> Fetch clean Markdown by appending `.md` to any page URL under https://signalwire.com/docs or requesting it with the HTTP header `Accept: text/markdown`. The root index at https://signalwire.com/docs/llms.txt lists the available documentation indexes.

# set_prompt_pom

> Set the prompt using a raw Prompt Object Model dictionary structure.

[prompt-add-section]: /docs/server-sdks/reference/python/agents/agent-base/prompt-add-section

[ref-agentbase]: /docs/server-sdks/reference/python/agents/agent-base

Set the prompt directly as a POM (Prompt Object Model) dictionary structure. This is
a low-level method that accepts the POM data format directly. For most cases, use
[`prompt_add_section()`][prompt-add-section]
to build POM prompts incrementally.

> **Info**
>
> The POM format structures prompts into titled sections with body text, bullet points,
> and nested subsections. This structure helps the AI model parse complex instructions
> more reliably than flat text.

## **Parameters**

**`pom`** `PromptObjectModel | list[dict[str, Any]]` — required

POM structure. A list of section dictionaries is converted to a
`PromptObjectModel` before it is stored; an existing `PromptObjectModel` is
accepted as-is. Each section dictionary has keys:

* `title` (str) -- Section heading
* `body` (str) -- Section body text
* `bullets` (list\[str]) -- Optional bullet points
* `numbered` (bool) -- Whether the section is numbered
* `subsections` (list\[dict]) -- Optional nested sections with the same keys

---

## **Returns**

[`AgentBase`][ref-agentbase] -- Returns self for method chaining.

## **Example**

```python {4}
from signalwire import AgentBase

agent = AgentBase(name="support", route="/support")
agent.set_prompt_pom([
    {
        "title": "Role",
        "body": "You are a customer support agent for Acme Corp."
    },
    {
        "title": "Guidelines",
        "body": "",
        "bullets": [
            "Always greet the caller by name if available",
            "Never share internal pricing details",
            "Escalate to a human if the caller requests it"
        ]
    },
    {
        "title": "Capabilities",
        "body": "You can help with the following:",
        "subsections": [
            {"title": "Orders", "body": "Look up order status, process returns"},
            {"title": "Billing", "body": "Check balances, explain charges"}
        ]
    }
])
agent.serve()
```

> Set the prompt using a raw Prompt Object Model dictionary structure.