Spark is a community support AI bot that helps teams “keep their AI up to date” by updating community knowledge. Once configured, Spark can automatically answer member questions using that knowledge.
In this 2025 Spark update, the emphasis is on making Spark easier to test, easier to configure, and more reliable when searching across knowledge sources—while also migrating the underlying model to Gemini.
What Spark AI does for community support
Spark is positioned as one of the most popular community features for support. The core idea is that Spark updates your community knowledge so the AI can automatically answer member questions.
The update also highlights two common use cases:
- Smaller projects: Spark helps deliver instant answers even when you’re not on Discord 24/7.
- Larger communities: Spark reduces repetitive questions moderators would otherwise answer, helping moderators spend less time on common topics.
Why Spark reduces repetitive questions for moderators
For bigger communities, the “instant answers” benefit isn’t only about speed—it’s also about reducing repetition. When Spark can handle routine member questions, moderators can focus on higher-signal requests rather than answering the same topics repeatedly.
This is part of how the 2025 Spark update is framed: Spark supports community support workflows by answering member questions automatically once the knowledge base is configured.
How to test Spark faster (add the bot + tag)
The video outlines an easier way to get a feel for Spark during testing.
According to the speaker, you can introduce the Spark bot into your community and tag the bot to quickly understand how it works. This makes the initial evaluation simpler before you go deeper into Spark Chat configuration.
If your goal is to validate the experience with real community questions, this “add and tag” approach is positioned as the fastest starting point.
New configuration: multi-agent across multiple channels
Beyond quick testing, the 2025 Spark update adds more customization options through Spark Chat configuration.
One of the key upgrades discussed is support for a multi-agent setup. Instead of relying on a single agent, you can configure multiple agents across multiple channels.
This matters for community structure because different channels often have different purposes (for example, distinct topics or workflows). Multi-agent configuration gives you a way to tailor how Spark responds depending on where members ask.
Expanded knowledge sources (Gbook, text uploads, technical docs)
A major theme in the update is expanding how Spark integrates knowledge.
Gbook integration
The 2025 update includes Gbook integration. This supports using Gbook as a source of knowledge so Spark can reference it when answering questions.
Manual text uploads
The update also adds options to submit knowledge via manual text uploads.
Technical documents (including GitHub-related content)
Additionally, the video mentions adding technical documents, including content you can reference from GitHub.
Overall, the goal is to broaden what Spark can draw on, so your community’s “known answers” can come from more than a single place.
Improved backend search: chunking + Gemini migration
Two important improvements are covered together in the search and model behavior.
Search improvements to reduce hallucinations (chunking)
On the backend, Spark’s search is improved by chunking documents into smaller pieces.
Chunking is described as a way to help reduce hallucinations by improving how the model retrieves relevant segments of information.
Migration from OpenAI to Gemini
The 2025 Spark update also includes a model migration: the video states that Spark’s model was migrated from OpenAI to Gemini.
The speaker ties this migration to practical conversation improvements:
- Better alignment with the team’s speaking style
- Support for longer conversations
- Improved context recall
- Chats that feel more personal
In other words, the Gemini change is presented not only as a backend swap, but as an upgrade to how Spark conversations behave over time.
What the 2025 Spark update changes in practice
Putting the changes together, the 2025 Spark update is designed to improve community support in four ways:
- Faster validation: Add the Spark bot to the community and tag it to test quickly.
- More flexible setup: Configure multiple agents across multiple channels via Spark Chat configuration.
- Broader knowledge integration: Use Gbook, manual text uploads, and technical documents (including GitHub-referenced content).
- More reliable answers: Improve search with document chunking and improve chat behavior through the OpenAI to Gemini migration.
Conclusion
The 2025 Spark update focuses on making community support AI bots easier to test and customize, while improving how Spark finds and uses knowledge. With multi-agent configuration, expanded knowledge sources like Gbook and uploads, chunked search to reduce hallucinations, and a migration to Gemini for better conversation behavior, Spark is positioned to provide more effective automatic answers for both smaller and larger communities.
