🐫 CAMEL: The first and the best multi-agent framework. Finding the Scaling Law of Agents. https://www.camel-ai.org
Python agentai-societiesartificial-intelligencecommunicative-aicooperative-aideep-learninglarge-language-modelsmulti-agent-systemsnatural-language-processing
<a id="camel.toolkits.hybrid_browser_toolkit_py.agent"></a>
<a id="camel.toolkits.hybrid_browser_toolkit_py.agent.PlaywrightLLMAgent"></a>
## PlaywrightLLMAgent
```python
class PlaywrightLLMAgent:
```
High-level orchestration layer connecting page snapshots, an LLM,
and an action executor for browser automation.
**Parameters:**
- **user_data_dir** (str, optional): Path to the browser user data directory for persistent sessions. (default: :obj:`None`)
- **headless** (bool, optional): Whether to run the browser in headless mode. (default: :obj:`False`)
- **stealth** (bool, optional): Whether to enable stealth mode to avoid bot detection. (default: :obj:`False`)
- **model_backend** (BaseModelBackend, optional): The model backend to use for LLM calls. If :obj:`None`, defaults to :obj:`ModelFactory` with :obj:`ModelPlatformType.DEFAULT` and :obj:`ModelType.DEFAULT`. (default: :obj:`None`)
- **default_timeout** (int, optional): Default timeout in milliseconds for browser actions. If :obj:`None`, uses the configured default. (default: :obj:`None`)
- **short_timeout** (int, optional): Short timeout in milliseconds for quick browser operations. If :obj:`None`, uses the configured default. (default: :obj:`None`)
<a id="camel.toolkits.hybrid_browser_toolkit_py.agent.PlaywrightLLMAgent.__init__"></a>
### __init__
```python
def __init__(self):
```
<a id="camel.toolkits.hybrid_browser_toolkit_py.agent.PlaywrightLLMAgent._get_chat_agent"></a>
### _get_chat_agent
```python
def _get_chat_agent(self):
```
**Returns:**
ChatAgent: The existing or newly created :obj:`ChatAgent`
initialized with the system prompt and model backend.
<a id="camel.toolkits.hybrid_browser_toolkit_py.agent.PlaywrightLLMAgent._safe_parse_json"></a>
### _safe_parse_json
```python
def _safe_parse_json(self, content: str):
```
Safely parse JSON from an LLM response with multiple fallback
strategies.
Attempts direct parsing first, then regex-based extraction, then
line-by-line parsing. Falls back to a default structure if all
strategies fail.
**Parameters:**
- **content** (str): The raw LLM response string to parse.
**Returns:**
Dict[str, Any]: The parsed JSON as a dictionary, or a fallback
response structure if parsing fails.
<a id="camel.toolkits.hybrid_browser_toolkit_py.agent.PlaywrightLLMAgent._get_fallback_response"></a>
### _get_fallback_response
```python
def _get_fallback_response(self, error_msg: str):
```
Generate a fallback response structure for parsing failures.
**Parameters:**
- **error_msg** (str): A description of the error that caused the fallback.
**Returns:**
Dict[str, Any]: A default response dictionary with a
`__INLINE_CODE_0__` action and the error message as summary.
<a id="camel.toolkits.hybrid_browser_toolkit_py.agent.PlaywrightLLMAgent._llm_call"></a>
### _llm_call
```python
def _llm_call(
self,
prompt: str,
snapshot: str,
is_initial: bool,
history: Optional[List[Dict[str, Any]]] = None
):
```
Call the LLM via :obj:`ChatAgent` to get a plan and next action.
**Parameters:**
- **prompt** (str): The user task description.
- **snapshot** (str): The current page snapshot text.
- **is_initial** (bool): Whether this is the first LLM call for the task. If :obj:`True`, no action history is included.
- **history** (List[Dict[str, Any]], optional): List of previous actions and their results. Only used when `__INLINE_CODE_0__`
- **is**: obj:`False`. (default: :obj:`None`)
**Returns:**
Dict[str, Any]: The parsed LLM response containing `__INLINE_CODE_0__`
(List[str]) and `__INLINE_CODE_1__` (Dict[str, Any]).