Skip to main content
Junkie integrates powerful web search capabilities through Perplexity Sonar Pro and Exa Search, enabling agents to fetch accurate, real-time, source-backed information from the live web.

​
Overview

Web search tools provide:
  • Perplexity Sonar Pro: Real-time web search with source citations
  • Exa Search: Semantic search for research and analysis
  • Agent Integration: Dedicated search agents for team workflows
  • Real-time Data: Access to live web information

​
Perplexity Sonar Pro

Perplexity Sonar Pro provides real-time web search with accurate, source-backed information.

​
Configuration

# .env file
PROVIDER="https://your-openai-compatible-api.com"
CUSTOM_PROVIDER_API_KEY="your-api-key"

​
Agent Setup

From agent_factory.py:234-245: 
from agno.agent import Agent
from agno.models.openai import OpenAILike

perplexity_agent = Agent(
    id="pplx-agent",
    name="Perplexity Sonar Pro",
    model=OpenAILike(
        id="sonar-pro",
        base_url=PROVIDER,
        api_key=CUSTOM_PROVIDER_API_KEY
    ),
    add_datetime_to_context=True,
    timezone_identifier="Asia/Kolkata",
)
​
id
str
required
Agent identifier. Use "pplx-agent" for Perplexity agent.
​
name
str
required
Human-readable agent name.
​
model
OpenAILike
required
Model configuration with id="sonar-pro" for Perplexity.
​
add_datetime_to_context
bool
default:"False"
Include current date/time in agent context.
​
timezone_identifier
str
default:"UTC"
Timezone for datetime context (e.g., "Asia/Kolkata").

​
Model Configuration

from agno.models.openai import OpenAILike

sonar_model = OpenAILike(
    id="sonar-pro",
    base_url=PROVIDER,
    api_key=CUSTOM_PROVIDER_API_KEY
)
​
id
str
required
Model ID. Use "sonar-pro" for Perplexity Sonar Pro.
​
base_url
str
required
API endpoint URL for the model provider.
​
api_key
str
required
API key for authentication.

​
Usage Example

from agno.agent import Agent
from agno.models.openai import OpenAILike
from core.config import PROVIDER, CUSTOM_PROVIDER_API_KEY

perplexity_agent = Agent(
    id="pplx-agent",
    name="Perplexity Sonar Pro",
    model=OpenAILike(
        id="sonar-pro",
        base_url=PROVIDER,
        api_key=CUSTOM_PROVIDER_API_KEY
    ),
    add_datetime_to_context=True,
    timezone_identifier="Asia/Kolkata",
)

# Agent can now search the web
response = await perplexity_agent.run(
    "What are the latest developments in AI?"
)
print(response.content)
Exa provides semantic search capabilities for research and competitive analysis.

​
Configuration

# .env file
EXA_API_KEY="your-exa-api-key"

​
Toolkit Setup

from agno.tools.exa import ExaTools

exa_tools = ExaTools()
No parameters required - API key is loaded from EXA_API_KEY environment variable.

​
Agent Integration

From agent_factory.py:182-187: 
from agno.tools.exa import ExaTools

code_agent_tools = [
    MCPTools(transport="streamable-http", url="https://mcp.context7.com/mcp"),
    e2b_toolkit,
    ExaTools(),  # Add Exa search capabilities
]

code_agent = Agent(
    name="Code Agent",
    model=my_model,
    tools=code_agent_tools
)

​
Usage Example

from agno.agent import Agent
from agno.tools.exa import ExaTools

search_agent = Agent(
    name="Search Agent",
    model=my_model,
    tools=[ExaTools()],
    instructions="Use Exa to search for relevant information and research."
)

# Agent can now use Exa search
response = await search_agent.run(
    "Find recent research papers on transformer architectures"
)

​
Team Integration

Integrating web search agents into a team workflow: 
from agno.team import Team
from agno.agent import Agent
from agno.models.openai import OpenAILike
from agno.tools.exa import ExaTools

# Perplexity agent for real-time web search
perplexity_agent = Agent(
    id="pplx-agent",
    name="Perplexity Sonar Pro",
    model=OpenAILike(
        id="sonar-pro",
        base_url=PROVIDER,
        api_key=CUSTOM_PROVIDER_API_KEY
    ),
    add_datetime_to_context=True,
    timezone_identifier="Asia/Kolkata",
)

# Code agent with Exa search
code_agent = Agent(
    id="code-agent",
    name="Code Agent",
    model=my_model,
    tools=[ExaTools()],
)

# Create team with search capabilities
team = Team(
    name="Research Team",
    model=leader_model,
    members=[perplexity_agent, code_agent],
    instructions="Delegate search tasks to Perplexity agent for real-time info."
)

​
Use Cases

​
Real-Time News

perplexity_agent = Agent(
    name="Perplexity Sonar Pro",
    model=OpenAILike(id="sonar-pro", base_url=PROVIDER, api_key=API_KEY),
)

response = await perplexity_agent.run(
    "What are today's top technology news stories?"
)

​
Competitive Analysis

from agno.tools.exa import ExaTools

research_agent = Agent(
    name="Research Agent",
    model=my_model,
    tools=[ExaTools()],
)

response = await research_agent.run(
    "Compare pricing models of top AI API providers"
)

​
Fact Checking

perplexity_agent = Agent(
    name="Perplexity Sonar Pro",
    model=OpenAILike(id="sonar-pro", base_url=PROVIDER, api_key=API_KEY),
    add_datetime_to_context=True,
)

response = await perplexity_agent.run(
    "Verify the latest unemployment statistics for the US"
)

​
Research Papers

from agno.tools.exa import ExaTools

research_agent = Agent(
    name="Research Agent",
    model=my_model,
    tools=[ExaTools()],
)

response = await research_agent.run(
    "Find recent papers on large language model alignment"
)

​
Best Practices

  • Use Perplexity Sonar Pro for real-time, source-backed searches
  • Use Exa for semantic search and research tasks
  • Combine both for comprehensive search coverage
  • Consider cost and latency trade-offs
  • Always set add_datetime_to_context=True for time-sensitive searches
  • Configure appropriate timezone_identifier
  • Enables accurate “today”, “recent”, “latest” queries
  • Important for news and real-time information
  • Provide clear instructions about when to use search
  • Specify source citation requirements
  • Define search scope and depth
  • Guide agents on handling ambiguous queries
  • Delegate search tasks to specialized agents
  • Use Perplexity agent for real-time queries
  • Use code agent with Exa for research
  • Avoid redundant searches across agents

​
Environment Variables

# .env file

# Provider configuration (for Perplexity)
PROVIDER="https://your-openai-compatible-api.com"
CUSTOM_PROVIDER_API_KEY="your-api-key"

# Exa Search
EXA_API_KEY="your-exa-api-key"

# Optional: Custom timezone
TIMEZONE="Asia/Kolkata"

​
Complete Example

Full implementation from agent_factory.py: 
from agno.agent import Agent
from agno.team import Team
from agno.models.openai import OpenAILike
from agno.tools.exa import ExaTools
from tools.e2b_tools import E2BToolkit
from agno.tools.mcp import MCPTools

# Perplexity Sonar Pro agent
perplexity_agent = Agent(
    id="pplx-agent",
    name="Perplexity Sonar Pro",
    model=OpenAILike(
        id="sonar-pro",
        base_url=PROVIDER,
        api_key=CUSTOM_PROVIDER_API_KEY
    ),
    add_datetime_to_context=True,
    timezone_identifier="Asia/Kolkata",
)

# Code agent with Exa search
code_agent_tools = [
    MCPTools(transport="streamable-http", url="https://mcp.context7.com/mcp"),
    e2b_toolkit,
    ExaTools(),
]

code_agent = Agent(
    id="code-agent",
    name="Code Agent",
    role="Designing and executing complex code",
    model=OpenAILike(
        id="gpt-5",
        base_url=PROVIDER,
        api_key=CUSTOM_PROVIDER_API_KEY,
    ),
    tools=code_agent_tools,
)

# Create team with search capabilities
team = Team(
    name="Hero Team",
    model=leader_model,
    members=[perplexity_agent, code_agent],
    instructions="Delegate search and research tasks appropriately."
)

​
Troubleshooting

​
API Key Issues

import os

# Check Exa API key
if not os.getenv("EXA_API_KEY"):
    print("Warning: EXA_API_KEY not set")

# Check provider configuration
if not os.getenv("CUSTOM_PROVIDER_API_KEY"):
    print("Warning: CUSTOM_PROVIDER_API_KEY not set")

​
Model Configuration

# Verify Perplexity model setup
model = OpenAILike(
    id="sonar-pro",  # Must be exactly "sonar-pro"
    base_url=PROVIDER,  # Check this URL
    api_key=CUSTOM_PROVIDER_API_KEY  # Verify key is valid
)

​
Timezone Issues

# Verify timezone setting
agent = Agent(
    name="Search Agent",
    model=model,
    add_datetime_to_context=True,  # Must be True
    timezone_identifier="Asia/Kolkata",  # Use valid IANA timezone
)

​
See Also

E2B Sandbox

Execute code with web scraping capabilities

MCP Tools

Extend search with Firecrawl and other MCP services

Build docs developers (and LLMs) love

Get started for freeTalk to us