Privacy Policy
Konvra develops and operates this public document service. Processing depends on the tool and the options you confirm. This policy describes the current application behavior and identifies where external provider practices also apply.
Scope and information handled
This policy covers Konvra’s public website and tools. No account or permanent online document library is currently provided. Depending on the feature, information handled includes documents or text you choose to process, cloud-request filenames and settings, browser preferences, operational events, and technical request information. A document may contain personal information about you or others; include it only if you have authority to process it.
Local document processing
Tools marked on-device process document content in your browser. Local annotation, visual signatures, editable forms, redaction, image-only sanitization, scan enhancement, searchable-PDF OCR, inspection, organization, and local exports do not upload document content for their operation. Local OCR loads bundled worker software, language data, and fonts from Konvra; those asset requests do not send your document. Files, previews, and prepared results use browser memory and temporary browser objects. Limited operational metadata may still reach Konvra. Website and asset requests are ordinary network activity; local processing does not mean zero network connections. Temporary result handoff uses application memory and is cleared on reload.
Confirmed server processing and external AI
Some tools require temporary processing on Konvra’s server: Office or HTML to PDF, editable PDF to DOCX, password protection or authorized unlocking, repair, strong compression, and available signature or PDF/A checks. After the server-processing confirmation, the selected file, its filename, operation settings, and any password you supply are transmitted to Konvra. The native document worker is configured to block Internet access, including external images and fonts; its availability depends on the deployment. This server path does not send the document to an AI model. A supplied password is used for the requested operation, not added to saved browser preferences or routine tool-event payloads.
Hybrid conversion first reads text locally. After cloud OCR confirmation, pages needing vision can be packaged into PDF batches and sent with filenames and settings; configured extraction or vision services may receive those batches or page images. After AI confirmation, summaries, document questions, translation, and structured extraction send recovered or pasted text, questions, relevant conversation history, requested fields or schemas, and output settings to Konvra and the configured model services. The provider may receive retrieved page passages rather than every source page. OpenRouter is the default gateway. Upstream provider retention, logging, and training practices depend on routing and account settings; Konvra does not promise universal zero retention or no training across providers. Review the linked policies before using external AI.
Temporary files and retention
Konvra does not intentionally maintain a permanent repository of uploaded documents or generated results. Native server processing writes input and output to a per-job temporary directory and attempts cleanup after success, errors, or cancellation. Server jobs are bounded to two minutes; interruption or cleanup failure can leave artifacts, so guaranteed immediate erasure is not promised. Browser downloads remain on your device until you remove them. Local and AI workspaces are temporary; there is no account-backed document library, team archive, or durable conversation history.
Configured persistent aggregate counters expire 120 days after their latest update; activity renews that expiration. Rate-limit identifiers use the configured window and cleanup rules. Native-worker limits use hashed network identifiers in process memory. In-memory counters can remain for the server process lifetime. Hosting logs, backups, and external-provider records have separate retention practices that application cleanup does not control.
Operational events and network information
Konvra records tool views, input-added and job-started events, success or failure, language, timestamps, and progress information such as job identifier, stage, page count, file byte count, duration, outcome, and error category. Server records also include guardrail events, request/error counts, model identifiers, and AI cost. These events support reliability, troubleshooting, abuse prevention, and service budgeting. Routine tool-event payloads are designed to exclude document content and filenames.
Requests necessarily reach hosting infrastructure with technical information such as IP address and request headers; host logs may retain it. AI and extraction rate limits use a truncated hash of a network identifier, or a shared fallback when one is unavailable. A hashed identifier is not a guarantee of anonymity. The current application does not sell your documents or use them to train its own models.
Browser storage and preferences
Browser storage remembers AI and cloud-OCR confirmations, output language, summary length, interface preferences, and workflow recipe settings. Recipes store settings, not document bytes, filenames, or content. The native server-processing checkbox is separate from AI confirmation; review the tool disclosure before each use. Password-entry values are temporary interface state and are not saved as preferences. Locale routing may use a preference cookie. Uncheck remembered AI or OCR confirmation to disable later requests that require it, or clear site data to reset preferences. This does not undo completed transmissions or delete downloads, hosting logs, or provider records.
Service providers and processing locations
Konvra’s server and hosting infrastructure receive files and settings when you choose native server processing, and technical request information when you use the website. Configured telemetry storage receives aggregate counters and rate-limit keys. External OCR, vision, and AI services receive only the batches, images, text, history, fields, or settings used by the confirmed external operation. OpenRouter and its upstream providers are recipients when that gateway is used. Native-worker Internet blocking does not prevent the browser-to-server upload or hosting infrastructure from handling the request. Providers may operate in other countries; no deployment region, residency guarantee, certification, or transfer safeguard is claimed without verification.
Choose how your documents are processed
Choose on-device tools when document content must stay in your browser. Local searchable-PDF OCR is available separately from cloud OCR. You may decline native-server or external-AI confirmation; the corresponding operation will remain unavailable. Read the actual tool label rather than assuming every conversion or OCR mode is local. Cancelling attempts to stop work but cannot recall data already received by a server or provider. Remove unnecessary sensitive content before authorized uploads. Routine operational events are sent automatically; the current interface does not provide an analytics preference control.
Security and sensitive documents
Konvra separates browser-local tools, confirmed native server uploads, and external AI paths; keeps provider keys server-side; validates requests; and applies limits. The native worker runs only when its Internet-blocking sandbox and required engines are available. These controls do not establish absolute security or certification. Password protection encrypts the PDF when supported; it does not secure copies already disclosed, prove identity, or guarantee that every reader enforces document permissions. Do not upload content that law, contracts, or professional duties prohibit from server or external processing.
Privacy rights and available controls
Depending on applicable law, you may have rights to access, correct, delete, restrict, or object to processing, withdraw consent, receive portable data, or complain to a competent regulator. These terms do not waive those rights. Local files and downloads can be managed on your device, and remembered choices can be reset in browser settings. These controls do not delete provider records or hosting logs. Konvra does not currently publish a dedicated privacy-request channel; the Help Center provides self-service processing guidance.
Age and automated output
External AI and cloud-OCR features are intended for adults aged 18 or older. Local OCR is a separate browser feature. Do not process children’s personal information without the authority required by applicable law. Answers, translation, extracted fields, and summaries assist document work; Konvra does not make legally significant eligibility, employment, credit, or similar decisions about people.
Policy updates
Changes to application behavior or disclosures will be reflected in this policy and its effective date. Review the current policy before choosing a new cloud feature. External provider policies can change independently; their published policies govern their own processing. The Help Center explains supported files, browser requirements, and recovery steps.