# Changelog All notable changes to EveAI will be documented in this file. The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). ## 3.1.14-beta Release date: ### Added - Introduction of User Actions - TBC - Additional configuration options for Agents: temperature and llm_model can now be configured (if allowed in Agent configuration) ### Changed - Improvement of RAG Specialist, including proofreading on generated output. - Specialist Editor - separate modal editors for Agents, Tasks and Tools to allow for more complex configuration. ### Removed - PartnerRagRetriever model - not used. ### Fixed - Bug fix where client appears to return no result on an interaction, due to connections without correct search path (out of the connection pool) ## 3.1.13-beta Release date: 2025-10-17 ### Added - Introduce consent for DPA and T&C in the administrative client - Refuse activities in the administrative client if no consent is given or consent needs to be renewed ### Changed - Centralise content versioning and markdown rendering in the administrative client ### Fixed - Improvement of reset password and confirm email address functionality. ## 3.1.12-beta Release date: 2025-10-03 ### Added - TRA-90: Additional positive & negative KO Criteria Questions added to TRAICIE_KO_QUESTIONS asset and interaction and evaluation in TRAICIE_SELECTION_SPECIALIST ## 3.1.11-beta Release date: 2025-10-03 ### Changed - Improved marked integration in the Chat Client, allowing for more complex markdown rendering ### Fixed - TRA-86: Chat client sometimes 'hangs' waiting for a response from the server - TRA-92: Form rendering given some extra space in AI message bubble ## 3.1.7-beta Release date: 2025-09-30 ### Added - Pushgateway for logging business events to Prometheus ### Changed - Prometheus deployment improvements and alignment with Pushgateway - Metrics logging in Business events to support multiple pod and processes - Maximum height for AI message in chat input also available in desktop client - AI message rendering now allows markdown - markdown rendering defined in a centralized utility ### Fixed - Bug preventing correct loading of cache busted css en js in eveai_app solved - Fix on rare bug preventing marked component to be displayed in SideBarExplanation ### Security - DOM checks on markdown text to prevent XSS ## 3.1.3-beta Release date: 2025-09-25 ### Added - Cache busting for static files ### Changed - Build process optimised for cache busting ### Fixed - TRA-83 - numerous changes to the mobile version of the chat client ## 3.1.2-beta Release date: 2025-09-23 ### Changed - Several improvements to the mobile version of the chat client ## 3.1.1-alfa Release date: 2025-09-22 ### Fixed - TRA-76 - Send Button color changes implemented - TRA-72 - Translation of privacy statement and T&C - TRA-73 - Strange characters in Tenant Make Name - TRA-77 - Adapted Scroll behavior for Chat Client in Message History ### Security - In case of vulnerabilities. ## 3.1.0-alfa Release date: 2025-09-12 ### Added - Configuration of the full k8s staging environment - k8s installation manual (cluster-install.md) ### Changed - Minio Bucket based approach adapted to a Folder-based approach - Startup scripts of all eveai services adapted to a more generic and reusable approach - Base image creation for all eveai services (iso building from scratch) - Enabled secure Redis connection (TLS + un/pw) + connection pooling for Session, Caches and Pub/Sub - Enable secure PostgreSQL connection (TLS) - Integration of monitoring possibilities in cluster (Prometheus, Grafana, Alertmanager, PgAdmin4, Redis Insight, Flower) - Introduction of Bunny.net as CDN & WAF - Isolation of initialisation tasks (in eveai_app) into a separate 'Job' container eveai_ops - Static files served from static pull zone (storage) on bunny.net - Improving health and readyness checks - Migration of dev and test with changes required for k8s - More comprehensive error communication system ## 3.0.1-beta Release date: 2025-08-21 ### Changed - Podman now replaces Docker for building images - Local registry now replaces Docker Hub - Start for k8s integration - RAG possibilities throughout usage of the TRAICIE_SELECTION_SPECIALIST ### Fixed - TRA-67 - initial edit of Tenant Make --> 2-step proces - TRA-68 - Correction of javascript for json editor, resulting in Asset Changes not being saved - TRA-70 - Wrong File Size display for Assets - TRA-69 - Wrong number of questions in TRAICIE_KO_INTERVIEW_DEFINITION_SPECIALIST (required correction in TRACIE_ROLE_DEFINITION_SPECIALIST) ### Security - In case of vulnerabilities. ## 3.0.0-beta Release date: 2025-08-15 ### Added - Mobile Support for the chat client. - Additional visual clues for chatbot and human messages in the chat client ### Changed - Additional visual clues for chatbot and human messages in the chat client - Adaptation (new version) of TRAICIE_SELECTION_SPECIALIST to further humanise interactions with end users. (Introduction of an additional interview phase to allow divergence from the interview scenario for normal questions, and convergence to the interview scenario. - Humanisation of cached interaction messages (random choice) - Adding specialist configuration information to be added as arguments for retrievers ## 2.3.12-alfa ### Added - Modal display of privacy statement and terms & conditions documents in eveai_Chat_client - Consent-flag ==> Consent on privacy & terms... - Customisation option added to show or hide DynamicForm Title (and icon) - Session Header defaults clickable, opening selection views for Partner, Tenant and Catalog - ### Changed - Document Processing View - show 'Finished Processing' iso 'Processing' to have more logical visual indicators - TRAICIE_SELECTION_SPECIALIST now no longer shows question to start selection procedure at initialisation. ### Fixed - Error Messages for adding documents in 'alert' - Correction of error in Template variable replacement, resulting in missing template variable value ## 2.3.11-alfa ### Added - RQC (Recruitment Qualified Candidate) export to EveAIDataCapsule ### Changed - Adapt configuration possibilities for Chat Client - Progress Tracker (client) level of information configuration - Definition of an Active Region in the client to ensure proper understanding - Adapting TRAICIE_SELECTION_SPECIALIST to retrieve preferred contact times using a form iso free text - Improvement of DynamicForm en FormField to handle boolean values. ## 2.3.10-alfa ### Added - introduction of eveai-listview that is sortable and filterable (using tabulator), with client-side pagination - Introduction of PARTNER_RAG retriever, PARTNER_RAG_SPECIALIST and linked Agents and Tasks, to support for documentation RAG - Domain model diagrams added - Addition of LicensePeriod views and form ### Changed - npm build now includes building of css files - npm build takes information from sourcefiles, defined in the correct component locations - eveai.css is now split into more maintainable, separate css files - adaptation of all list views in the application - Chat-client converted to vue components and composables ## 2.3.9-alfa ### Added - Translation functionality for Front-End, configs (e.g. Forms) and free text - Introduction of TRACIE_KO_INTERVIEW_DEFINITION_SPECIALIST - Introduction of intelligent Q&A analysis - HumanAnswerServices - Full VA-versioin of TRAICIE_SELECTION_SPECIALIST - EveAICrewAI implementation guide ### Changed - Allowed Languages and default_language part of Tenant Make - Refinement of EveAI Assets to define Partner Assets and allow storage of json - Improvements of Base & EveAICrewAI Specialists - Catalogs & Retrievers now fully type-based, removing need for end-user definition of Tagging Fields - RAG_SPECIALIST to support new possibilities ## 2.3.8-alfa ### Added - Translation Service - Automagic HTML Processor - Allowed languages defined at level of Tenant Make ### Changed - For changes in existing functionality. - Allow to activate / de-activate Processors - Align all document views with session catalog - Allow different processor types to handle the same file types - Remove welcome message from tenant_make customisation, add to specialist configuration ### Fixed - Adapt TRAICIE_ROLE_DEFINITION_SPECIALIST to latest requirements - Allow for empty historical messages - Ensure client can cope with empty customisation options - Ensure only tenant-defined makes are selectable throughout the application - Refresh partner info when adding Partner Services ### Security - In case of vulnerabilities. ## 2.3.7-alfa ### Added - Basic Base Specialist additions for handling phases and transferring data between state and output - Introduction of URL and QR-code for MagicLink ### Changed - Logging improvement & simplification (remove Graylog) - Traicie Selection Specialist v1.3 - full roundtrip & full process ## 2.3.6-alfa ### Added - Full Chat Client functionality, including Forms, ESS, theming - First Demo version of Traicie Selection Specialist ## 2.3.5-alfa ### Added - Chat Client Initialisation (based on SpecialistMagicLink code) - Definition of framework for the chat_client (using vue.js) ### Changed - Remove AllowedLanguages from Tenant - Remove Tenant URL (now in Make) - Adapt chat client customisation options ### Fixed - Several Bugfixes to administrative app ## 2.3.4-alfa ### Added - Introduction of Tenant Make - Introduction of 'system' type for dynamic attributes - Introduce Tenant Make to Traicie Specialists ### Changed - Enable Specialist 'activation' / 'deactivation' - Unique constraints introduced for Catalog Name (tenant level) and make name (public level) ## 2.3.3-alfa ### Added - Add Tenant Make - Add Chat Client customisation options to Tenant Make ### Changed - Catalog name must be unique to avoid mistakes ### Fixed - Ensure document version is selected in UI before trying to view it. - Remove obsolete tab from tenant overview ## 2.3.2-alfa ### Added - Changelog display - Introduction of Specialist Magic Links ### Fixed - dynamic fields for adding documents / urls to dossier catalog - tabs in latest bootstrap version no longer functional - partner association of license tier not working when no partner selected - data-type dynamic field needs conversion to isoformat - Add public tables to env.py of tenant schema ## 2.3.1-alfa ### Added - Introduction of ordered_list dynamic field type (using tabulator) ### Changed - Bring configuration of PROCESSOR_TYPES & CATALOG_TYPES to new config standard - Specialist Editor: move general information in tab - Role Definition Specialist creates Selection Specialist from generated competencies - Improvements to Selection Specialist (Agent definition to be started) ## 2.3.0-alfa ### Added - Introduction of Push Gateway for Prometheus - Introduction of Partner Models - Introduction of Tenant and Partner codes for more security - Introduction of 'Management Partner' type and additional 'Partner Admin'-role - Introduction of a technical services layer - Introduction of partner-specific configurations - Introduction of additional test environment - Introduction of strict no-overage usage - Introduction of LicensePeriod, Payments & Invoices - Introduction of Processed File Viewer - Introduction of Traicie Role Definition Specialist - Allow invocation of non-interactive specialists in administrative interface (eveai_app) - Introduction of advanced JSON editor - Introduction of ChatSession (Specialist Execution) follow-up in administrative interface - Introduce npm for javascript libraries usage and optimisations - Introduction of new top bar in administrative interface to show session defaults (removing old navbar buttons) ### Changed - Add 'Register'-button to list views, replacing register menu-items - Add additional environment capabilities in docker - PDF Processor now uses Mistral OCR - Allow additional chunking mechanisms for very long chunks (in case of very large documents) - Allow for TrackedMistralAIEmbedding batching to allow for processing long documents - RAG & SPIN Specialist improvements - Move mail messaging from standard SMTP to Scaleway TEM mails - Improve mail layouts - Add functionality to add a default dictionary for dynamic forms - AI model choices defined by Ask Eve AI iso Tenant (replaces ModelVariables completely) - Improve HTML Processing - Pagination improvements - Update Material Kit Pro to latest version ### Removed - Repopack implementation ==> Using PyCharm's new AI capabilities instead ### Fixed - Synchronous vs Asynchronous behaviour in crewAI type specialists - Nasty dynamic boolean fields bug corrected - Several smaller bugfixes - Tasks & Tools editors finished ### Security - In case of vulnerabilities. ## 2.2.0-alfa ### Added - Mistral AI as main provider for embeddings, chains and specialists - Usage measuring for specialists - RAG from chain to specialist technology - Dossier catalog management possibilities added to eveai_app - Asset definition (Paused - other priorities) - Prometheus and Grafana - Add prometheus monitoring to business events - Asynchronous execution of specialists ### Changed - Moved choice for AI providers / models to specialists and prompts - Improve RAG to not repeat historic answers - Fixed embedding model, no more choices allowed - clean url (of tracking parameters) before adding it to a catalog ### Deprecated - For soon-to-be removed features. ### Removed - Add Multiple URLs removed from menu - Old Specialist items removed from interaction menu - ### Fixed - Set default language when registering Documents or URLs. ## 2.1.0-alfa ### Added - Zapier Refresh Document - SPIN Specialist definition - from start to finish - Introduction of startup scripts in eveai_app - Caching for all configurations added - Caching for processed specialist configurations - Caching for specialist history - Augmented Specialist Editor, including Specialist graphic presentation - Introduction of specialist_execution_api, introducting SSE - Introduction of crewai framework for specialist implementation - Test app for testing specialists - also serves as a sample client application for SSE - ### Changed - Improvement of startup of applications using gevent, and better handling and scaling of multiple connections - STANDARD_RAG Specialist improvement - ### Deprecated - eveai_chat - using sockets - will be replaced with new specialist_execution_api and SSE ## 2.0.1-alfa ### Added - Zapîer Integration (partial - only adding files). - Addition of general chunking parameters (chunking_heading_level and chunking_patterns) - Addition of DocX and markdown Processor Types ### Changed - For changes in existing functionality. ### Deprecated - For soon-to-be removed features. ### Removed - For now removed features. ### Fixed - Ensure the RAG Specialist is using the detailed_question - Wordpress Chat Plugin: languages dropdown filled again - OpenAI update - proxies no longer supported - Build & Release script for Wordpress Plugins (including end user download folder) ### Security - In case of vulnerabilities. ## 2.0.0-alfa ### Added - Introduction of dynamic Retrievers & Specialists - Introduction of dynamic Processors - Introduction of caching system - Introduction of a better template manager - Modernisation of external API/Socket authentication using projects - Creation of new eveai_chat WordPress plugin to support specialists ### Changed - Update of eveai_sync WordPress plugin ### Fixed - Set default language when registering Documents or URLs. ### Security - Security improvements to Docker images ## 1.0.14-alfa ### Added - New release script added to tag images with release number - Allow the addition of multiple types of Catalogs - Generic functionality to enable dynamic fields - Addition of Retrievers to allow for smart collection of information in Catalogs - Add dynamic fields to Catalog / Retriever / DocumentVersion ### Changed - Processing parameters defined at Catalog level iso Tenant level - Reroute 'blank' paths to 'admin' ### Deprecated - For soon-to-be removed features. ### Removed - For now removed features. ### Fixed - Set default language when registering Documents or URLs. ### Security - In case of vulnerabilities. ## 1.0.13-alfa ### Added - Finished Catalog introduction - Reinitialization of WordPress site for syncing ### Changed - Modification of WordPress Sync Component - Cleanup of attributes in Tenant ### Fixed - Overall bugfixes as result from the Catalog introduction ## 1.0.12-alfa ### Added - Added Catalog functionality ### Changed - For changes in existing functionality. ### Deprecated - For soon-to-be removed features. ### Removed - For now removed features. ### Fixed - Set default language when registering Documents or URLs. ### Security - In case of vulnerabilities. ## 1.0.11-alfa ### Added - License Usage Calculation realised - View License Usages - Celery Beat container added - First schedule in Celery Beat for calculating usage (hourly) ### Changed - repopack can now split for different components ### Fixed - Various fixes as consequence of changing file_location / file_name ==> bucket_name / object_name - Celery Routing / Queuing updated ## 1.0.10-alfa ### Added - BusinessEventLog monitoring using Langchain native code ### Changed - Allow longer audio files (or video) to be uploaded and processed - Storage and Embedding usage now expressed in MiB iso tokens (more logical) - Views for License / LicenseTier ### Removed - Portkey removed for monitoring usage ## 1.0.9-alfa - 2024/10/01 ### Added - Business Event tracing (eveai_workers & eveai_chat_workers) - Flower Container added for monitoring ### Changed - Healthcheck improvements - model_utils turned into a class with lazy loading ### Deprecated - For soon-to-be removed features. ### Removed - For now removed features. ### Fixed - Set default language when registering Documents or URLs. ## 1.0.8-alfa - 2024-09-12 ### Added - Tenant type defined to allow for active, inactive, demo ... tenants - Search and filtering functionality on Tenants - Implementation of health checks (1st version) - Provision for Prometheus monitoring (no implementation yet) - Refine audio_processor and srt_processor to reduce duplicate code and support larger files - Introduction of repopack to reason in LLMs about the code ### Fixed - Refine audio_processor and srt_processor to reduce duplicate code and support larger files ## 1.0.7-alfa - 2024-09-12 ### Added - Full Document API allowing for creation, updating and invalidation of documents. - Metadata fields (JSON) added to DocumentVersion, allowing end-users to add structured information - Wordpress plugin eveai_sync to synchronize Wordpress content with EveAI ### Fixed - Maximal deduplication of code between views and api in document_utils.py ## 1.0.6-alfa - 2024-09-03 ### Fixed - Problems with tenant scheme migrations - may have to be revisited - Correction of default language settings when uploading docs or URLs - Addition of a CHANGELOG.md file ## 1.0.5-alfa - 2024-09-02 ### Added - Allow chatwidget to connect to multiple servers (e.g. development and production) - Start implementation of API - Add API-key functionality to tenants - Deduplication of API and Document view code - Allow URL addition to accept all types of files, not just HTML - Allow new file types upload: srt, mp3, ogg, mp4 - Improve processing of different file types using Processor classes ### Removed - Removed direct upload of Youtube URLs, due to continuous changes in Youtube website ## 1.0.4-alfa - 2024-08-27 Skipped ## 1.0.3-alfa - 2024-08-27 ### Added - Refinement of HTML processing - allow for excluded classes and elements. - Allow for multiple instances of Evie on 1 website (pure + Wordpress plugin) ### Changed - PDF Processing extracted in new PDF Processor class. - Allow for longer and more complex PDFs to be uploaded. ## 1.0.2-alfa - 2024-08-22 ### Fixed - Bugfix for ResetPasswordForm in config.py ## 1.0.1-alfa - 2024-08-21 ### Added - Full Document Version Overview ### Changed - Improvements to user creation and registration, renewal of passwords, ... ## 1.0.0-alfa - 2024-08-16 ### Added - Initial release of the project. ### Changed - None ### Fixed - None [Unreleased]: https://github.com/username/repo/compare/v1.0.0...HEAD [1.0.0]: https://github.com/username/repo/releases/tag/v1.0.0