Web Search

Basic Usage

from SimplerLLM.language.llm import LLM, LLMProvider

llm = LLM.create(provider=LLMProvider.OPENAI, model_name="gpt-4o")

response = llm.generate_response(
    prompt="What are the latest developments in AI?",
    web_search=True,
    full_response=True
)

print(response.generated_text)

Note: Set full_response=True to access web sources. Without it, only the generated text is returned.

Supported Providers

Provider	Enum Value	Example Model	Web Search
OpenAI	LLMProvider.OPENAI	gpt-4o	web_search=True
Anthropic	LLMProvider.ANTHROPIC	claude-sonnet-4-5-20250929	web_search=True
Google Gemini	LLMProvider.GEMINI	gemini-2.5-flash	web_search=True
Perplexity	LLMProvider.PERPLEXITY	sonar-pro	Always enabled

Accessing Sources

Web sources are available on the web_sources field when using full_response=True:

response = llm.generate_response(
    prompt="Recent breakthroughs in quantum computing",
    web_search=True,
    full_response=True
)

for source in response.web_sources or []:
    print(f"{source['title']}: {source['url']}")

Each source contains:

Key	Description
title	Source page title
url	Source URL

Switching Providers

The web_search parameter works the same across providers:

from SimplerLLM.language.llm import LLM, LLMProvider

# OpenAI
llm = LLM.create(provider=LLMProvider.OPENAI, model_name="gpt-4o")

# Anthropic
llm = LLM.create(provider=LLMProvider.ANTHROPIC, model_name="claude-sonnet-4-5-20250929")

# Gemini
llm = LLM.create(provider=LLMProvider.GEMINI, model_name="gemini-2.5-flash")

All use the same call:

response = llm.generate_response(
    prompt="Latest news on renewable energy",
    web_search=True,
    full_response=True
)

Perplexity

Perplexity models always include web search -- no need to set web_search=True.

from SimplerLLM.language.llm import LLM, LLMProvider

llm = LLM.create(provider=LLMProvider.PERPLEXITY, model_name="sonar-pro")

response = llm.generate_response(
    prompt="What happened in tech this week?",
    search_recency_filter="week",
    search_domain_filter=["techcrunch.com", "theverge.com"],
    full_response=True
)

print(response.generated_text)

for source in response.web_sources or []:
    print(f"{source['title']}: {source['url']}")

Parameter	Type	Description
search_domain_filter	list[str]	Limit to specific domains (prefix with - to exclude)
search_recency_filter	str	"day", "week", "month", or "year"
return_images	bool	Include images in search results
return_related_questions	bool	Suggest related queries

Available models: sonar, sonar-pro, sonar-reasoning-pro, sonar-deep-research.

Structured Output with Web Search

Combine web search with Pydantic model generation:

from pydantic import BaseModel
from SimplerLLM.language.llm import LLM, LLMProvider
from SimplerLLM.language.llm_addons import generate_pydantic_json_model

class ResearchSummary(BaseModel):
    topic: str
    key_findings: list[str]
    sources_used: int

llm = LLM.create(provider=LLMProvider.OPENAI, model_name="gpt-4o")

response = generate_pydantic_json_model(
    model_class=ResearchSummary,
    prompt="Summarize recent advances in quantum computing",
    llm_instance=llm,
    web_search=True,
    full_response=True
)

print(response.model_object.key_findings)

for source in response.web_sources or []:
    print(f"{source['title']}: {source['url']}")

See Structured Output for more on Pydantic model generation.

Vision + Web Search

Combine image analysis with web search to get context-aware responses about visual content:

from SimplerLLM.language.llm import LLM, LLMProvider

llm = LLM.create(provider=LLMProvider.OPENAI, model_name="gpt-4o")

response = llm.generate_response(
    prompt="What is this landmark and what are its current visiting hours?",
    images=["https://example.com/landmark.jpg"],
    web_search=True,
    full_response=True
)

print(response.generated_text)

for source in response.web_sources or []:
    print(f"{source['title']}: {source['url']}")

This also works with structured output:

from pydantic import BaseModel
from SimplerLLM.language.llm_addons import generate_pydantic_json_model

class LandmarkInfo(BaseModel):
    name: str
    location: str
    visiting_hours: str
    description: str

result = generate_pydantic_json_model(
    model_class=LandmarkInfo,
    prompt="Identify this landmark and find its current visiting hours",
    llm_instance=llm,
    images=["path/to/photo.jpg"],
    web_search=True
)

print(result.name)
print(result.visiting_hours)

Async Usage

import asyncio
from SimplerLLM.language.llm import LLM, LLMProvider

llm = LLM.create(provider=LLMProvider.OPENAI, model_name="gpt-4o")

async def main():
    response = await llm.generate_response_async(
        prompt="Latest AI safety research",
        web_search=True,
        full_response=True
    )
    print(response.generated_text)

asyncio.run(main())