Release Notes

- Added the Contract Builder, which asks you simple questions about a deal and builds a ready-to-send statement of work from pre-approved wording

- Security updates applied

- Raw scanned-text tables are now hidden, so documents only show the tables you need

- Simplified search results

- Search now runs through the main application instead of a separate service

- Retired the AI chat feature

- Removed the leftover AI chat page from the app

- Removed the app's remaining connections to the old chat service

- Removed unused integration

- Removed the document notes feature

- Removed unused background services to keep document processing simple and reliable

- Fixed an issue where a document tab crashed under certain conditions

- Deleting a tag table no longer erases the original extracted data for that document

- Uploading a file over the size limit now tells you why it was not added

- Fixed an issue where the calendar widget was slightly off-screen

- Fixed the home page so it keeps loading correctly after the old chat service was retired

- Product and tag reporting is now available through the Kiwi MCP server

- New reporting APIs for product data and tag data

- Added a simplified product view to make product reporting faster and easier to read

- Added a new Ramp Duration tag for extracted contract data

- Security fixes and dependency patches applied

- Product start and end dates in reporting are now stored as proper dates, so they sort and filter correctly

- Moved the remaining admin tools into the main application for a single, consistent experience

- Improved our internal build and automated testing process to help us ship changes more safely

- Fixed an error that stopped a document from opening when it was linked to another document

- Fixed an issue where document content was missing from search results

- Fixed search results that linked back to a page which would not open

- Uploading a file that is too large now shows a clear message instead of an error page

- Fixed an issue where the ServiceNow template export did not download

- Added a tenant picker, so you can see and choose which workspace you are working in

- You can now switch between tenants without the page reloading

- Added a location code column to extracted tables

- Security fixes and dependency patches applied

- Standardised application logging across services for more consistent monitoring and diagnostics

- Removed unused legacy search code

- Corrected a pricing error where some prices read from documents were stored far larger than the amount shown in the document

- Fixed the Reports page showing up blank instead of displaying reports

- Fixed a dashboard search that could hang without ever returning results when the answer included source references

- Fixed search only looking at the documents currently loaded on screen, so documents beyond the first 100 were never found

- Document diagnostics, a new tool that explains why a document failed to process, with a step-by-step guide for fixing it

- Accuracy Report, a new report showing how accurate extracted data is

- Total recurring equipment fees now appear in the product table

- Monthly internet pricing now includes a Package Product column, priced by package average

- Security updates applied

- Fixed sign-up on the sample analysis page, where the "Continue" button would fail

- Fixed issue where clicking a field to check it against the document changed the extracted value

- Extracted addresses now keep their line breaks instead of running together

- Fields now show the correct input control instead of a plain text box

- Uploaded documents are sorted by document type again, instead of all showing as "Unclassified"

- Additional security hardening applied

- Fixed issue where the document list's pagination showed an incorrect total count

- Fixed issue where the Assign Columns dialog did not populate table cells after applying

- Fixed issue where setting a table row as a grandchild would silently delete the row on save

- Locked and published documents now show a clear read-only indicator

- Fixed issue where the table row action menu could scroll out of view on long tables

- Fixed issue where newly added tables were not visible until the page was refreshed

- Fixed issue where deleting a tag table always failed

- Fixed issue where table cell focus no longer advanced automatically after selecting a document

- Fixed issue where a normalization error icon incorrectly appeared on empty table cells

- Removed unnecessary placeholder dashes shown in the UI

- Fixed issue where an internal-only field was visible in the document attributes view

- Fixed issue where an AI extraction error could silently skip tagging and block the document processing queue

- Fixed issue where documents could get stuck and block manual tagging due to a backend processing error

- Applied additional security patches to backend dependencies

- Improved HTTP security headers and cookie handling

- Proof-of-concept MCP server for programmatic access to Kiwi documents and data

- Experimental Kiwi Reports feature

- Bulk transform and bulk republish for documents

- New UI for managing a document's canonical extraction run

- Refactored table editing for improved reliability and performance

- Fixed issue where the term type dropdown value did not save

- Fixed issue where table cell values were lost after deleting and re-adding a column

- Fixed an authentication edge case affecting deactivated accounts

- Fixed issue where a table cell would clear its value when re-clicked after selecting a document

- Fixed issue where a field would reorder position each time a related option was selected

- Fixed issue where currency values with a trailing comma were saved incorrectly

- Fixed issue where certain boolean fields were not restored correctly after a page reload

- Fixed an issue with the password reset flow

- Strengthened access controls and session security

- Applied security patches across backend dependencies

- FE V2 API updates

- External file upload monitoring page (independent client upload) v2

- Fixed security vulnerabilities

- Patched issue where buttons where difficult to read

- Corrected file upload error handling

- Login issue for external login monitoring

- Client Specific Data Calculation Added

- Independent Potential Client Upload Page

- External File Upload Monitor (for independent client upload)

- External file upload document type and malicious content validation (for independent client upload)

- Security fix added

- Fixed bug where company names were not appearing in document table

- Enhanced document error handling and retry system

- File upload functionality to separate administrative portal added

- Human review document state added

- Added a complete currency extraction to be applied across the document

- Created company, address, and product data verification process

- Pipeline simplification to reduce technical debt

- Improved outputs by adding verification validation for several data points

- Table accuracy and word map improvements

- Users can choose which extracted data to export

- User expereice improved by disabling inputs in certain document status

- Fixed issue where new company names required the user to highlight the company name twice

- Fixed issue where users could not connect to Retool

- Fixed error using search functionality

- Fixed issue where all AI tagging highlights were missing

- Resolved parsing errors in the pipeline

- Fixed search returning a redirect error

- Fixed issue where users could not filter documents by status

- Fixed issue where user could not update specific inputs

- Removed extra data in output step

- Fixed document type misclassification issue

- Fixed issue where Retool data was not loading

- Fixed issue where long document titles were effecting document upload

- Security Patches Applied

- Added verification for company, address, and product data

- ExtID and WordMap verification testing added

- Enhanced document error handling and retry system

- File upload enhancements

- Pipeline human review state added

- Improved table accuracy and word maps

- Improved currency AI extraction

- Table accuracy improvements

- Transform pipeline now correctly ignores items where export is false

- Resolved parsing errors in the pipeline

- Fixed search returning a 301 redirect error

- Fixed issue where users could not filter documents by status

- Inputs are now correctly disabled when document is in a non-editable status

- Fixed issue where attributes could not be cleared

- Fixed Word Maps behavior in transform

- Fixed issue causing document type misclassification

- Fixed issue where Retool data was not loading

- Created new centralized error tracking dashboard

- Error codes added for improved debugging

- Documents now enter an error state in the event that AI accuracy needs to be reviewed

- Added database update to support better accuracy tracking for tables

- User can now delete cells and shift rows up in tables

- When document is in a non editable status, inputs are clearly disabled

- Backend performance improved

- Security patches applied

- Document lists created so users can arbitrarily organize documents

- Users can now reorder rows in tables

- Added back most recent updates to reporting

- Had issue where id's were being shown instead of the actual names of accounts

- Fixed issue where update broke search

- Highlighting was appearing on incorrect page

- Document processing has been sped up from 20 minutes to about 7

- Users can open the advanced options tab for documents

- Users could not correct documents if they were in the appropriate state

- Large update broke reporting. This has been resolved

- Issue where documents where not processing after major backend updates

- Users could not view individual document details

- User could not view previously uploaded documents

- Added Document grid refresh so the user does not have to refresh the page to check document status

- Added Error States for faster bug fixes

- Auto generated API documentation added

- Uploaded documents now flow through states and various checks unless they encounter an error

- Internal team debug interface added

- Internal team tag comparison from command line added

- Added document table extraction to our new accuracy checks

- Document grid filters are now reflected in the url so they persist on refresh

- Improved search functionality

- Accuracy check information added to database for tracking and visibility

- User can directly update tag states from API for testing purposes

- Users can now tag synonyms in document for values present in dropdowns. Simply, tagging is easier

- Added a variety of actions to cells to improve table tagging

- Users can now clear content from cells in tables

- Added captcha to external webpage

- Added products as a dropdown tag

- Added a keyboard shortcut to set hierarchy for table tags

- Users could not open documents after update

- Document detail view is now visible again

- Issue where tag would not completely delete has been addressed

- Resolved problem where users could not log into application

- Documents are now moving through states as expected

- Tags are no longer being overwritten when documents are reprocessed

- Documents are now reprocessing as expected

- Fixed issue where AI was not reformatting dates extracted from documents correctly

- New list of document states is now populating dropdown

- Tables can now add chunks of duplicate data if that is how the data is presented in the customer document

- Fixed issue where incorrect urls made it impossible for the user to log into the application

- Users can now easily use new Assign Data feature in table columns that have dropdown values

- Security Fixes applied

- Tags save on enter or when user clicks away

- User can now mass update a column in a table from data extracted by AI

- Users can now automatically flag errors with known issues

- Beginning of back end for flagging errors in documents for human review

- Ability for users to set heirarchy or delete rows in tables in bulk

- User can now set rows as parents in a table

- Back end support for setting parents and children in tables

- Updated backend AI pipeline to new version

- Updated table parser to new version

- Updated table view to change between horizontal and vertical viewing

- Backend update where hanging processes are automatically cleared more regularly

- User can now drag down a cell to automatically copy down its contents

- User can now insert missing rows into the middle of a table

- Updated table columns to delete data when column is removed from table

- Save table column header order

- Tag tables has been moved to veritcla view with resizable panel

- Add Document Page Rotation for Landscape Docs

- Add Caution modal when deleting multiple rows

- Table cell focus will automatically move to cell below upon save

- Table styling has been update to mirror common spreadsheet user interfaces

- User can now nativgate the table with arrow keys

- When a user uploads a document with a completely new customer, that customer will be automatically added to the database

- User can designate table column order in the column menu

- Deleting documents was not completely deleting documents from the database

- Duration tags were not correctly saving as blanks

- Documents were not processing, fixed quickly

- Tags like renewal interval was being extracted by AI but not saving

- New customer names were being incorrectly matched in existing dropdown

- Columns were moving after user added data to first cell

- Table cells could not accept negative values

- Documents were uploading, but tags were not being found

- User could not create a new tag table for a document

- User could not open table dropdowns

- When the total currency was 0, the currency was USD, even when the rest of the table was different currency

- User could not remove columns created by AI

- Page crashes when updating initial term, net payment term and auto renewal interval tags

- Cannot clear Recurring Frequency tag

- Deleting a row adds rows to table

- Removing values from 'Term' column in table results in 'undefined undefined'

- User Can't rotate last page more than 1/4 turn

- Table cells data was duplicating down all existing rows

- Security Fixes applied

- Columns were reordering back to alphabetical when the user added a column

- User was getting kicked out of the instance even with activity in the app

- Backend Patching

- Security Updates Applied

- Table dropdowns now populate correctly

- User interface now correctly displays information that is generated by the agent

- Tables now always have an empty row, ready to create an additional row of data

- Added cookie consent popup to the app

- User can reorder table tagging columns as needed

- More tag terms are highlighted in the document for easy user review

- Advanced tab now contains OCR tables from uploaded documents

- Updated how various tags are identified by the agent to improve overall accuracy

- Removed redundant search code

- Removed unnecessary monitoring software

- AI responses standardized when a value is not found

- Currency properly displaying in UI when found by agent

- Cursor now stays in search bar when performing a keyword search of a document

- Term is correctly displayed in agent generated table

- Super users can reset dev instances

- Automatic upload retry has been fixed. Document uploads are now very stable

- Patched security fix

- Large documents now upload with no errors

- Added 'Quote' as a new document type

- Added a Configurable Embedding to make longer documents even more accurate than before

- Added an automatic retry to increase document load success

- Added configurable pipeline so different clients can have different AI agents

- Added accuracy tracking to our table structure

- Added a title to each AI generated table

- Standardized that data should be saved as tables to the database

- Added tag type to all tags

- Current AI version is now viewable in the DB

- Added bespoke citations to Kiwi Search

- Table Appearance

- QA Export

- Long Document tagging was non-functional

- Database was not updating after

- Security Updates Applied

- Created automatic clean up for any long running processes

- Tables now have searchable summaries to improve accuracy

- Various strategies implemented to continue to improve tagging accuracy for longer than average documents

- Improved Tags: Account ID, Total, and Currency

- Added a title to each AI generated table

- Standardized that data should be saved as tables to the database

- Added tag type to all tags

- Current AI version is now viewable in the DB

- Added bespoke citations to Kiwi Search

- Long Document tagging was non-functional

- Database was not updating after

- Improved search accuracy

- Add an automatic retry for document upload in case of failure

- Added citations to our search

- Chat was not returning responses

- Fixed testing issue

- Kiwi Chat responses have improved formatting

- Separated AI from application

- Dynamic example queries for search

- Created search with citations

- Added analytics to better track bugs

- Created a way to add document adjacent data to the app to inform reporting

- Security updates applied

- Created document upload permissions

- Users can now export configurations from existing instances

- Users can easily add configurations to new instances in bulk

- Added user monitoring for better bug tracking

- Added a simple user wiki to add non-document specific information that can add useful context

- Clear instance

- Document upload queue now will always complete jobs instead of timing out

- Security updates applied

- Users can update table tags

- User cursor stable when updating table name

- File attachment icon removed from chat (no longer needed)

- Documents appear once and only once in document grid upon upload

- Security updates applied

- AI Table processing error

- AI Table processing timing out

- Admin styles not applied

- Dynamic AI chat name

- Dynamic example queries for search

- Scrollbar updated for windows

- Security updates applied

- New AI pipeline implemented, reports are showing an over 30% increase in overall accuracy

- IBM table detect and fill added to on-premises solutions

- AI now auto detects the document type

- Resize Chat history

- Security updates applied

- Page redirect fixed

- Document loading error resolved

- Chat Search

- Chat Empty state

- Resize Chat history

- Clauses can now span multiple pages

- Chat Alpha

- Security Updates Implemented

- Created tool to extract clauses from documents and present to user

- Created feature to easily show or hide features in progress on production deployments

- Implemented new security fixes

- Improved failed document process display

- Created tool to extract clauses from documents and present to user

- Created feature to easily show or hide features in progress on production deployments

- Fixed login issue

- Reports page in view not refreshing when new status applied

- User groups and attributes were not being applied to ai search

- Page not refreshing when doc type is updated

- Cannot sort by status in doc grid

- Business partner drop down not populating in slide out filter

- Parsing error on table data

- Bug with AI Pipeline

- Bug when logging in

- Processing status and processing status message to uploaded documents

- Delete functions for testing

- RAG package updates, automated multi tenant schema creation, synchronized filter and delete, updated links on summary page

- Inputs were not updating to reviewed

- Currency dropdown was not populating after several clicks

- Could not select items directly within document for tagging

- Business partner dropdown was crashing page on open

- Add business partner was failing

- Enhanced debugging for AI training

- Enhanced table extraction and tagging

- Added AI pipeline export for documents

- Security fixes implemented