[Feat] Add EXA AI Search API to LiteLLM

[ishaan-jaff]

[Feat] Add EXA AI Search API to LiteLLM

Relevant issues

#15314

Provides a unified interface for search, use Exa AI API here

Adds https://docs.exa.ai/reference/search

Relevant issue

#15314

Add Exa AI Search Provider to Unified Search API

Summary

This PR adds Exa AI Search API support to LiteLLM's unified search interface, enabling users to perform web searches using Exa AI alongside existing providers (Perplexity, Tavily, Parallel AI). Exa AI provides intelligent embeddings-based neural search and traditional keyword search, automatically choosing the best approach for each query.

Usage

Basic Search

import litellm

# Simple Exa AI search
response = litellm.search(
    query="Latest research in LLMs",
    custom_llm_provider="exa_ai"
)

for result in response.results:
    print(f"{result.title}: {result.url}")
    print(f"{result.snippet}\n")

With Options

import litellm

# Search with domain filtering and max results
response = litellm.search(
    query="latest AI developments",
    custom_llm_provider="exa_ai",
    max_results=10,
    search_domain_filter=["arxiv.org", "paperswithcode.com"],
    country="US"
)

With Exa-Specific Features

import litellm

# Use Exa-specific parameters (passed through automatically)
response = litellm.search(
    query="AI research papers",
    custom_llm_provider="exa_ai",
    max_results=10,
    type="neural",  # Exa-specific: 'keyword', 'neural', 'fast', 'auto'
    category="research paper",  # Exa-specific: focus on research papers
    startPublishedDate="2024-01-01T00:00:00.000Z",  # Exa-specific
    endPublishedDate="2024-12-31T00:00:00.000Z"  # Exa-specific
)

Async Search

import litellm

response = await litellm.asearch(
    query="machine learning research",
    custom_llm_provider="exa_ai",
    max_results=5
)

Environment Setup

export EXA_API_KEY="your-exa-api-key"

# Optional: Override API base
export EXA_API_BASE="https://api.exa.ai"

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/litellm/ directory, Adding at least 1 test is a hard requirement - see details
• I have added a screenshot of my new test passing locally
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem

Type

🆕 New Feature
✅ Test

Changes

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
litellm	Ready	Preview	Comment	Oct 22, 2025 0:06am