Perplexity Integration

Perform web searches and leverage Perplexity's advanced LLM models for real-time information retrieval.

---

Timbal integrates with Perplexity to provide advanced search and LLM capabilities.

This integration allows you to perform web searches and leverage Perplexity's models directly within your Timbal workflows.

Prerequisites

Before using the Perplexity integration, you'll need:

1. A Perplexity API key – Get your API key here
2. Store your obtained API key in an environment variable named PERPLEXITY_API_KEY to facilitate its use by the tools.

Search

Description

The search step allows you to perform a web search or interact with Perplexity's LLMs using a variety of parameters for customization.

Example

from timbal.steps.perplexity import search

result = await search(query="What is the capital of France?")

Parameters

Parameter	Type	Description	Required
query	str	Query to search for.	Yes
model	str	Model to use. Default: sonar.	No
system_prompt	str	System prompt to guide the LLM's behavior and role.	No
temperature	float	Amount of randomness in the response. Ranges from 0 to 2. Default: 0.2	No
top_p	float	Nucleus sampling threshold. Ranges from 0 to 1. Default: 0.9	No
search_domain_filter	list/None	Limit citations to specific domains. Only available in certain tiers.	No
return_images	bool	Whether to return images in the response. Only available in certain tiers.	No
return_related_questions	bool	Whether to return related questions. Only available in certain tiers.	No
search_recency_filter	Literal['month', 'week', 'day', 'hour']	Restrict search results to a recent time interval.	No
top_k	int	Number of tokens to keep for top-k filtering. 0 disables.	No
stream	bool	Whether to stream the response.	No
presence_penalty	float	Penalize new tokens based on presence. Ranges from -2.0 to 2.0.	No
frequency_penalty	float	Penalize new tokens based on frequency. >1.0 discourages repetition.	No
response_format	dict/None	Enable structured outputs with a JSON or Regex schema.	No

Agent Integration Example

from timbal.steps.perplexity import search
from timbal import Agent

agent = Agent(
  tools=[search]
)

response = await agent.complete(prompt="What are the latest AI trends?")

Notes

• For more advanced usage, see the Perplexity API documentation.