> ## Documentation Index
> Fetch the complete documentation index at: https://zarna.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Backend Tech Stack

> Complete breakdown of backend technologies and services

## Core Technologies

### FastAPI

Modern, high-performance Python web framework.

**Why FastAPI**:

* Automatic OpenAPI documentation
* Built-in validation with Pydantic
* Async/await support
* High performance (comparable to Node.js)
* Type hints throughout

### Python 3.8+

Language runtime with extensive library ecosystem.

### Uvicorn

ASGI server for production deployment.

```python theme={null}
import uvicorn

uvicorn.run(
    "app.main:app",
    host="0.0.0.0",
    port=8000,
    reload=True  # Development only
)
```

## AI & ML Libraries

<AccordionGroup>
  <Accordion title="Anthropic SDK">
    **Purpose**: Claude AI integration

    **Models Used**:

    * claude-3-opus (highest quality)
    * claude-3-sonnet (balanced)
    * claude-3-haiku (fast)

    **Use Cases**: Report generation, document analysis, CRM queries
  </Accordion>

  <Accordion title="OpenAI SDK">
    **Purpose**: Alternative AI provider

    **Models**: GPT-4, GPT-3.5-turbo

    **Use Cases**: Embeddings, fallback AI
  </Accordion>

  <Accordion title="AutoGen">
    **Purpose**: Multi-agent orchestration

    **Features**: Agent teams, tool use, conversation management
  </Accordion>

  <Accordion title="Transformers">
    **Purpose**: Local model inference

    **Use Cases**: Embeddings, classification
  </Accordion>
</AccordionGroup>

## Document Processing

<AccordionGroup>
  <Accordion title="Docling">
    **Purpose**: Primary document extraction

    **Supports**: PDF, DOCX, PPTX, HTML
    **Features**: Layout analysis, table extraction, image extraction
  </Accordion>

  <Accordion title="PDFPlumber">
    **Purpose**: PDF text and table extraction

    **Advantages**: Better table detection than PyPDF2
  </Accordion>

  <Accordion title="EasyOCR">
    **Purpose**: Optical Character Recognition

    **Languages**: 80+
    **Features**: GPU support, high accuracy
  </Accordion>
</AccordionGroup>

## Database & Storage

<AccordionGroup>
  <Accordion title="Supabase Client">
    **Purpose**: PostgreSQL database access

    **Features**:

    * Database queries
    * Auth management
    * File storage
    * Real-time subscriptions
  </Accordion>

  <Accordion title="SQLAlchemy (Optional)">
    **Purpose**: ORM for complex queries

    **Use Cases**: Advanced joins, aggregations
  </Accordion>
</AccordionGroup>

## External Integrations

<AccordionGroup>
  <Accordion title="Composio SDK">
    **Purpose**: Unified OAuth and API access

    **Integrations**: Gmail, Outlook, Drive, Calendar, 100+ more
  </Accordion>

  <Accordion title="Exa SDK">
    **Purpose**: AI-powered web search

    **Use Case**: Company sourcing and research
  </Accordion>
</AccordionGroup>

## Data Processing

<AccordionGroup>
  <Accordion title="Pandas">
    **Purpose**: Data manipulation and analysis

    **Use Cases**: Excel processing, data transformation
  </Accordion>

  <Accordion title="NumPy">
    **Purpose**: Numerical computations

    **Use Cases**: Financial calculations, analytics
  </Accordion>
</AccordionGroup>

## Development Tools

* **Black**: Code formatting
* **pytest**: Testing framework
* **mypy**: Type checking
* **pylint**: Code linting

## Dependencies

**Total**: 145+ Python packages in `requirements.txt`

**Key Packages**:

```txt theme={null}
fastapi==0.109.0
uvicorn[standard]==0.27.0
pydantic==2.5.0
python-jose[cryptography]==3.3.0
python-multipart==0.0.6
supabase==2.3.0
anthropic==0.8.0
openai==1.6.0
autogen==0.2.0
docling==1.0.0
easyocr==1.7.0
pdfplumber==0.10.0
pandas==2.1.0
numpy==1.24.0
```

## Performance

* **Requests/sec**: 1,000+ on single instance
* **Async I/O**: Non-blocking operations
* **Connection pooling**: Reuse database connections
* **Agent pools**: Eliminate 5-11s cold start

## Next Steps

<CardGroup cols={2}>
  <Card title="Backend Overview" icon="server" href="/backend/overview">
    Architecture overview
  </Card>

  <Card title="Services" icon="cog" href="/backend/services/crm-agent">
    Core services
  </Card>

  <Card title="API Reference" icon="code" href="/api-reference/introduction">
    API endpoints
  </Card>
</CardGroup>