### Answer Based on the available sources, there isn't official documentation specifically explaining Retell AI's caching mechanism. However, a community discussion touches on this issue: Retell AI may cache certain responses for performance reasons. Currently, there isn't a direct method provided in the documentation to clear this cache. [(1) ]() Here are some steps that may help resolve your agent being stuck on 2025 references: 1. Update the agent's configuration or re-deploy it — this might help clear residual cached data 2. Check for fine-tuning examples — someone on your team might have added a fine-tuning example in the agent config that takes precedence over other data sources, effectively masking related entries 3. Create a new agent for testing — this can help determine if the issue is with the specific agent's configuration or a broader system behavior 4. Ensure complete removal and re-addition of any Knowledge Base — residual links can sometimes persist If the issue persists after trying these steps, you may want to reach out to Retell support directly for assistance with cache clearing. Would you like me to provide a general answer about how LLM date references typically work, or would you prefer to contact Retell support for more specific guidance?

​ To learn more, see the sources I considered: [[1] knowledge base caching](https://discord.com/channels/1186424376300679209/1360231379933069362) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer LLMs have a knowledge cutoff date, which means they cannot internally know the current date or accurately determine relative references like "today," "tomorrow," or "upcoming January." [(1) ]() How to fix date reference issues in Retell AI: 1. Use dynamic variables — Add

Current time is {{current_time}}

to your prompt and the LLM will know the current date and time. 2. Use the calendar variable — For more comprehensive date awareness, add

{{current_calendar_[timezone]}}

to your prompt. This provides the AI with the correct calendar for the next 14 days in your specified timezone. [(2) ]() 3. Inject via API — If you're using the API, you can embed the current date using dynamic variables through the

retell_llm_dynamic_variables

parameter or

default_dynamic_variables

in your LLM configuration. [(1) ]() [(3) ]() Example of setting default dynamic variables in your LLM config:

json
"default_dynamic_variables": {
  "customer_name": "John Doe"
}

[(3) ]() For your specific issue where the agent is stuck on 2025 instead of 2026, ensure your dynamic variable or prompt includes the correct current date reference, and consider updating your agent configuration to refresh any cached prompt data.

​ To learn more, see the sources I considered: [[1] Date Recognition Issue in the LLM](https://discord.com/channels/1186424376300679209/1293970874365640864) [[2] Returning Incorrect DAY when in ISO Format.](https://discord.com/channels/1186424376300679209/1418668209443180675) [[3] Get Retell LLM](https://docs.retellai.com/api-references/get-retell-llm) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.