December 11th, 2025

We spent the last month working with the Bellevue DS pilot team to annotate AI-generated answers using a new feedback form, then building improvements based on what we learned.
The team helped us immensely, running through testing sessions, throwing a range of questions at our Chat Assistant, and telling us how the responses held up. That feedback drove improvements to our AI systems, helping the Chat and Email Assistants better answer questions. We’re excited for you to see the results.
Check out the improvements at https://bellevuewa.govstream.ai/
Your assistants now:
Understand longer conversations better
Find more relevant documents
Cite sources more accurately
Can pull addresses from messages and use GIS lookups without manual entry in the project context sidebar

If you asked a tough question in November and it struggled, ask it again. The answer should be noticeably better. Check out some examples below:


We rebuilt how the system searches for information. It now uses multiple search methods at once and goes beyond exact matches to find other conceptually relevant sources. It keeps better track of context in back-and-forth conversations.
We removed the pending and resolved status tracking and kept what matters:
A simple table showing what you submitted
A dedicated page for full details when you click on an item
Form and dashboard now match perfectly


This is your feedback in action - going from annotating questions to improving the product in about a month.
Ask a question that you threw at it in November. Try Chat or Email Assistant again.
https://bellevuewa.govstream.ai/
Keep sending feedback and telling us what works and what we miss. That feedback drives what comes next.
Chat Assistants responses no longer return empty citations
Citations now better handle formatting and special characters in URLs
Timezones now reflect correctly in feedback views
Feedback ratings now record correctly
v1.5.0 released December 2025