### Answer Based on the available sources, I don't have specific documentation for Plivo integration with Retell AI. However, I can provide general guidance for SIP trunking setup that should help resolve your "destination not found" error. For custom telephony providers using elastic SIP trunking, you need to configure both origination (inbound) and termination (outbound) settings [(1) ]() . Key configuration points: Retell SIP server URI:

sip:sip.retellai.com

IP block for traffic:

18.98.16.120/30

(All regions) ,

143.223.88.0/21

(certain United States traffic) ,

161.115.160.0/19

(certain United States traffic) Transport methods supported: - TCP (Recommended) :

sip:sip.retellai.com;transport=tcp

- UDP:

sip:sip.retellai.com;transport=udp

- TLS:

sip:sip.retellai.com;transport=tls

For your "destination not found" error, check: 1. Origination setup in Plivo - Ensure you've specified Retell's SIP server address

sip.retellai.com

as the origination SIP URI [(1) ]() [(2) ]() 2. Termination setup - Verify the termination SIP URI, username, and password are correctly configured in Plivo and imported into Retell 3. Number import - After configuring your SIP trunk in Plivo, you need to import the number to Retell so it knows how to route the call For more detailed troubleshooting, check out these potentially useful sources: - [Custom Telephony Overview]() - [Vonage]() (similar SIP trunking setup process) Would you like me to provide a general answer based on my knowledge instead?

​ To learn more, see the sources I considered: [[1] Custom Telephony Overview](https://docs.retellai.com/deploy/custom-telephony) [[2] Vonage](https://docs.retellai.com/deploy/vonage) ​

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

### Answer You cannot directly override the prompt itself via the inbound webhook, but you can override agent-level settings that affect how the agent behaves [(1) ]() . The inbound webhook response supports these override fields : - `override_agent_id`: Override which agent handles the call - `override_agent_version`: Override the agent version - `dynamic_variables`: Set dynamic variables for personalization - `metadata`: Set metadata for the call - `agent_override`: Override specific agent settings per-call Using

agent_override

for prompt-like behavior: The

agent_override

object lets you modify agent behavior without changing the saved agent . For Retell LLM agents, you can override : - `retell_llm.begin_message`: The opening greeting - `retell_llm.model`: The LLM model used - `retell_llm.model_temperature`: Response creativity - `retell_llm.knowledge_base_ids`: Which knowledge bases to use For Conversation Flow agents, similar overrides apply under

conversation_flow

. Example response to modify agent behavior:

json
{
  "call_inbound": {
    "agent_override": {
      "retell_llm": {
        "begin_message": "Hi {{customer_name}}, thanks for calling.",
        "model_temperature": 0.2
      }
    },
    "dynamic_variables": {
      "customer_name": "John Doe"
    }
  }
}

While you cannot replace the entire prompt, these overrides combined with dynamic variables give you significant control over how the agent responds on each call .

​ To learn more, see the sources I considered: [[1] Inbound webhook](https://docs.retellai.com/features/inbound-call-webhook) ​

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

### Answer I don't have information in the available sources explaining the specific reasoning behind why Retell AI designed the inbound webhook to not support full prompt override [(1) ]() . The documentation shows what is supported through the inbound webhook - including

override_agent_id

,

override_agent_version

,

dynamic_variables

,

metadata

, and

agent_override

for partial agent settings - but doesn't explain the design rationale for these specific limitations. What you can do instead: If you need different prompts for different inbound scenarios, you have these options: 1. Use

override_agent_id

- Create multiple agents with different prompts and route to them based on your inbound webhook logic [(1) ]() [(2) ]() [(3) ]() [(4) ]() 2. Use dynamic variables - Inject context-specific information that your prompt can reference using

{{variable_name}}

syntax [(1) ]() [(5) ]() 3. Use

agent_override

- Modify specific agent settings like

begin_message

,

model_temperature

, or

knowledge_base_ids

per-call [(1) ]() For the architectural reasoning behind this design choice, you might want to reach out to Retell support directly or check their community forums for discussions on this topic.

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

​ To learn more, see the sources I considered: [[1] Inbound webhook](https://docs.retellai.com/features/inbound-call-webhook) [[2] Discord message from @katecardenas_67927 on 2025-09-17](https://discord.com/channels/1186424376300679209/1186462578310910112/1417939440860528783) [[3] Discord message from @dedmeat on 2025-09-09](https://discord.com/channels/1186424376300679209/1186462578310910112/1415076614563303506) [[4] Importing existing phone number to multiple agents via API](https://discord.com/channels/1186424376300679209/1418871771045564509) [[5] Dynamic Variables](https://docs.retellai.com/build/dynamic-variables) ​

### Answer Yes, this approach should work. Dynamic variables are injected into your Retell LLM prompt and tool descriptions [(1) ]() [(2) ]() , and you can reference them using the

{{variable_name}}

syntax [(3) ]() . How it works: When you set

general_prompt

to

{{system_prompt}}

[(1) ]() [(2) ]() , and then provide the actual prompt content through dynamic variables, it will be injected at runtime [(1) ]() [(4) ]() [(5) ]() . Example setup: Create your Retell LLM with:

json
{
  "general_prompt": "{{system_prompt}}"
}

[(1) ]() Then when making a call or via inbound webhook, pass the full prompt:

json
{
  "retell_llm_dynamic_variables": {
    "system_prompt": "You are a helpful assistant. Your goal is to..."
  }
}

[(6) ]() [(4) ]() Important notes: - Dynamic variables work in

general_prompt

,

begin_message

, and tool descriptions [(1) ]() [(2) ]() [(3) ]() - They're represented as key-value pairs of strings [(1) ]() [(4) ]() [(5) ]() - You can set them via inbound webhook using the

dynamic_variables

field (for calls) or

retell_llm_dynamic_variables

(when creating calls directly) [(6) ]() [(4) ]() This gives you the flexibility to dynamically control your entire system prompt per call while maintaining a single agent configuration.

​ To learn more, see the sources I considered: [[1] Create Retell LLM](https://docs.retellai.com/api-references/create-retell-llm) [[2] Update Retell LLM](https://docs.retellai.com/api-references/update-retell-llm) [[3] Issue using retell_llm_dynamic_variables](https://discord.com/channels/1186424376300679209/1223405124882534470) [[4] Create Phone Call](https://docs.retellai.com/api-references/create-phone-call) [[5] Register Phone Call](https://docs.retellai.com/api-references/register-phone-call) [[6] Inbound webhook](https://docs.retellai.com/features/inbound-call-webhook) ​

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

### Answer Based on the available Discord sources, I found relevant discussions about using dynamic variables for inbound calls, though not specifically about using them to replace the entire system prompt [(1) ]() [(2) ]() . Key findings from Discord discussions: The community has discussed setting dynamic variables for inbound calls through the inbound webhook . One user specifically asked about populating dynamic variables during inbound calls and was directed to use the inbound webhook approach [(2) ]() . Important limitation to be aware of: There's a documented issue where dynamic variables via inbound webhook break if non-string values are sent [(3) ]() . The documentation clearly states that all values in

retell_llm_dynamic_variables

must be strings - numbers, booleans, or other data types are not supported . For your specific approach: While I don't see a Discord discussion specifically about using

{{system_prompt}}

as the entire general_prompt, the approach is technically supported by the platform. Dynamic variables can be used in prompts, begin messages, and tool configurations [(4) ]() . To set dynamic variables for inbound calls: You can supply dynamic variables in the inbound webhook response [(5) ]() [(4) ]() :

json
{
  "call_inbound": {
    "dynamic_variables": {
      "system_prompt": "Your full prompt text here..."
    }
  }
}

[(5) ]() Critical requirement: Ensure your

system_prompt

value is a string. Any attempt to pass non-string values will cause the dynamic variables to fail [(3) ]() .

​ To learn more, see the sources I considered: [[1] Dynamic Variables For Inbound Calls](https://discord.com/channels/1186424376300679209/1352197272770973786) [[2] Populating Dynamic Variables During An Inbound Call](https://discord.com/channels/1186424376300679209/1228587811188379649) [[3] Dynamic Variables via inbound webhook break if non-string values are sent](https://discord.com/channels/1186424376300679209/1420429628979282011) [[4] Dynamic Variables](https://docs.retellai.com/build/dynamic-variables) [[5] Inbound webhook](https://docs.retellai.com/features/inbound-call-webhook) ​

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

​ To learn more, see the sources I considered: [[1] Create Retell LLM](https://docs.retellai.com/api-references/create-retell-llm) [[2] Get Retell LLM](https://docs.retellai.com/api-references/get-retell-llm) [[3] List Retell LLMs](https://docs.retellai.com/api-references/list-retell-llms) [[4] Update Retell LLM](https://docs.retellai.com/api-references/update-retell-llm) ​

### Answer Yes, you can use one Retell LLM Response Engine across multiple agents [(1) ]() [(2) ]() [(3) ]() . The Retell LLM is a separate resource that you create and can then attach to different agents [(1) ]() . Each Retell LLM has a unique

llm_id

that you can reference when configuring agents [(2) ]() [(1) ]() . How it works: When you create a Retell LLM, it returns an

llm_id

[(1) ]() :

json
{
  "llm_id": "oBeDLoLOeuAbiuaMFXRtDOLriTJ5tSxD",
  "last_modification_timestamp": 1703413636133,
  "version": 1,
  "is_published": false,
  "model": "gpt-4.1"
}

You can then use this same

llm_id

when creating or configuring multiple agents [(2) ]() [(1) ]() . This allows you to maintain consistent LLM behavior (prompts, tools, states, etc.) across different agents while still being able to customize other agent-specific settings like voice, phone numbers, or webhook configurations [(1) ]() [(2) ]() . Benefits of this approach: - Centralized prompt management - update the LLM once and all connected agents reflect the changes [(4) ]() - Consistency across multiple agents [(2) ]() [(3) ]() - Easier testing and iteration [(4) ]() You can view all your Retell LLMs using the list endpoint [(3) ]() , and update them as needed [(4) ]() .

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

### Answer Whether sharing one Retell LLM across multiple agents is a good practice depends on your specific use case [(1) ]() . When it's beneficial: - **Consistent behavior needed**: If multiple agents should handle conversations similarly with the same prompts, tools, and logic - **Centralized management**: Update the LLM once and all connected agents reflect the changes - **Easier maintenance**: Single source of truth for your conversational logic When it might not be ideal: - **Different use cases**: If agents serve distinct purposes (e.g., sales vs. support vs. appointment booking) , separate LLMs allow better customization [(2) ]() - **A/B testing**: Testing different prompt strategies requires separate LLM configurations [(1) ]() - **Complex state management**: If agents need very different conversation flows, separate LLMs with their own states make more sense [(1) ]() [(3) ]() Best practice recommendation: Based on the Prompt Engineering Guide, use sectional prompts and consider Conversation Flow for complex tasks [(2) ]() . If your agents share the same core logic but need minor variations, one LLM with dynamic variables (like your

{{system_prompt}}

approach) provides flexibility [(1) ]() . However, if agents have fundamentally different purposes or require distinct tool sets, separate LLMs offer cleaner separation and easier debugging [(2) ]() [(1) ](

/api-references/update-retell-llm) . **

​ To learn more, see the sources I considered: [[1] Update Retell LLM](https://docs.retellai.com/api-references/update-retell-llm) [[2] Prompt Engineering Guide](https://docs.retellai.com/build/prompt-engineering-guide) [[3] Introducing Retell LLM Beta](https://www.retellai.com/changelog/introducing-retell-llm-beta) ​

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