This section provides an overview of frequently asked questions (FAQs) around Solink agents.
Do I need an AI accelerator chip installed on my Solink Recording Device in order to create, edit and/or use Solink agents?
No, an AI chip is not required.
What regions are agents in available in?
Currently agents are available in North America, but Solink is looking to expand into other regions in the near future.
How can I improve agent performance?
It is recommended to create an Event Flow for each agent and to actively respond to each event the agent creates in order to improve performance.
For instance, if you have an agent that checks to see if a customer has been greeted upon entering the store, you could set up an Event Flow to verify if the agent correctly flagged that the customer was/was not greeted (if the agent says "Yes", you can subsequently respond in the Event Flow with your own "Yes" or "No" to confirm or dispute the agent's answer).
Additionally, see the Agents Troubleshooting guide to improve your agent(s) if you encounter sub-optimal performance related to prompts or other technical issues.
See the Agents Prompting guide for help creating better prompts to improve agent performance.
Can agents notify me when it finds something of interest?
Yes - agents automatically flag the activity you are looking for, such as overnight break-ins, suspicious transactions, and safety hazards.
Can I create a custom agent?
Yes - when creating a new agent, select the Custom Solution button to reach out to our team directly to inquire about creating an agent that meets your needs.
How many agents can I have on my account?
You can activate as many agents as needed. However, please ensure to monitor your Agent credits, as agents will not run once your available credits have been exceeded.
How do agent hours (credits) work?
See Agent Hours for more information.
Do agents learn and improve performance over time?
Yes- agents remember what they've already observed and bring that memory into their next analysis.
The Overnight Guard, for instance, won't create a brand-new event every time it sees the same person continuing to loiter. It strings the activity into a single ongoing event. If that loitering then escalates (for example, into trespassing, and then into a break-in), all of it is captured as one continuous event, with the full context attached.