Overview
The AI Knowledge Assistant is an innovative UI component tailored to revolutionize user interactions within the Appian platform. By harnessing the power of large language models, this tool offers an immersive chatbot experience, allowing users not just to converse with state-of-the-art Generative AI models, but also source answers from their Appian Knowledge Center documents. This means, beyond traditional data retrieval, users can now semantically search and engage with their private data. Designed with adaptability in mind, the AI Knowledge Assistant seamlessly integrates with various organizational themes, ensuring a cohesive brand experience.
Key Features & Functionality
Notes:
The latest update includes usage alongside response metadata within the "assistant" messages. Please ensure you've updated to the most recent version of the plug-in.
Regarding "Included token usage info for OpenAI & Azure OpenAI request/response in the conversation saveInto"
When viewing the value of conversationSaveInto: local!conversation, It does showing any token usage information.
Please update to the latest Vector Database v3.2.3. This should resolve this issue as well as your previous "Processing Documents" bug. Hope this helps!
Thank you for updating.
However, after installing the update, we cannot get rid of the following message that blocks usage.
There seems be to no change possible/required to the function name or anything else in order to resolve this error.
Please review
Hi,
We have implemented this feature successfully in production. We have increased the Heap Max for app server by 1GB on beforehand.
The feature works, however, it continues to show 'Processing Documents' as a user-textballoon together with a loading circle. Even while that remains, the feature is usable.
The AI Knowledge Assistant does not work with the "Appian" mobile app on an iPhone. It just spins and does not allow you to type anything.
It does work from a browser, it just does not work from the mobile app.
Hi andersonc6744,
It looks like GPT is responding with ```markdown``` instead of specific markdown tags. We'll look into modifying the prompt we send behind the scenes, but in the meantime you can add "Make sure to never use the ```markdown``` tag. Always use specific markdown tags." to the system prompt which should curb this behavior.
I am using v3.1.3 with Vector DB v3.2.2 and Appian 24.2
The chat response from AI does not display properly.
When the problem occurs, you can see the background color of the response does not match the chat background. The text display in solid white background like a white box without borders. In addition, the text does not properly wrap. Instead, a horizontal slider bar is displayed under the white box.
Hi hansvt,
Happy to help. Feel free to reach out to techpartners@appian.com with the the component SAIL setup and the vector database configuration. I can attempt to debug from there.