Chat

The Chat module provides functionality for managing chat-based interactions with knowledge bases. It handles chat thread creation, message history, and generating responses using knowledge base search.

Public Methods

create_new_chat_thread

def create_new_chat_thread(chat_thread_params: ChatThreadParams, chat_thread_db: ChatThreadDB) -> str

Create a new chat thread in the database.

Arguments: - chat_thread_params: Parameters for the chat thread. Example:

{
    # Knowledge base IDs to use
    "kb_ids": ["kb1", "kb2"],

    # LLM model to use
    "model": "gpt-4o-mini",

    # Temperature for LLM sampling
    "temperature": 0.2,

    # System message for LLM
    "system_message": "You are a helpful assistant",

    # Model for auto-query generation
    "auto_query_model": "gpt-4o-mini",

    # Guidance for auto-query generation
    "auto_query_guidance": "",

    # Target response length (short/medium/long)
    "target_output_length": "medium",

    # Maximum tokens in chat history
    "max_chat_history_tokens": 8000,

    # Optional supplementary ID
    "supp_id": ""
}

- chat_thread_db: Database instance for storing chat threads.

Returns: Unique identifier (str) for the created chat thread.

get_chat_thread_response

def get_chat_thread_response(thread_id: str, get_response_input: ChatResponseInput, 
                           chat_thread_db: ChatThreadDB, knowledge_bases: dict)

Get a response for a chat thread using knowledge base search.

Arguments: - thread_id: Unique identifier for the chat thread. - get_response_input: Input parameters containing: - user_input: User's message text - chat_thread_params: Optional parameter overrides - metadata_filter: Optional search filter - chat_thread_db: Database instance for chat threads. - knowledge_bases: Dictionary mapping knowledge base IDs to instances.

Returns: Formatted interaction containing: - user_input: User message with content and timestamp - model_response: Model response with content and timestamp - search_queries: Generated search queries - relevant_segments: Retrieved relevant segments with file names and types - message: Error message if something went wrong (optional)

Chat Types

ChatThreadParams

Type definition for chat thread parameters.

ChatThreadParams = TypedDict('ChatThreadParams', {
    'kb_ids': List[str],              # List of knowledge base IDs to use
    'model': str,                     # LLM model name (e.g., "gpt-4o-mini")
    'temperature': float,             # Temperature for LLM sampling (0.0-1.0)
    'system_message': str,            # Custom system message for LLM
    'auto_query_model': str,          # Model for generating search queries
    'auto_query_guidance': str,       # Custom guidance for query generation
    'target_output_length': str,      # Response length ("short", "medium", "long")
    'max_chat_history_tokens': int,   # Maximum tokens in chat history
    'thread_id': str,                 # Unique thread identifier (auto-generated)
    'supp_id': str,                   # Optional supplementary identifier
})

ChatResponseInput

Input parameters for getting a chat response.

ChatResponseInput = TypedDict('ChatResponseInput', {
    'user_input': str,                          # User's message text
    'chat_thread_params': Optional[ChatThreadParams],  # Optional parameter overrides
    'metadata_filter': Optional[MetadataFilter], # Optional search filter
})

MetadataFilter

Filter criteria for knowledge base searches.

MetadataFilter = TypedDict('MetadataFilter', {
    'field_name': str,       # Name of the metadata field to filter on
    'field_value': Any,      # Value to match against
    'comparison_type': str,  # Type of comparison ("equals", "contains", etc.)
}, total=False)

You can use these types to ensure type safety when working with the chat functions. For example:

# Create chat thread parameters
params: ChatThreadParams = {
    "kb_ids": ["kb1"],
    "model": "gpt-4o-mini",
    "temperature": 0.2,
    "system_message": "You are a helpful assistant",
    "auto_query_model": "gpt-4o-mini",
    "auto_query_guidance": "",
    "target_output_length": "medium",
    "max_chat_history_tokens": 8000,
    "supp_id": ""
}

# Create chat response input
response_input: ChatResponseInput = {
    "user_input": "What is the capital of France?",
    "chat_thread_params": None,
    "metadata_filter": None
}