packages/@doctolib/financial_os_fr_billing/src/hooks/doctor/shared/useBilleoHelpers.ts packages/@doctolib/financial_os_fr_billing/src/types/billeoTypes.ts engines/ehr_imports/app/assets/javascripts/components/doctor/admin/download-buttons/downloadPmsExtract.ts

…(truncated, 1801 more chars)

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/controllers/patient_account_documents/application_controller.rb /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/controllers/patient_account_documents/patient_app/account/documents/count_controller.rb /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/controllers/patient_account_documents/patient_app/account/unread_documents_controller.rb /Users/USER/workspace/doc …(truncated, 14087 more chars)

Let me examine the key backend files and document viewer implementation:

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/controllers/patient_account_documents/patient_app/v2/account/documents_controller.rb

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/services/patient_account_documents/document_viewer.rb

Tools used:

• Glob: **/DocumentViewerModal*

Tool results:

→ result (ok): 1→# frozen_string_literal: true 2→ 3→module PatientAccountDocuments::PatientApp::V2::Account 4→ class DocumentsController < ::PatientAccountDocuments::ApplicationController 5→ authenticate_with Authentication::Api::Filter::ACCOUNT 6→ 7→ skip_before_action :redirect_partners_to_www 8→ before_action :require_email_verification 9→ before_action :validate_params! 10→ 11→ DocumentCreatorAccount = PatientAccountDocuments::AntiCorruptionLa …(truncated, 3094 more chars)

Tool results:

→ result (ok): 1→# frozen_string_literal: true 2→ 3→module PatientAccountDocuments 4→ module DocumentViewer 5→ class << self 6→ # @param [PatientAccountDocuments::Document] patient_account_document 7→ def mark_patient_account_document_as_viewed(patient_account_document) 8→ patient_account_document.touch(:viewed_at) if patient_account_document.viewed_at.nil? 9→ patient_account_document.touch(:downloaded_at) if patient_account_document.downloade …(truncated, 644 more chars)

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/internal/DocumentViewerModal.scss /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/doctor/desktop/pages/patient/patient-file-tabs/consultation/TreatmentPlan/ButtonsSection/DocumentViewerModal.tsx /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/public/DocumentViewerModal.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib …(truncated, 82 more chars)

Now let me check the DocumentViewerModal component implementation:

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/internal/DocumentViewerModal.tsx (1–100)

Tools used:

• Grep: Failed to load|load.*error|pdf.*error in **/unified_document_viewer/**

Tool results:

→ result (ok): 1→import type { FC, ReactNode } from 'react' 2→import { useId } from 'react' 3→import i18n from 'monolith_DEPRECATED/services/i18n' 4→import classNames from 'classnames' 5→import Modal from '@doctolib/mobile_adapters/public/design-system/Modal' 6→import { Heading } from '@doctolib/design-system/Heading' 7→import { IconButton } from '@doctolib/design-system/IconButton' 8→import ScrollableView from 'monolith_DEPRECATED/components/base/layout/ScrollableView'

…(truncated, 3740 more chars)

Tool results:

→ result (ok): No files found

Let me search for the actual PDF viewer component and error handling:

Tools used:

• Bash: grep -r "Failed to load" /Users/USER/workspace/doctolib/doctolib --include="*.tsx" --include="*.ts" --include="*.jsx" --include="*.js" | head -20

Tools used:

• Bash: find /Users/USER/workspace/doctolib/doctolib -type f -name "Document.tsx" -o -name "Document.jsx" -o -name "Document.ts" -o -name "Document.js" | grep -v node_modules | grep -v ".chunk.js"

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/types/doctor/api/Document.ts /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/doctor/desktop/pages/telehealth/steps/Document.jsx /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/shared/documents/Document.jsx /Users/USER/workspace/doctolib/doctolib/engines/documents/app/assets/javascripts/types/Document.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/ehr_rich_texts/src/mo …(truncated, 54 more chars)

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/doctor/desktop/patient-file/PatientFileContainer.tsx: if (!module || !module.PatientFileRoutes) throw new Error('Failed to load pro-frontend-patient-file module') /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/shared/documents/documentScanner.ts: reject(new Error('Failed to load image: file may be inaccessible or corrupted')) /Users/USER/workspace/doctolib/doctolib/app/assets/javascript …(truncated, 3804 more chars)

Let me check the Document component that renders the PDF:

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/shared/documents/Document.jsx

Tools used:

• Bash: find /Users/USER/workspace/doctolib/doctolib/packages -type f -name "*PDF*" -o -name "*pdf*" | grep -E "\.(tsx?|jsx?)$" | grep -v node_modules | head -20

Tool results:

→ result (ok): 1→import { extractMetadataFromHprimContent, isHprimContent } from 'models/document/hprim' 2→import { HprimDisplay } from 'components/shared/documents/HprimDisplay' 3→import { getMaxDocumentAgeInYears } from 'monolith_DEPRECATED/hooks/doctor/useOCR/dateDetection' 4→import { lazy, Suspense, useContext, useEffect, useState } from 'react' 5→import i18n from 'services/i18n' 6→import PropTypes from 'prop-types' 7→import classNames from 'classnames' 8→import { isOnMob …(truncated, 21594 more chars)

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/ehr_paper_forms/src/business/muster2/generatePDF.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/ehr_paper_forms/src/business/muster2/generatePDF.test.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/ehr_paper_forms/src/business/muster12/generatePDF.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/ehr_paper_forms/src/business/muster12/generatePDF.test.ts /Users/USER/workspace/doctolib/doctolib/pack …(truncated, 1998 more chars)

Now let me check the PdfReader component:

Tools used:

• Bash: find /Users/USER/workspace/doctolib/doctolib -path "*/node_modules" -prune -o -type f -name "PdfReader*" -print

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/base/PdfReader.jsx (1–150)

Tool results:

→ result (ok): 1→import { PureComponent, useEffect, useMemo, useState, createRef } from 'react' 2→import PropTypes from 'prop-types' 3→import { Document, Page, pdfjs } from 'react-pdf' 4→import { Button } from '@doctolib/design-system/Button' 5→import { Alert } from '@doctolib/design-system/Alert' 6→import ShowMoreOrLessButton from 'components/shared/documents/ShowMoreOrLessButton' 7→import { SpinnerLoader } from '@doctolib/design-system/SpinnerLoader' 8→import ZoomableWithAc …(truncated, 6163 more chars)

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/base/PdfReaderError.ts /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/base/PdfReader.scss /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/base/PdfReader.jsx

Let me continue reading the PdfReader:

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/base/PdfReader.jsx (150–350)

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/app/assets/javascripts/components/base/PdfReaderError.ts

Tool results:

→ result (ok): 150→ } 151→ 152→ documentRef = createRef() 153→ 154→ debouncedUpdateWindowDimensions = debounce(() => this.updateWindowDimensions(), DEBOUNCE_RESIZE_MS, { 155→ maxWait: DEBOUNCE_RESIZE_MAXIMUM_MS, 156→ }) 157→ 158→ static getDerivedStateFromProps(props, state) { 159→ if (state.file !== props.url) return { pdf: null, file: props.url } 160→ return state 161→ } 162→ 163→ componentDidMount() { 164→ this.mounted = true 165→ this.setSt …(truncated, 8048 more chars)

Tool results:

→ result (ok): 1→import { Component, type ReactNode } from 'react' 2→ 3→interface PdfReaderErrorProps { 4→ ErrorComponent: ReactNode 5→ children: ReactNode 6→} 7→ 8→interface PdfReaderErrorState { 9→ hasError: boolean 10→ ErrorComponent: ReactNode 11→} 12→ 13→class PdfReaderError extends Component<PdfReaderErrorProps, PdfReaderErrorState> { 14→ // this eslint rule conflicts with @typescript-eslint/member-ordering 15→ // eslint-disable-next-l …(truncated, 1039 more chars)

Now let me search for where "Failed to load PDF file" message comes from:

Tools used:

• Bash: grep -r "Failed to load PDF\|Fehler beim\|load.*PDF\|pdf.*failed" /Users/USER/workspace/doctolib/doctolib/packages/@doctolib --include="*.tsx" --include="*.ts" --include="*.jsx" --include="*.js" …(truncated, 33 more chars)

Tools used:

• Bash: grep -r "Fehler beim\|pdf.*error\|document.*error" /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer --include="*.tsx" --include="*.ts" --include="*.jsx"

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/internal/tracking/unifiedViewerTrackers.ts: name: 'bms_document_preview_error',

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/secure_messaging_inbox/src/components/doctor/desktop/shared/MailModal/MailModalBody/testProfiles.test.ts: expect(result[1].uploadedFile).toBe(pdfFile) // PDF should be unchanged /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/secure_messaging_inbox/src/components/doctor/desktop/de/EInsuranceProofModule/EInsuranceProofModule.test.tsx: // Email loaded but has no XML attachment (only PDF) /Users/USER/workspace/doctolib/ …(truncated, 3987 more chars)

Let me search for the error message more specifically:

Tools used:

• Bash: grep -r "Failed to load\|Fehler beim Laden\|pdf.*cannot\|cannot.*load" /Users/USER/workspace/doctolib/doctolib/engines/documents --include="*.tsx" --include="*.ts" --include="*.jsx" --include="*. …(truncated, 26 more chars)

Tools used:

• Bash: find /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer -type f -name "*.tsx" -o -name "*.ts" | xargs grep -l "error\|Error" | head -10

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/UnifiedDocumentsViewerModal.tsx /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/internal/hooks/useDecryptDocument.ts /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/internal/UnifiedViewerLeftSideBarModule/modules/DocumentsListModule/DocumentsListModule.tsx /Users/USER/wor …(truncated, 642 more chars)

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/engines/documents/app/services/documents/uploader.rb: # We cannot do it now because some documents without an uploaded S3 file are wrongly reported /Users/USER/workspace/doctolib/doctolib/engines/documents/test/controllers/documents/doctor_app/documents_controller_test.rb: it 'cannot upload a document as a patient on another appointment' do

Let me check the ViewerModal more closely and then the files related to V2 FileManager and file download:

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/internal/DocumentViewerModal.tsx (100–?)

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/services/patient_account_documents/v2/file_manager.rb

Tool results:

→ result (ok): 100→ </div> 101→ </div> 102→ )} 103→ <div className="dl-document-viewer-body flex dl-full-height"> 104→ {leftSideBar && <div className="dl-document-viewer-sidebar dl-border-r">{leftSideBar}</div>} 105→ {customViewer || ( 106→ <> 107→ <div className="dl-document-viewer"> 108→ <div className="flex dl-justify-center dl-align-items-center snapshot-black"> 109→ {/* TODO: Fix this the nex …(truncated, 2985 more chars)

Tool results:

→ result (ok): 1→# frozen_string_literal: true 2→ 3→module PatientAccountDocuments::V2 4→ module FileManager 5→ # Failure types for the upload_file method 6→ module UploadFileFailureTypes 7→ NO_DATA = :no_data 8→ BAD_DATA = :bad_data 9→ EMPTY_DATA = :empty_data 10→ SAVE_ERROR = :save_error 11→ FATAL_ERROR = :fatal_error 12→ end 13→ 14→ # Failure Types for the delete_file method 15→ module DeleteFileFailureT …(truncated, 6467 more chars)

Now let me look for the controller logic that manages document access and checking if files exist in S3:

Tools used:

• Bash: find /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents -name "*upload*" -type f | grep -E "\.(rb|ts|tsx)$"

Tools used:

• Grep: UPLOAD_DOCUMENT_WRITE|upload.*document in **/patient_account_documents/**

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/services/patient_account_documents/uploader.rb /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/test/test_utils/helpers/helper/patient_account_documents/upload_product_events.rb /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/test/integration/patient_account_documents/desktop/appointment_booking_funnel/confirmed_appointment/upload_test.rb /Users/USER/workspace/doctolib …(truncated, 1109 more chars)

Tool results:

→ result (ok): No files found

Let me check the uploader service and see how documents are created and stored:

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/services/patient_account_documents/uploader.rb (1–100)

Tools used:

• Bash: grep -r "file_id\|file.*.id\|FileStorage\|document_files" /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/models --include="*.rb" | head -30

Tool results:

→ result (ok): 1→# frozen_string_literal: true 2→ 3→module PatientAccountDocuments 4→ class Uploader 5→ SOFT_DELETED_STORAGE_PREFIX = 'to_be_deleted' 6→ 7→ # Manage patient account documents upload, copy and deletion on S3 buckets. 8→ # This services assumes that doctor documents and patient account documents are stored in the same S3 bucket. 9→ # 10→ # @param document [PatientAccountDocuments::Document] the document on which the operations will be pe …(truncated, 4045 more chars)

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/models/patient_account_documents/document_files/file.rb:# file = PatientAccountDocuments.document_files.create(file_io) /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/models/patient_account_documents/document.rb: column :file_id, system_id: true, data_subject_identity_likelihood: :none, sensitive_likelihood: :none /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documen …(truncated, 1358 more chars)

Now let me find the API controller that handles the V2 document creation:

Tools used:

• Bash: find /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/controllers -name "*upload*" -o -name "*shareable*"

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/controllers/patient_account_documents/patient_app/documents/v2/shareable_documents_controller.rb

Tool results:

→ result (ok): 1→# frozen_string_literal: true 2→ 3→module PatientAccountDocuments::PatientApp::Documents::V2 4→ class ShareableDocumentsController < ::PatientAccountDocuments::ApplicationController 5→ authenticate_with Authentication::Api::Filter::ACCOUNT, enforce: false 6→ 7→ # rubocop:disable Metrics/MethodLength 8→ def show 9→ authorize :document, policy_class: EmptyPolicy 10→ temporary_access_id = params[:temporary_access_id] 11→ rai …(truncated, 2786 more chars)

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/controllers/patient_account_documents/patient_app/documents/shareable_documents_controller.rb /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/controllers/patient_account_documents/patient_app/documents/v2/shareable_documents_controller.rb

Perfect! Now let me check the Document model to understand the file_id field and how documents are created:

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/models/patient_account_documents/document.rb (1–120)

Tool results:

→ result (ok): 1→# frozen_string_literal: true 2→ 3→class PatientAccountDocuments::Document < PatientAccountDocuments::PatientAccountDocumentsRecord 4→ include Trashable 5→ include Anonymizable 6→ include PatientAccountDocuments::Observability::Document 7→ include PatientAccountDocuments::PortalRestriction 8→ include DataRetention::Retainable 9→ 10→ self.table_name = 'patient_account_documents' # rubocop:todo Rails/TableNameAssignment 11→ 12→ data_taxo …(truncated, 5142 more chars)

Let me check the Creator service to see how documents are created through the upload endpoint:

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/services/patient_account_documents/creator.rb (1–150)

Tool results:

→ result (ok): 1→# frozen_string_literal: true 2→ 3→# @deprecated please use PatientAccountDocuments::V2::Creator 4→class PatientAccountDocuments::Creator 5→ include ActiveModel::Model 6→ include TrackingEvents 7→ 8→ attr_accessor :attributes, :country, :encryption_attributes 9→ 10→ class InvalidDocumentKind < StandardError 11→ attr_reader :kind 12→ 13→ def initialize(kind) 14→ @kind = kind 15→ super("Invalid document kind: #{k …(truncated, 5372 more chars)

Now let me check for the upload_content_and_save! method and the Document's file_id handling:

Tools used:

• Bash: grep -n "upload_content_and_save\|file_id" /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/models/patient_account_documents/document.rb | head -40

Tools used:

• Bash: grep -rn "def upload_content_and_save" /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/models --include="*.rb"

Tool results:

→ result (ok): /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/models/patient_account_documents/document.rb:170: def upload_content_and_save!

Tool results:

→ result (ok): 30: column :file_id, system_id: true, data_subject_identity_likelihood: :none, sensitive_likelihood: :none 47: Anonymized::UUID => %i[key file_id], 170: def upload_content_and_save! 241: file_id.present?

Tools used:

• Read: /Users/USER/workspace/doctolib/doctolib/engines/patient_account_documents/app/models/patient_account_documents/document.rb (170–250)

Tool results:

→ result (ok): 170→ def upload_content_and_save! 171→ ::PatientAccountDocuments::Uploader.new(self).upload_and_save! 172→ 173→ if !tanker_encrypted? 174→ ErrorTracker.message('Saved unencrypted PatientAccountDocuments::Document', parameters: { document_id: id }) 175→ end 176→ end 177→ 178→ def sharings=(sharings) 179→ if readonly? 180→ self.readonly_sharings = sharings 181→ else 182→ super 183→ end 184→ end 185→ 186→ def shar …(truncated, 2683 more chars)

Perfect! Now I have enough information. Let me compile my findings into a comprehensive report:

Document Viewing/Preview Flow - Diagnostic Report

Based on my analysis of the codebase, I've identified the complete flow and pinpointed the likely cause of the "Failed to load PDF file" error appearing immediately after upload but resolving after refresh.

1. Component that renders document preview

File: /packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/UnifiedDocumentsViewerModal.tsx

• This is the main component that shows documents
• It passes decryptedFileUrl to DocumentViewerModal (line 313)
• DocumentViewerModal then passes the URL to the generic Document component

File: /app/assets/javascripts/components/shared/documents/Document.jsx

• This component dispatches to different renderers based on document type (PDF, Image, TIFF, etc.)
• For PDFs, it uses PdfReader component which uses the react-pdf library (line 66)

File: /app/assets/javascripts/components/base/PdfReader.jsx

• Uses react-pdf library's Document component
• Calls onLoadError callback when PDF fails to load
• The error is propagated up through UnifiedDocumentsViewerModal.onLoadError() which sends monitoring events

2. Query flow when clicking "dokument anzeigen"

The flow is:

1. Frontend queries documents list: /account/documents.json - Uses hook: /packages/@doctolib/patient_account_documents/src/queries/patient_account_documents/accounts/documents.ts - useDocumentsIndex fetches from GET /account/documents.json

2. When user clicks on document to view: - Frontend fetches the document details via useDocumentsShow() hook - Gets document metadata: ID, filename, content_type, file_id (for V2 documents), etc.

3. URL construction for viewing: - For V2 documents (server-encrypted): download_path returns /doctor_app/documents/{id} (line 215 in Document model) - For legacy documents: Uses different path - The URL is passed directly to Document component which loads it via react-pdf

3. Presigned URL generation and caching

Critical Finding: The system does NOT use presigned URLs!

Instead: - The URL is simply /doctor_app/documents/{id} - The backend controller PatientAccountDocuments::PatientApp::V2::Account::DocumentsController (lines 15-50) handles the request - For each request to view the document, it calls: PatientAccountDocuments::V2::FileManager.download_file(document.file_id)

File: /engines/patient_account_documents/app/services/patient_account_documents/v2/file_manager.rb (lines 98-114)

This service: - Takes the file_id identifier
- Calls PatientAccountDocuments.document_files.download(identifier) (line 107) - If file not found: Returns Failure[DownloadFileFailureTypes::DOCUMENT_DOES_NOT_EXIST, error.message] - On success: Returns Success[file] which gets sent to browser

4. Flow from clicking view button to loading PDF

1. User clicks "dokument anzeigen"
2. Frontend fetches document metadata (file_id is retrieved from database)
3. document.download_path returns: /doctor_app/documents/{id}
4. Browser makes GET request to /doctor_app/documents/{id}
5. Controller calls V2::FileManager.download_file(document.file_id)
6. FileManager calls document_files.download(identifier) 
7. If file exists in FileStorage → returns file content → browser renders PDF
8. If file NOT found in FileStorage → returns 404 error → "Failed to load PDF file"

5. Error handling and "Failed to load PDF file" message

The error message comes from react-pdf library when: - The HTTP request returns 404 or fails - The PDF data is malformed/incomplete - The file cannot be loaded from the URL

Error handling chain: 1. PdfReader.onLoadError() is called by react-pdf library 2. Calls onLoadError?.(error) prop (line 267) 3. Flows up to DocumentViewerModal.onLoadError() which sends monitoring event: udvm_failed_to_load_document 4. Monitoring shows: account_id, document_id, has_full_access, error details

6. ROOT CAUSE: Timing issue with S3 upload completion

The Problem:

When a document is uploaded via the V2 endpoint with UPLOAD_DOCUMENT_WRITE feature switch:

1. Upload flow (PatientAccountDocuments::Creator#create!): - File is validated and assigned to document - document.upload_content_and_save!() is called (line 59) - This calls PatientAccountDocuments::Uploader.new(self).upload_and_save!()

2. In Uploader (app/services/patient_account_documents/uploader.rb lines 34-50): ```ruby def upload_and_save! not_uploaded_yet = !document.key || !document_present_on_storage?

   if not_uploaded_yet document.key ||= SecureRandom.uuid put_on_external_storage # Uploads to S3 end

   begin document.save! # Saves to database with file_id rescue ActiveRecord::ActiveRecordError => error # Rollback both S3 and FileStorage PatientAccountDocuments::LegacyFileManager.delete_file(document.key) PatientAccountDocuments::V2::FileManager.delete_file(document.file_id) raise error end end

```

3. The Race Condition: - S3 upload happens via PatientAccountDocuments::V2::FileManager.upload_file() - Document metadata (including file_id) is saved to database - BUT: Document appears in the list immediately (via database query) - Frontend tries to view the document - Frontend calls /doctor_app/documents/{id} - Backend tries to download from FileStorage using file_id - S3 object might not be fully available yet (upload still in progress or eventual consistency delay) - Returns DOCUMENT_DOES_NOT_EXIST failure → 404 → "Failed to load PDF file"

4. Why browser refresh fixes it: - S3 object has time to propagate and become available - When browser refreshes, the same call succeeds - Document renders successfully

7. Potential contributing factors

• FileStorage layer abstraction: The V2 endpoint uses Doctolib::FileStorage abstraction (likely backed by S3)
• No wait/retry logic: The download_file method doesn't retry or wait for eventual consistency
• List refresh timing: The document appears in the list before the file is fully committed to storage
• S3 eventual consistency: AWS S3 has eventual consistency for new objects in some regions

Key Files to Monitor:

• /engines/patient_account_documents/app/services/patient_account_documents/v2/file_manager.rb - Line 107: download call
• /engines/patient_account_documents/app/services/patient_account_documents/uploader.rb - Lines 34-50: upload sequence
• /engines/patient_account_documents/app/models/patient_account_documents/document.rb - Line 240: server_encrypted_content_available? check
• /packages/@doctolib/unified_document_viewer/src/unified/viewer-modal/UnifiedDocumentsViewerModal.tsx - Line 157-180: Error tracking

The timing issue occurs because the document metadata is persisted to the database before the file is guaranteed to be fully available in S3, creating a window where the document appears available but the file cannot be retrieved.