🔍 Code Extractor

A dataclass that stores configuration settings for a chat interface integrated with a RAG (Retrieval-Augmented Generation) engine, managing search parameters, data sources, and model settings.

Purpose

ChatConfiguration serves as a centralized configuration container for chat interfaces that use RAG engines. It manages settings for document retrieval (collections, search thresholds, topK results), search behavior (web search, extensive search, memory), filtering options (keyword and reference filtering), and LLM parameters (model, temperature, tokens). The class provides backward compatibility with legacy field names and includes methods for serialization/deserialization to/from dictionaries. It's designed to match RAG engine defaults and can be persisted or transmitted between components.

Source Code

class ChatConfiguration:
    """Configuration for chat interface linked to a text section - matches RAG engine defaults"""
    # Data sources
    collections: List[str] = None
    
    # Core RAG engine settings (matching flow_control defaults)
    enableSearch: bool = True
    enableWebSearch: bool = False
    enableMemory: bool = False  # RAG default is false
    enableExtensiveSearch: bool = False
    topK: int = 5
    searchThreshold: float = 0.7
    detailLevel: str = "balanced"  # "Balanced" -> "balanced"
    maxSources: int = 10
    targetSummaryTokens: int = 8000  # matches RAG default
    extensiveChunks: int = 100  # matches RAG default extensive_search_chunks
    summaryTokens: int = 5000
    
    # Keyword and reference filtering (matching RAG defaults)
    enableKeywordFiltering: bool = False
    enableReferenceFiltering: bool = True
    referenceThreshold: float = 0.3  # matches RAG relevance_threshold
    manualKeywords: List[str] = None
    
    # Instructions
    detailedInstructions: str = ""
    instructionTemplate: str = ""
    
    # Legacy fields (keep for backward compatibility but with RAG-aligned defaults)
    system_prompt: str = ""
    model: str = "gpt-4o"  # matches RAG actual model
    temperature: float = 0.0  # matches RAG default temp
    max_tokens: int = 4000  # more reasonable default
    context_window: int = 10
    auto_save: bool = True
    reference_mode: str = "citations"
    selected_collections: List[str] = None  # alias for collections
    uploaded_documents: List[str] = None
    use_reranking: bool = False
    min_score: float = 0.5
    diversity_threshold: float = 0.8
    system_instructions: str = ""  # legacy
    detailed_instructions: str = ""  # legacy
    
    def __post_init__(self):
        if self.collections is None:
            self.collections = []
        if self.selected_collections is None:
            self.selected_collections = []
        if self.uploaded_documents is None:
            self.uploaded_documents = []
        if self.manualKeywords is None:
            self.manualKeywords = []
    
    def to_dict(self) -> Dict:
        return asdict(self)
    
    @classmethod
    def from_dict(cls, data: Dict) -> 'ChatConfiguration':
        """Create ChatConfiguration from dict with backward compatibility"""
        # Handle backward compatibility and field name mapping
        converted_data = {}
        
        # Field mappings from old names to new names
        field_mappings = {
            'enable_search': 'enableSearch',
            'enable_web_search': 'enableWebSearch', 
            'enable_memory': 'enableMemory',
            'enable_extensive_search': 'enableExtensiveSearch',
            'top_k': 'topK',
            'search_threshold': 'searchThreshold',
            'detail_level': 'detailLevel',
            'max_sources': 'maxSources',
            'target_summary_tokens': 'targetSummaryTokens',
            'extensive_chunks': 'extensiveChunks',
            'summary_tokens': 'summaryTokens',
            'enable_keyword_filtering': 'enableKeywordFiltering',
            'enable_reference_filtering': 'enableReferenceFiltering',
            'reference_threshold': 'referenceThreshold',
            'manual_keywords': 'manualKeywords',
            'detailed_instructions': 'detailedInstructions',
            'instruction_template': 'instructionTemplate'
        }
        
        # Get valid field names for ChatConfiguration
        import inspect
        valid_fields = set(inspect.signature(cls).parameters.keys())
        
        # Convert old field names to new field names
        for old_key, new_key in field_mappings.items():
            if old_key in data and new_key in valid_fields:
                converted_data[new_key] = data[old_key]
            elif new_key in data and new_key in valid_fields:
                converted_data[new_key] = data[new_key]
        
        # Handle collections field (prioritize 'collections' over 'selected_collections')
        if 'collections' in valid_fields:
            if 'collections' in data:
                converted_data['collections'] = data['collections']
            elif 'selected_collections' in data:
                converted_data['collections'] = data['selected_collections']
            
        # Copy remaining fields directly (only if they're valid)
        for key, value in data.items():
            if key not in field_mappings and key != 'selected_collections' and key in valid_fields:
                converted_data[key] = value
                
        return cls(**converted_data)

Parameters

Name	Type	Default	Kind
bases	-	-

Parameter Details

collections: List of collection names/IDs to search within. Defaults to empty list if None. Primary field for specifying data sources.

enableSearch: Boolean flag to enable/disable search functionality. Default is True.

enableWebSearch: Boolean flag to enable/disable web search capabilities. Default is False.

enableMemory: Boolean flag to enable/disable conversation memory. Default is False to match RAG defaults.

enableExtensiveSearch: Boolean flag to enable/disable extensive search mode for deeper retrieval. Default is False.

topK: Integer specifying the number of top results to retrieve from search. Default is 5.

searchThreshold: Float threshold (0.0-1.0) for minimum similarity score in search results. Default is 0.7.

detailLevel: String indicating the level of detail in responses. Default is 'balanced'. Expected values: 'balanced', 'concise', 'detailed'.

maxSources: Maximum number of sources to include in responses. Default is 10.

targetSummaryTokens: Target token count for generated summaries. Default is 8000 to match RAG defaults.

extensiveChunks: Number of chunks to retrieve in extensive search mode. Default is 100.

summaryTokens: Token budget for summary generation. Default is 5000.

enableKeywordFiltering: Boolean to enable/disable keyword-based filtering of results. Default is False.

enableReferenceFiltering: Boolean to enable/disable reference-based filtering. Default is True.

referenceThreshold: Float threshold (0.0-1.0) for reference relevance filtering. Default is 0.3.

manualKeywords: List of manually specified keywords for filtering. Defaults to empty list if None.

detailedInstructions: String containing detailed instructions for the chat behavior. Default is empty string.

instructionTemplate: String template for formatting instructions. Default is empty string.

system_prompt: Legacy field for system-level prompts. Default is empty string.

model: LLM model identifier. Default is 'gpt-4o' to match RAG defaults.

temperature: Float (0.0-2.0) controlling randomness in model responses. Default is 0.0 for deterministic output.

max_tokens: Maximum tokens in model response. Default is 4000.

context_window: Number of previous messages to include in context. Default is 10.

auto_save: Boolean to enable/disable automatic saving of chat state. Default is True.

reference_mode: String specifying how references are displayed. Default is 'citations'. Expected values: 'citations', 'inline', 'footnotes'.

selected_collections: Legacy alias for 'collections' field. Maintained for backward compatibility.

uploaded_documents: List of uploaded document identifiers. Defaults to empty list if None.

use_reranking: Boolean to enable/disable result reranking. Default is False.

min_score: Minimum score threshold for including results. Default is 0.5.

diversity_threshold: Float threshold for result diversity filtering. Default is 0.8.

system_instructions: Legacy field for system instructions. Default is empty string.

detailed_instructions: Legacy field for detailed instructions. Default is empty string.

Return Value

Instantiation returns a ChatConfiguration object with all configuration attributes initialized. The to_dict() method returns a Dict containing all configuration fields as key-value pairs. The from_dict() class method returns a new ChatConfiguration instance created from a dictionary, handling backward compatibility with legacy field names.

Class Interface

Methods

Attributes

Name	Type	Description	Scope
collections	List[str]	List of collection names/IDs to search within	instance
enableSearch	bool	Flag to enable/disable search functionality	instance
enableWebSearch	bool	Flag to enable/disable web search capabilities	instance
enableMemory	bool	Flag to enable/disable conversation memory	instance
enableExtensiveSearch	bool	Flag to enable/disable extensive search mode	instance
topK	int	Number of top results to retrieve from search	instance
searchThreshold	float	Minimum similarity score threshold for search results	instance
detailLevel	str	Level of detail in responses (balanced, concise, detailed)	instance
maxSources	int	Maximum number of sources to include in responses	instance
targetSummaryTokens	int	Target token count for generated summaries	instance
extensiveChunks	int	Number of chunks to retrieve in extensive search mode	instance
summaryTokens	int	Token budget for summary generation	instance
enableKeywordFiltering	bool	Flag to enable/disable keyword-based filtering	instance
enableReferenceFiltering	bool	Flag to enable/disable reference-based filtering	instance
referenceThreshold	float	Threshold for reference relevance filtering	instance
manualKeywords	List[str]	List of manually specified keywords for filtering	instance
detailedInstructions	str	Detailed instructions for chat behavior	instance
instructionTemplate	str	Template for formatting instructions	instance
system_prompt	str	Legacy field for system-level prompts	instance
model	str	LLM model identifier (e.g., gpt-4o)	instance
temperature	float	Controls randomness in model responses (0.0-2.0)	instance
max_tokens	int	Maximum tokens in model response	instance
context_window	int	Number of previous messages to include in context	instance
auto_save	bool	Flag to enable/disable automatic saving of chat state	instance
reference_mode	str	How references are displayed (citations, inline, footnotes)	instance
selected_collections	List[str]	Legacy alias for collections field	instance
uploaded_documents	List[str]	List of uploaded document identifiers	instance
use_reranking	bool	Flag to enable/disable result reranking	instance
min_score	float	Minimum score threshold for including results	instance
diversity_threshold	float	Threshold for result diversity filtering	instance
system_instructions	str	Legacy field for system instructions	instance
detailed_instructions	str	Legacy field for detailed instructions	instance

Dependencies

• dataclasses
• typing
• inspect

Required Imports

from dataclasses import dataclass, asdict
from typing import List, Dict
import inspect

Usage Example

# Create a new configuration with defaults
from dataclasses import dataclass, asdict
from typing import List, Dict
import inspect

config = ChatConfiguration()

# Create with custom settings
config = ChatConfiguration(
    collections=['documents', 'research_papers'],
    enableSearch=True,
    enableWebSearch=True,
    topK=10,
    searchThreshold=0.8,
    model='gpt-4o',
    temperature=0.2,
    detailedInstructions='Focus on technical accuracy'
)

# Convert to dictionary for serialization
config_dict = config.to_dict()

# Create from dictionary (with backward compatibility)
old_format_dict = {
    'enable_search': True,
    'top_k': 5,
    'selected_collections': ['docs'],
    'detail_level': 'balanced'
}
config_from_dict = ChatConfiguration.from_dict(old_format_dict)

# Access attributes
print(config.collections)
print(config.topK)
print(config.model)

Best Practices

• Always use from_dict() class method when deserializing from external sources to ensure backward compatibility with legacy field names.
• The __post_init__ method automatically initializes list fields to empty lists if None, preventing NoneType errors.
• Use 'collections' field instead of 'selected_collections' for new code, though both are supported for compatibility.
• The class is immutable after creation unless explicitly modified. Consider creating new instances for different configurations.
• Field names use camelCase for new fields (enableSearch, topK) but snake_case for legacy fields (system_prompt, max_tokens).
• When persisting configuration, use to_dict() to ensure all fields are captured correctly.
• The from_dict() method validates fields against the class signature, ignoring invalid fields to prevent errors.
• Default values are aligned with RAG engine defaults for consistency across the system.
• Temperature of 0.0 provides deterministic responses; increase for more creative outputs.
• The searchThreshold and referenceThreshold values should be tuned based on your data quality and retrieval needs.

Tags

Similar Components

AI-powered semantic similarity - components with related functionality:

• class DocChatRAG 63.6% similar

  Main RAG engine with three operating modes: 1. Basic RAG (similarity search) 2. Extensive (full document retrieval with preprocessing) 3. Full Reading (process all documents)

  From: /tf/active/vicechatdev/docchat/rag_engine.py

• function process_chat_background 63.3% similar

  Processes chat requests asynchronously in a background thread, managing RAG engine interactions, progress updates, and session state for various query modes including basic, extensive, full_reading, and deep_reflection.

  From: /tf/active/vicechatdev/docchat/app.py

• function init_chat_engine_v1 62.7% similar

  Initializes a global chat engine instance using OneCo_hybrid_RAG and validates its configuration by checking for required attributes like available_collections and data_handles.

  From: /tf/active/vicechatdev/vice_ai/app.py

• class ChatSession_v1 62.1% similar

  A dataclass representing a chat session associated with a specific text section in a document, managing conversation messages, context, and references.

  From: /tf/active/vicechatdev/vice_ai/models.py

• function basic_rag_example 62.0% similar

  Demonstrates a basic RAG (Retrieval-Augmented Generation) workflow by initializing a DocChatRAG engine, executing a sample query about document topics, and displaying the response with metadata.

  From: /tf/active/vicechatdev/docchat/example_usage.py

class ChatConfiguration:
    """Configuration for chat interface linked to a text section - matches RAG engine defaults"""
    # Data sources
    collections: List[str] = None
    
    # Core RAG engine settings (matching flow_control defaults)
    enableSearch: bool = True
    enableWebSearch: bool = False
    enableMemory: bool = False  # RAG default is false
    enableExtensiveSearch: bool = False
    topK: int = 5
    searchThreshold: float = 0.7
    detailLevel: str = "balanced"  # "Balanced" -> "balanced"
    maxSources: int = 10
    targetSummaryTokens: int = 8000  # matches RAG default
    extensiveChunks: int = 100  # matches RAG default extensive_search_chunks
    summaryTokens: int = 5000
    
    # Keyword and reference filtering (matching RAG defaults)
    enableKeywordFiltering: bool = False
    enableReferenceFiltering: bool = True
    referenceThreshold: float = 0.3  # matches RAG relevance_threshold
    manualKeywords: List[str] = None
    
    # Instructions
    detailedInstructions: str = ""
    instructionTemplate: str = ""
    
    # Legacy fields (keep for backward compatibility but with RAG-aligned defaults)
    system_prompt: str = ""
    model: str = "gpt-4o"  # matches RAG actual model
    temperature: float = 0.0  # matches RAG default temp
    max_tokens: int = 4000  # more reasonable default
    context_window: int = 10
    auto_save: bool = True
    reference_mode: str = "citations"
    selected_collections: List[str] = None  # alias for collections
    uploaded_documents: List[str] = None
    use_reranking: bool = False
    min_score: float = 0.5
    diversity_threshold: float = 0.8
    system_instructions: str = ""  # legacy
    detailed_instructions: str = ""  # legacy
    
    def __post_init__(self):
        if self.collections is None:
            self.collections = []
        if self.selected_collections is None:
            self.selected_collections = []
        if self.uploaded_documents is None:
            self.uploaded_documents = []
        if self.manualKeywords is None:
            self.manualKeywords = []
    
    def to_dict(self) -> Dict:
        return asdict(self)
    
    @classmethod
    def from_dict(cls, data: Dict) -> 'ChatConfiguration':
        """Create ChatConfiguration from dict with backward compatibility"""
        # Handle backward compatibility and field name mapping
        converted_data = {}
        
        # Field mappings from old names to new names
        field_mappings = {
            'enable_search': 'enableSearch',
            'enable_web_search': 'enableWebSearch', 
            'enable_memory': 'enableMemory',
            'enable_extensive_search': 'enableExtensiveSearch',
            'top_k': 'topK',
            'search_threshold': 'searchThreshold',
            'detail_level': 'detailLevel',
            'max_sources': 'maxSources',
            'target_summary_tokens': 'targetSummaryTokens',
            'extensive_chunks': 'extensiveChunks',
            'summary_tokens': 'summaryTokens',
            'enable_keyword_filtering': 'enableKeywordFiltering',
            'enable_reference_filtering': 'enableReferenceFiltering',
            'reference_threshold': 'referenceThreshold',
            'manual_keywords': 'manualKeywords',
            'detailed_instructions': 'detailedInstructions',
            'instruction_template': 'instructionTemplate'
        }
        
        # Get valid field names for ChatConfiguration
        import inspect
        valid_fields = set(inspect.signature(cls).parameters.keys())
        
        # Convert old field names to new field names
        for old_key, new_key in field_mappings.items():
            if old_key in data and new_key in valid_fields:
                converted_data[new_key] = data[old_key]
            elif new_key in data and new_key in valid_fields:
                converted_data[new_key] = data[new_key]
        
        # Handle collections field (prioritize 'collections' over 'selected_collections')
        if 'collections' in valid_fields:
            if 'collections' in data:
                converted_data['collections'] = data['collections']
            elif 'selected_collections' in data:
                converted_data['collections'] = data['selected_collections']
            
        # Copy remaining fields directly (only if they're valid)
        for key, value in data.items():
            if key not in field_mappings and key != 'selected_collections' and key in valid_fields:
                converted_data[key] = value
                
        return cls(**converted_data)