Class ZCL_LLM_CLIENT_OLLAMA

AI Generated documentation.

Overview

The ZCL_LLM_CLIENT_OLLAMA class implements a client for interacting with Ollama LLM services. It inherits from ZCL_LLM_CLIENT_BASE and provides specialized handling for Ollama-specific features.

Public methods:

• get_client: Factory method that creates and returns a new Ollama client instance
• constructor: Initializes the Ollama client with client and provider configurations

The class handles:

• Authentication via optional API keys
• Request/response formatting specific to Ollama
• Structured output processing
• Tool calls management
• Message parsing and formatting

Dependencies

• Parent class: ZCL_LLM_CLIENT_BASE
• ZCL_LLM_HTTP_CLIENT_WRAPPER: For HTTP communication
• ZCL_LLM_COMMON: Utility class for JSON handling
• BAdI implementation for encryption handling

Details

The class implements a specialized HTTP client for Ollama with focus on proper request/response handling:

flowchart TD
    A[Client Request] --> B[Build Request JSON]
    B --> C[HTTP Request]
    C --> D[Handle Response]
    D --> E{Check Status}
    E -->|Success| F[Parse Response]
    E -->|Error| G[Handle Error]
    F --> H[Process Tool Calls]
    F --> I[Handle Structured Output]
    H --> J[Return Response]
    I --> J
    G --> J

Key implementation aspects:

• Supports streaming control via stream parameter
• Handles Ollama-specific format parameters for structured output
• Implements custom options handling in Ollama format
• Provides detailed error handling with retry capabilities for rate limits and timeouts
• Processes tool calls with argument parsing and validation
• Supports encrypted authentication via custom headers