Parallel

Use Parallel with Kern for AI-optimized web search and content extraction.

ParallelTools enable an Agent to perform AI-optimized web search and content extraction using Parallel's APIs.

Prerequisites

The following example requires the parallel-web library and an API key which can be obtained from Parallel.

1uv pip install -U parallel-web

1export PARALLEL_API_KEY=***

Example

The following agent will search for information on AI agents and autonomous systems, then extract content from specific URLs:

1from kern.agent import Agent
2from kern.tools.parallel import ParallelTools
3
4agent = Agent(
5    tools=[
6        ParallelTools(
7            enable_search=True,
8            enable_extract=True,
9            max_results=5,
10            max_chars_per_result=8000,
11        )
12    ],
13    markdown=True,
14)
15
16# Should use parallel_search
17agent.print_response(
18    "Search for the latest information on 'AI agents and autonomous systems' and summarize the key findings"
19)
20
21# Should use parallel_extract
22agent.print_response(
23    "Extract information about the product features from https://parallel.ai and https://docs.parallel.ai"
24)

Toolkit Params

Parameter	Type	Default	Description
`api_key`	`Optional[str]`	`None`	Parallel API key. If not provided, will use PARALLEL_API_KEY environment variable.
`enable_search`	`bool`	`True`	Enable Search API functionality for AI-optimized web search.
`enable_extract`	`bool`	`True`	Enable Extract API functionality for content extraction from URLs.
`all`	`bool`	`False`	Enable all tools. Overrides individual flags when True.
`max_results`	`int`	`10`	Default maximum number of results for search operations.
`max_chars_per_result`	`int`	`10000`	Default maximum characters per result for search operations.
`beta_version`	`str`	`"search-extract-2025-10-10"`	Beta API version header.
`mode`	`Optional[str]`	`None`	Default search mode. Options: "one-shot" or "agentic".
`include_domains`	`Optional[List[str]]`	`None`	Default domains to restrict results to.
`exclude_domains`	`Optional[List[str]]`	`None`	Default domains to exclude from results.
`max_age_seconds`	`Optional[int]`	`None`	Default cache age threshold. When set, minimum value is 600 seconds.
`timeout_seconds`	`Optional[float]`	`None`	Default timeout for content retrieval.
`disable_cache_fallback`	`Optional[bool]`	`None`	Default cache fallback behavior.

Toolkit Functions

Function Description

parallel_search Search the web using Parallel's Search API with natural language objective. Parameters include objective (str) for natural-language description, search_queries (list) for traditional keyword queries, max_results (int) for maximum number of results, and max_chars_per_result (int) for maximum characters per result. Returns JSON formatted string with URLs, titles, publish dates, and relevant excerpts.

parallel_extract Extract content from specific URLs using Parallel's Extract API. Parameters include urls (list) for URLs to extract from, objective (str) to guide extraction, search_queries (list) for targeting relevant content, excerpts (bool, default=True) to include text snippets, max_chars_per_excerpt (int) to limit excerpt characters, full_content (bool, default=False) to include complete page text, and max_chars_for_full_content (int) to limit full content characters. Returns JSON formatted string with extracted content in clean markdown format.

Function	Description
`parallel_search`	Search the web using Parallel's Search API with natural language objective. Parameters include `objective` (str) for natural-language description, `search_queries` (list) for traditional keyword queries, `max_results` (int) for maximum number of results, and `max_chars_per_result` (int) for maximum characters per result. Returns JSON formatted string with URLs, titles, publish dates, and relevant excerpts.
`parallel_extract`	Extract content from specific URLs using Parallel's Extract API. Parameters include `urls` (list) for URLs to extract from, `objective` (str) to guide extraction, `search_queries` (list) for targeting relevant content, `excerpts` (bool, default=True) to include text snippets, `max_chars_per_excerpt` (int) to limit excerpt characters, `full_content` (bool, default=False) to include complete page text, and `max_chars_for_full_content` (int) to limit full content characters. Returns JSON formatted string with extracted content in clean markdown format.

Parallel

Prerequisites

Example

Toolkit Params

Toolkit Functions

Developer Resources