v0.4.9

New Features

• Allow users to configure LLM inference parameters per bot from the UI. #303 by @jessieweiyi

Bug Fixes

Enhancements

• When enableMistral is true, the app name in the frontend UI is updated to "Bedrock Chat". #285 by @hustshawn
• Explicitly specify the AWS region when accessing DynamoDB resources to prevent AccessDeniedException. #304 by @GitSumito
• Implemented parallel processing and detailed structural analysis (unstructured.partition.pdf) for embedding PDFs to improve performance. #301 by @fsatsuki
• Added Italian translation. #282 by @angelo-malatacca

v0.4.8

New Features

• Add thumbs up/down buttons to each message to collect user feedback. #256 by @statefb, @wadabee

Bug Fixes

• Fixed a deploy error regarding IdP. #254 by @k70suK3-k06a7ash1

Enhancements

• Skip embedding process if there are no updates to the knowledge, chunk_size, or chunk_overlap in the bot console. #273 by @fsatsuki
• Add support for Mistral models. #253 by @hustshawn
• Add Spanish language support. #258 by @DTheunis
• Update backend README with all the required environment variables to run it locally. #263 by @clementvp
• Add math equation rendering support. #262 by @Taikono-Himazin

v0.4.7

New Features

• Support for the latest Claude 3 Opus model. #250 by @gabrielkoo

Bug Fixes

• Fixed an issue where the chat history title was displayed in the wrong language. #247 by @kylelee24
• Fixed an issue causing the embedding job to fail by adding the poppler dependency. #244 by @skeppeler
• Fixed an issue where the IDP configuration was not reflected correctly on the login screen. #242 by @statefb

Enhancements

• Improved the embedding sync process by passing only the DynamoDB record keys through the EventBridge pipe instead of the entire record, avoiding the 8KB character limit. #233 by @JoshuaToth
• Added support for Progressive Web App (PWA) and optimized icons. #245 by @wadabee

v0.4.6

New Features

• Google Auth as an optional feature for user sign-in. #128 by @kyosuke-kobayashi-lvgs
• Added the ability to customize LLM parameters such as chunk size and overlap from UI. #188 by @2814109
• A shutdown feature to stop Aurora Serverless at scheduled times. #228 by @2814109
• Support for custom OIDC providers for user authentication. #213
• Ability to save the web app to the iOS home screen as an icon. #176 by @amaynez
• Added Chinese (simplified & traditional) language support. #226 by @dennischang
• Ability to restrict sign-up to specific email domains. #234 by @fsatsuki
• A feature to show related documents as references in bot responses. #191

Bug Fixes

• Fixed an issue that prevented starring a shared bot without first having a conversation with it. #203
• Fixed an error that occurred when sending large images. #201

Enhancements

• Improved UX on mobile devices by fixing the page height to extend below the address bar. #181 by @alandao
• Added a preview image dialog when clicking on images in chat messages. #199
• Reduced the task startup time by 80% for the embedding job by enabling the SOCI index. #185 by @tmokmss

v0.4.5

New Features

• Added API publication and Admin dashboard features. Allows users to publish their customized bots as stand-alone APIs and analyze usage for each user / bot on the administrator dashboard.
• Added French and German language support to the i18n framework. #160 by @jeremylatorre and @dmetzler, #168 by @kennell

Bug Fixes

• Fixed an issue where markdown could not be used for external knowledge. #175
• Fixed the '+' new chat button on the top-right corner to correctly navigate to the new chat page. #180 by @alandao

Enhancements

• Updated Llama Index to the latest version. #163

v0.4.4

New Features

• Added support for Claude 3 Haiku.
• Reduced the minimum Aurora Capacity Units (ACUs) from 2 to 0.5 for Aurora Serverless, allowing for more cost-effective database scaling. by @kyosuke-kobayashi-lvgs
• Enabled querying Aurora directly from the Management Console by activating the Data API, providing a convenient way to interact with the database.

Bug Fixes

• Fixed an issue where file uploads were not properly reflected when updating custom bots, ensuring that all changes are now accurately applied. #113 by @Taikono-Himazin
• Resolved a problem where uploads sometimes failed due to DNS resolution issues.

Enhancements

• Added a link to the Anthropic Prompt Library within the prompt examples, providing users with access to a wide range of helpful prompts and templates.

v0.4.3

Fix bug: RAG functionality not working

v0.4.2

Multi-modal chat by Claude 3 (Sonnet).

• More powerful performance than Claude 2
• Accept both text and images on chat

v0.4.1

Fix #102

Calculate number of token based on text length when split the text to invoke cohere-multilingual

v0.4.0

External Knowledge Feature

Utilizing RAG (Retrieval Augmented Generation): This update enables the use of the following information for customized bot:

• Web (html)
• Text Data (txt, csv, markdown, etc.)
• PDF
• Microsoft Office (pptx, docx, xlsx)
• YouTube Subtitles

NOTE: Knowledge can be provided through URLs or file uploads.

A simple logic has been implemented. For more details, visit here.