> ## Documentation Index
> Fetch the complete documentation index at: https://company-research-agent.pratikdani.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Architecture Overview

> Understanding the Company Research Agent architecture

# Architecture Overview

The Company Research Agent is built using CrewAI, a powerful framework for orchestrating multiple AI agents to work together on complex tasks. The architecture combines specialized agents, custom tools, and the Apify platform to deliver comprehensive company research.

## High-Level Architecture

<CardGroup cols={2}>
  <Card title="CrewAI Framework" icon="users">
    Coordinates multiple AI agents working together to research and analyze company data
  </Card>

  <Card title="Apify Integration" icon="robot">
    Provides web scraping and data collection capabilities through specialized actors
  </Card>

  <Card title="Custom Tools" icon="wrench">
    Purpose-built tools for interacting with various data sources
  </Card>

  <Card title="LLM Integration" icon="brain">
    Uses Google's Gemini model for analysis and report generation
  </Card>
</CardGroup>

## Core Components

### 1. CrewAI Crew

The `CompanyResearchCrew` class orchestrates three specialized agents:

```python theme={null}
@CrewBase
class CompanyResearchCrew:
    def crew(self) -> Crew:
        return Crew(
            agents=[
                self.researcher(),
                self.data_analyst(),
            ],
            tasks=[
                self.research_company(),
                self.analyze_data(),
            ],
            process=Process.sequential
        )
```

### 2. Specialized Agents

Each agent has a specific role in the research process:

* **Researcher**: Gathers raw company information
* **Data Analyst**: Processes and analyzes the collected data
* **Content Compiler**: Formats the findings into structured reports

### 3. Custom Tools

The system includes specialized tools for different data sources:

```python theme={null}
tools=[
    CompanyNewsSearchTool(actor=self.actor),
    ProfessionalProfilesTool(actor=self.actor),
    LinkedInScraperTool(actor=self.actor),
    CrunchbaseScraperTool(actor=self.actor),
    PitchBookScraperTool(actor=self.actor),
    GoogleSearchTool(actor=self.actor)
]
```

### 4. Task Flow

The research process follows a sequential workflow:

1. Company research using domain name
2. Data analysis and insight extraction
3. Report compilation and formatting

## Integration Points

### Apify Platform

* **Actors**: Uses specialized Apify actors for web scraping
* **Dataset**: Stores and manages research results
* **Key-Value Store**: Handles intermediate data storage

### LLM Integration

```python theme={null}
self.llm = LLM(
    model="gemini/gemini-2.0-flash-lite",
    temperature=0.7,
    api_key=os.getenv("GOOGLE_API_KEY")
)
```

## Data Flow

1. **Input**: Company domain name
2. **Research Phase**: Multiple tools gather data in parallel
3. **Analysis Phase**: Data is processed and analyzed
4. **Output**: Structured report in JSON format

## Error Handling

The system includes robust error handling:

* Tool-level error handling for each data source
* Agent-level task retry mechanisms
* Crew-level process management

## Performance Optimization

* Concurrent data collection where possible
* Rate limiting controls
* Memory management for large datasets

<Note>
  The architecture is designed to be modular and extensible. New agents, tools, or data sources can be added without major modifications to the core system.
</Note>