Ideas for Parashift as submitted to our Feature Upvote board. Ideas are ordered by 'new' and the 50 top matches are included. pr_oh2s0pbfmgiimvw 2026-04-22T06:07:57Z sug_gxgt85ck9d1j6ax 2026-02-19T13:11:41Z 2026-04-22T06:06:18Z A new API Key is valid for 6 month. I would like to have a notification, before an API Key is expiring, so there will be no interuption for the customer just because we did't remarked it. sug_fkgtai9zj5mfw7p 2026-03-24T14:43:06Z 2026-04-22T06:06:16Z Today we need to edit each users permissions individually. It would make editing permissions a lot easier if you could use groups to edit many users at once. Even better if you could push these permissions between tenants. sug_wn2untgymnhgojs 2026-03-31T15:48:29Z 2026-04-22T06:06:13Z Sometimes I need to create a new doctype that is variation of a similar existing document type. If I copy the existing one, both are using the same fields. Although I don't want any shared element between those 2 doctypes. I'd love to have a "Copy with new fields" option sug_kwdqnqr971gsluh 2026-02-09T08:30:23Z 2026-02-10T05:31:16Z Es wäre hilfreich, wenn man für das Training die einzelnen Werte je Feld auf Zeichenebene, z.B. mithilfe der Pfeiltasten markieren könnte. Ist so etwas vorgesehen? sug_plkre3ueizzvhlg 2026-02-03T12:41:40Z 2026-02-10T05:31:02Z For admins who are active in many tenants, the list is long and not sorted. Having it sort in alphabetical order would make the list more clear and easier to find a certain tenant. This would be the easiest improvement to implement, I think. Another idea is to be able to make certain tenants "invisible" in the "Personal settings" page, so it would not show in the quick list. Ofcourse users should be able to undo this too and make tenants visible again. sug_rdnuv1psmioaybj 2026-01-20T07:20:38Z 2026-02-10T05:30:34Z Hey there The resolution of the outbound PDF are differ from the resolutuion of the inbound. The problem is that parashift turns the pdf to an image and it will converted in an image with 2500 pixel longest side. If you have set the checkox "Generate Output PDF" in the upoad-config it must be so, that the inbound resolution is the same as the outbound, for example 300dpi. Actually this differ and when the result of some docments before and some documents from parashift merge you got an PDF with different page sizes. Thanks & Best regards sug_u7btds7jfkfhbh9 2025-11-26T08:55:31Z 2026-02-10T05:30:15Z The tenant search in the menu over "Powered by Parashift" is only case-sensitve. It would be nice to search non-case-sensitive. sug_bhtnlwr9eer8xn2 2025-10-28T12:53:33Z 2025-11-12T05:49:03Z To keep the EV view clean, for example, if not the training option or not for validation is active, the "hide" field should be available. sug_vxi7vciuslcybgd 2025-10-22T11:21:19Z 2025-11-12T05:55:09Z - separation-validation via API - remove/add batches to training sug_kf7k4jtkmssevpi 2025-08-04T13:01:24Z 2025-11-12T05:55:17Z It would be a useful feature if there were an option to pass data up or down from the tenant's perspective. For example, the tenant could only learn from its parent, but the learning data of this tenant would only be available there and would not be passed up to the parent and distributed to other children. sug_ozggbzyixsldxxs 2025-08-04T12:02:23Z 2025-11-12T05:49:30Z Is there a way to ensure that only documents that have been poorly recognized end up in QC, while the documents still go through? The background to this is that some documents were not recognized properly in one of our child tenant and now need to be manually trained. We would therefore like to have a filter that allows us to identify which documents were recognized below a certain value. Despite this, the documents should still be processed and the JobRouter processes should not prevent them from being processed. Filtering allows the customer to dive right into Parashift and train these documents. sug_5cioicfpmnekosq 2025-07-29T10:03:13Z 2025-11-12T05:55:28Z sug_xypul9ugprso9jn 2025-07-02T10:18:47Z 2025-11-12T05:55:34Z sug_25q2gutyep8svrr 2025-07-01T12:01:01Z 2025-11-12T05:56:17Z potential value for: Healthcare Monitoring feature integration sug_h1soje37vzbpfjv 2025-06-24T14:22:05Z 2025-06-24T14:22:05Z Users expressed a need for more screen space when working with both the document viewer and the validation form. sug_g8avaudzz5txss3 2025-06-18T14:51:39Z 2025-06-19T06:56:39Z Capture tokens from two different pages is currently impossible. sug_yj9zzwj84ug9r3u 2025-05-14T13:59:36Z 2025-11-12T05:57:44Z "We understand that currently a single model version is used across all tenants for a given document type. To enhance flexibility and control, we propose implementing model versioning with the ability to manage deployment and usage of specific model versions per tenant. This feature would enable: - Targeted training and controlled rollouts - The ability to rollback to previous model versions if needed - Safe testing of new model versions in isolated environments - Comparative analysis of model performance across versions to track improvements or regressions This would provide significantly more control and transparency in the model lifecycle and performance monitoring." sug_mukkeyzxswfi24s 2025-05-14T13:58:07Z 2025-05-14T13:58:06Z We aim to use shared tenants (e.g., dev, test, prod) that include all relevant document types. To maintain strict control over the training process, we would like these tenants to operate in read-only mode. Our existing training procedures are tightly managed and should remain the sole channel for model updates and modifications. Therefore, we request a feature that allows setting an entire tenant to read-only, ensuring that no new documents can be uploaded or used for training within that environment. This would help avoid confusion and prevent any unintended training activities. sug_v769fw5txqkctvp 2025-05-12T12:08:38Z 2025-05-12T12:11:57Z According to assurance requirements, Edge is sometimes the preferred choice and Chrome is only made available by corporate IT departments in exceptional cases. In this respect, it would be helpful if the Parashift Platform was also offered optimized for the Edge browser in addition to Chrome. Overall, everything seems to work, but Chrome works a little better with larger workloads. sug_a0ktop5shnk2qn6 2025-04-15T06:50:19Z 2025-05-12T12:11:58Z Filter function to find documents that were manually validated during any workflow step. sug_d5vpfjyxludbfyk 2025-03-13T08:58:02Z 2025-05-12T12:11:59Z add a fraud tag to id documents if machine readable data and human readable data is not consistent. sug_oke8qsmu3i6c2qv 2025-02-11T18:19:30Z 2025-05-12T12:12:00Z Given that the page count is readily available, we would appreciate if it could also be included in the JSON output sug_jmdhockfdmbqmet 2025-02-03T09:16:30Z 2025-02-03T10:26:52Z It would be great if the filter options in document overview could be customised and you could filter by vendor or invoice number, for example. sug_nctwcholvoeoboe 2025-01-31T09:36:01Z 2025-11-12T05:57:33Z In the case of a multi-page document, it occasionally happens that previously correct pages are rotated incorrectly when pages are rotated. This must be switched off or be customisable. sug_vooizn1arzrmnvh 2025-01-29T10:52:25Z 2025-02-03T10:27:09Z Our engine with rotation document need to be updated... sug_ttvamnmh9aclxew 2024-11-18T14:54:16Z 2025-02-03T10:27:12Z It is about standardizing the interpretation of data from right to left. - If you move the Hebrew words in the editor, the value is automatically rotated correctly. - However, if the value is supplied via detection, the rotation does not take place.  We now have the option of defining a transformer for each field 3 2 1 becomes 1 2 3  However, this is not possible with line breaks, as can be seen in the example sug_haub5yppcvv7z0e 2024-11-08T13:19:05Z 2025-11-12T05:56:34Z I have two customers who want to use a RAG system to use LLMs on a larger context. In both cases it is necessary to create multiple vector DBs that can be used dynamically in POST calls. One would like to be able to build these based on document types and tags, the other would like to create this either via a batch upload (at least 5-10 GB of files) or via document IDs that should be embedded. sug_adqavtoitsllkhi 2024-11-08T08:26:33Z 2025-11-12T05:57:07Z Other services make it easier to understand what webhooks to activate when, for example, a user/system needs to follow-up with an intervention. E.g., "Activate all webhooks where intervention is required" and it automatically checks all relevant ones. sug_sm27qevwnjou7hc 2024-11-01T07:56:05Z 2025-11-12T05:50:08Z I want to be able to upload multiple files as context to an LLM (not as a batch) and receive info on where the LLM found information (document, page or section when, for example, the LLM provided a summary on specific parts of a document). sug_hv9zg7zlj1sog7m 2024-11-01T07:52:25Z 2025-11-12T05:50:26Z In the context of a virtual data room use case, we would like to be able to send not only a document or a set of documents to the platform, but also an LLM prompt that is then applied. Use case: a prospective buyer looking at company data in the data room wants to get specific answers on a document or set of documents, or extract data such as contract clauses that they specify on the fly, and then compare those clauses across multiple contracts. sug_hagpdrs7milekdm 2024-10-31T12:04:14Z 2024-11-01T13:16:30Z Please allow to select and copy OCR text from the document view. In the document view it's clearly visible, that the content has been recognized by OCR. When hovering the mouse over any content it's highlighted. However, the current settings do not allow selecting and copying it into the clipboard. Currently, if you need a useful information from the document, you must read and write it manually, which is laborious and prone to error. This restriction is unfounded. sug_pxkpemb1pszixdl 2024-10-31T09:17:58Z 2025-11-12T05:50:54Z It should be possible to have checks over the extracted Data. For example if there are 4 line items with amounts, the sum of these items should match to the extracted total amount. sug_nnwb8oqnrbbvvmq 2024-10-03T13:24:39Z 2025-11-12T05:51:04Z As mentioned in the ticket with the reference number 3851103598, the processing of large documents in Parashift fails. As a result, the documents cannot be further processed in a downstream process. I therefore request an increase in the maximum document size from 20 MB to 200 MB. sug_1nnx71utfrp3shj 2024-09-23T06:24:02Z 2024-11-01T13:16:30Z This applies in particular to uploads via e-mail. Currently, all attached documents of a mail are merged and treated as one document in the classification and extraction steps (unless a separation has been configured). However, the output files provided are then only available individually. It would be good to be able to configure this so that all documents are merged into one PDF, even for downloading. Use case: Invoices and work reports are submitted separately by the supplier, but should be available as one document in the subsequent approval workflow and for archiving. sug_0rwqxtk22gle2k4 2024-09-23T06:15:00Z 2024-11-01T13:17:02Z Uploaded documents that cannot be processed due to their mime type should be automatically converted to PDF. This is of course pointless for audio or video files, but there are various quasi PDF variants that cannot be processed so far. sug_9yxskspsorxf1e4 2024-09-13T07:03:51Z 2025-11-12T05:51:21Z This information is required to detect inactive users in order to have up-to-date access management, which is important for security reasons. Users with the “user admins” right should have access to this information. sug_vttmea8onwxn8hx 2024-09-02T15:23:52Z 2024-11-01T13:17:02Z Add the ability to assign users to teams. Add the option to assign document-types to teams. Members of a specific team shall only be able to see documents assigned to their team. Usage of validation levels shall be possible. sug_8wawrmukhl909rx 2024-08-13T15:48:34Z 2024-11-01T13:17:02Z Add logs of changes to fields, fieldsets and document types to the UI. sug_rjucjtqzixdayzg 2024-07-10T12:04:38Z 2024-11-01T13:17:02Z For some clients, it would help to restrict the classification on the first page of a document because the following pages are attachments. sug_nd7tkoxbslvzddl 2024-06-14T16:28:28Z 2025-11-12T05:59:46Z One of the customers would like to use the chat function outside the Parashift system boundary. sug_urnwjleh6rw0pal 2024-05-27T12:07:40Z 2024-11-01T13:17:19Z We'd like to use the extraction screen for the end-users auf our application but we are not allowed due to data access issues. Use case: A user sends a document from our application to Parashift via API. If recognition is not good enough this user should be able to do the validation in Parashift. The problem is, that the user would also see all other documents from the other users. But this user should only be allowed to see the documents that he sent via API. His "own" documents. Therefore it must be possible to set the access-user(s) via API on a document. A new role like "Validation of own documents" would be needed in Parashift. sug_maltzh7khhidfkd 2024-05-02T10:10:07Z 2025-11-12T05:56:47Z Enable chat with document outside of Parashift's GUI. sug_rfmoghnlfrzweql 2024-04-24T08:20:09Z 2025-11-12T05:51:54Z SSO Login Ability to log in via Entra ID user account. Automatic Detection Post-login automatic identification and data retrieval via GraphAPI from Entra ID/OAuth endpoint. Role-based Access Control Apply user roles assigned in Entra ID; assign different permissions in the - application based on roles. Role Management Central role assignment through Entra ID; changes to reflect after re-login. Security and Compliance Adherence to current security standards for secure authentication. Technology Use of OpenID Connect via OAuth 2.0 Flow. Validation Ability to validate Identity Tokens (JWT). Token Claims Extraction of user roles from token claims. sug_6hodwxkds9nzozg 2024-04-22T08:00:52Z 2024-11-01T13:17:54Z If you manually capture a value for a field with transformers, contrary to the automatic extraction, you have to manually add the transformation. We need automatic transformation on manual validation, similar to the currently existing data output type: date transformation. sug_o7e8cslrzrqtse7 2024-04-17T09:23:36Z 2025-11-12T06:01:03Z A large insurance customer wants Parashift to automatically identify and redact sensitive data such as direct or indirect identifying documents. sug_dzeshexqcmwkdza 2024-04-09T06:29:38Z 2024-11-01T13:33:50Z sug_xnh9y8d2h82ig2u 2024-04-04T09:05:49Z 2025-02-10T12:57:30Z Certain customers have documents that are classified according to different confidentiality levels. Some of them would like to be able to make this classification in Parashift. sug_ylethxowyubd0d8 2024-03-26T09:40:02Z 2024-11-01T13:17:54Z API call that returns a list of all active (production) document types with fields for the tenant sug_shjgqn2syyn87mr 2024-03-13T14:38:10Z 2024-11-01T13:18:15Z 1. Since the zoom level and resolution are not sufficent, in order to see details on large documents (e.g. large plans), it should also be possible to download PDF documents. sug_xy3ec9gbxbuf4ud 2024-03-13T14:22:19Z 2024-11-01T13:19:55Z Classification Validation: In first level processing, for selection of a class a click on the class in the dropdown is needed. It is not possible to select it with the keyboard by e.g. pressing enter. Could be more efficient, if using only keyboard and if there would be configurable Shortcuts.