### Answer --- I don't have specific documentation about why an agent would suddenly start reading JSON back to callers. However, this behavior typically occurs when function/tool call responses aren't being handled properly by the LLM. A few things to check: 1. Prompt instructions — Make sure your

general_prompt

or

state_prompt

explicitly instructs the agent to interpret tool results naturally and not read raw data to the caller. [(1) ]() 2. Tool response handling — When a function returns JSON, the LLM should be summarizing or acting on that data, not reciting it. This is usually controlled through prompt engineering. [(2) ]() 3. Model setting — Note that the available models documented include options like

gpt-4.1

,

gpt-4.1-mini

,

gpt-5

,

gpt-5-mini

, etc. [(1) ]() I don't see "GPT 5.1" as a documented model option — you may want to verify your exact model configuration. 4. Response variables — If using custom functions, ensure response variables are properly configured so data is extracted and used contextually rather than passed raw to speech output. [(3) ]() I'd recommend reviewing your prompt to add explicit instructions like "Never read raw JSON or technical data to the caller — always summarize information conversationally." Check out these potentially useful sources for more details: - [Function Calling Overview]() - [Custom Function]() - [Get Retell LLM]() Would you like me to provide a general a

nswer on best practices for handling tool responses in voice agents?

​ To learn more, see the sources I considered: [[1] Get Retell LLM](https://docs.retellai.com/api-references/get-retell-llm) [[2] Function Calling Overview](https://docs.retellai.com/build/single-multi-prompt/function-calling) [[3] Custom Function](https://docs.retellai.com/build/conversation-flow/custom-function) ​

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

Appreciate the feedback! Feel free to ask me again with additional guidance or context.

@AI Evy Are there any network issues with GPT today? I'm curious why this issue just started happening.

### Answer I don't have real-time information about current network issues or outages with GPT or Retell AI. To check for any ongoing issues: - Subscribe to the Retell Status Page at [status.retellai.com]() to get notified about any platform issues [(1) ]() Retell maintains 24/7 latency monitoring and alerting systems, and guarantees >99.9% uptime . If there's a known issue affecting GPT models, it would typically be reflected on the status page. Regarding GPT-5 specifically, there has been at least one community-reported issue where GPT-5 agents got stuck in loops repeating questions — switching back to GPT-4 resolved it in that case [(2) ]() . The changelog also notes that GPT-5 models "can be verbose in replies" [(3) ]() . If you're experiencing sudden behavioral changes, you might try: 1. Checking the status page for any alerts 2. Temporarily switching to GPT-4.1 or GPT-5 Mini to see if the issue persists

​ To learn more, see the sources I considered: [[1] Reliability Overview](https://docs.retellai.com/reliability/reliability-overview) [[2] Issue with GPT-5 model.](https://discord.com/channels/1186424376300679209/1411997452315263059) [[3] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) ​

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

Appreciate the feedback! Feel free to ask me again with additional guidance or context.