Improving Copilot Observability in SAS Viya: Introducing Viya GenAIMonitor (Community Project)

[vitorlima86]

As Copilot capabilities become more widely used in SAS Viya, understanding how these features are being adopted in real environments is increasingly important.

Questions such as who is using Copilot, how often, and in what context often arise—from customers, administrators, and support teams alike. While safety mechanisms and guardrails are fundamental, observability plays a key role in building trust, supporting troubleshooting, and enabling meaningful feedback.

This post introduces Viya GenAIMonitor, an unofficial, community-led tool created to explore these observability needs.

 

Because Observality Matters

The initial goal of this project was to enable a clearer way to retrieve and review prompts and responses generated by Copilot. Access to real conversations can help:

• Review how users interact with Copilot features
• Identify response quality issues, including potential hallucinations
• Support troubleshooting and customer-reported scenarios
• Inform future improvements through real usage examples

As development progressed, the scope naturally expanded to include broader usage observability, helping teams better understand how Copilot is being used across different environments.

What Is Viya4 GenAI Monitor?

Viya GenAIMonitor is a lightweight, browser-based dashboard that connects to a SAS Viya Platform through a local Python proxy.

It enables read-only access to SAS Viya Copilot conversations, making it possible to:

• View active and historical Copilot chat sessions
• Explore conversations by user, chat ID, or free-text search
• Understand usage patterns across environments
• Safely retrieve chats for troubleshooting or analysis

The design follows a few strict principles:

• Read-only access
• Local-first architecture
• No external services
• No stored credentials

Everything stays within the environment.

Intended Audience

This community tool may be useful for:

• Technical Support teams retrieving conversations to assist with analysis and troubleshooting
• Customer administrators seeking visibility into Copilot usage within their environments
• End users or power users revisiting previous chats or commonly used prompts
• Anyone who needs to retrieve or review SAS Viya Copilot conversations for learning or analysis purposes

Community and Contribution

Viya GenAIMonitor is a community-developed tool and is not officially supported by SAS Standard Support.

That said, feedback and discussion are encouraged. Customers and community members are welcome to:

• Ask questions
• Share observations or suggestions
• Provide feedback based on real-world usage

These conversations can take place through the project’s GitHub page or within the SAS Communities platform, depending on what is most convenient.

Community input plays an important role in shaping ideas, identifying opportunities for improvement, and exploring how Copilot observability can better support real customer needs.

Closing Thoughts

As Copilot features continue to evolve, visibility into real-world usage can help improve understanding, support, and trust. This project represents one community-driven exploration of that space.

If observability of Copilot usage in SAS Viya is relevant to your work, you are welcome to review the project, share feedback, or contribute.

[vitorlima86]

Sorry, I forgot to add the link to the page:

GitHub project: Viya GenAIMonitor

Questions or feedback can also be shared through this post.

[HelissonMota]

New Features available:

• 📌 Pinned chats for quick access
• 💬 Ability to add comments to each chat
• 🔁 Replay function for conversations
• 🆔 One-click auto-copy of Chat ID

Data & Export Enhancements:

• 📊 Download CSV data from All and History (with period-based sorting)
• 📈 Export graph views for all users or individual users (PNG or PDF formats)

New Visualizations:

• 🔥 Heatmap chart
• 🧭 User journey chart (based on selected user)

UI Improvements:

• ✨ Refactored interface for a smoother experience

[HelissonMota]

🚀 Viya GenAI Monitor Update — New Features Just Dropped!

Today we're excited to share four new features that make monitoring your SAS Viya Copilot activity faster, deeper, and more flexible.

🗜️ Compact Mode
Need to see more at a glance? Compact Mode reduces card padding so you can fit more cards on screen without scrolling.

📅 History Mode: Custom Date Range
The "Custom…" period option is now available in History Mode, letting you define a precise from/to date range to focus on exactly the timeframe you care about.

⚖️ Chat Comparison
Wondering how two conversations stack up? The new Chat Comparison view lets you place two conversations side by side for a clear, direct comparison.

🔍 User Deep-Dive Panel
Click any username on a chat card, to instantly open a per-user activity summary. Great for understanding individual usage patterns at a glance.

Get the Code

Use git commands below to clone only that project.
technical-support-code/usage/administration/sas-gen-ai-monitor-vk at main · sascommunities/technical...

git clone --no-checkout --depth 1 --filter=blob:none \
https://github.com/sascommunities/technical-support-code.git .
git sparse-checkout init --no-cone
git sparse-checkout set /usage/administration/sas-gen-ai-monitor-vk
git checkout

[gabriel_ovalle]

I have been working on a Docker container deployment for the Viya GenAIMonitor. Simply add the following Dockerfile to the git project directory, build it and run it.

FROM python:3.11
WORKDIR /app
COPY . .
RUN chmod +x gen-ai-monitor.sh
CMD ["sh", "-c", "./gen-ai-monitor.sh start && tail -f /dev/null"]
EXPOSE 8899

The commands to build it are as such (you may need to sudo them):

docker build -t genai-monitor .
docker run -d -p 0.0.0.0:8899:8899 genai-monitor

You may need to change how the gen-ai-monitor.sh file is saved, in order to avoid errors with CRLF and LF storage.

From that point, you should be able to easily access the GenAIMonitor tool via localhost.

[vitorlima86]

Thanks a lot Gabriel, I'll coordinate with Helisson to add the Docker file into the repository.