Unique AI Documentation

Release 2024.40

2 min read

Release Date:

Announcements

💥Breaking change in 2024.40

info

Only impacts Benchmarking service for External Modules

Benchmarking service for External Modules requires the last message to set a completedAt attribute.

Why?

To enable benchmarking of modules/assistants that take long to execute/complete and are not killed by a timeout.

Details

  • To determine when an assistant module has been finished generating it’s output, the benchmarking will check if the completedAt attribute was set to start the benchmarking evaluation. Earlier, a timeout of two min was set and so caused problems for assistants that could take longer.

  • This change requires all external assistant modules to set the completedAt attribute after the last assistant message was created to enable benchmarking.:

unique_sdk.Message.modify({
            "user_id": "user_id",
            "company_id": "company_id",
            "id": "message_id",
            "chatId": "chat_id",
            "text": "text", // text of the latest assistant message
            "originalText": "text", // text of the latest assistant message
            "references": "references",
            "debugInfo": "debug_info",
            "completedAt": datetime.now().strftime("%Y-%m-%d %H:%M:%S.%f"), // time of now
   })

Features

  • Implemented Deeplinking to the Chat App with Auto Space Selection and Prompt Execution. You can read more here

  • Implemented Telemetry Data Collection. You can read more here
    The following data are now being collected:

    • Number of users

    • Number of scopes

    • Number of groups

    • Number of chats

    • Number of spaces

    • Number of user messages

    • Number of assistant messages

    • DAU/WAU/MAU aggregation

    • Number of positive and negative feedback

Improvements

  • Optimized methods for performance and efficient handling of large data sets.

  • Improved speed of ingestion of pictures in the Chat

  • Added support for GPT-4o version 2024-08-04

  • Added User information to feedback export

  • Added support for changing the DB connection pool limit

  • Improved notification message when deleting a chat

  • The "Device Category" in the Detailed user interaction analytics now shows the total number of interactions by tracking mobile vs desktop interaction.

  • Improved chat page performance on large number of prompts thereby fixing the lag noticed on the page

  • Improved error message when attempting to upload files larger than the configured limit - currently set at 100MB - it’s the upload limit and therefore also the limit for the ingestion.

  • Cleaned up the Feedback report to include only relevant fields

Bug fixes

  • Fixed issue where analytics export fails on large dataset

  • When saving a prompt, warning text are no longer saved

  • Acronyms in the Translation Assistant are now preserved

  • Fix uploading of failed error when selecting markdown files

Changelog

Image Tags

 nextChat:            2024.40-a259a
 nextKnowledgeUpload: 2024.40-a259a
 nextTheme:           2024.40-a259a
 nextAdmin:           2024.40-63e9d
 nodeChat:            2024.40-b84b3
 nodeIngestion:       2024.40-d071a
 nodeIngestionWorker: 2024.40-d071a
 nodeTheme:           2024.40-d071a
 nodeScopeManagement: 2024.40-a259a
 assistantsCore:      2024.40-a03c4
 assistants-reranker  2024.40-a259a

Author

 

Last updated