Configuration Guide

KnowLang uses pydantic-settings for configuration management. Settings can be provided through environment variables, .env files, or programmatically.

Quick Start

Copy the example configuration:

cp .env.example .env

Modify settings as needed in .env

Core Settings

LLM Settings

# Default is Ollama with llama3.2
LLM__MODEL_NAME=llama3.2
LLM__MODEL_PROVIDER=ollama
LLM__API_KEY=your_api_key  # Required for providers like OpenAI

Supported providers:

ollama: Local models through Ollama
openai: OpenAI models (requires API key)
anthropic: Anthropic models (requires API key)

Embedding Settings

# Default is Ollama with mxbai-embed-large
EMBEDDING__MODEL_NAME=mxbai-embed-large
EMBEDDING__MODEL_PROVIDER=ollama
EMBEDDING__API_KEY=your_api_key  # Required for providers like OpenAI

Database Settings

# ChromaDB configuration
DB__PERSIST_DIRECTORY=./chromadb/mycode
DB__COLLECTION_NAME=code
DB__CODEBASE_DIRECTORY=./

Parser Settings

# Language support and file patterns
PARSER__LANGUAGES='{"python": {"enabled": true, "file_extensions": [".py"]}}'
PARSER__PATH_PATTERNS='{"include": ["**/*"], "exclude": ["**/venv/**", "**/.git/**"]}'

Chat Interface Settings

CHAT__MAX_CONTEXT_CHUNKS=5
CHAT__SIMILARITY_THRESHOLD=0.7
CHAT__INTERFACE_TITLE='Code Repository Q&A Assistant'

Advanced Configuration

Using Multiple Models

You can configure different models for different purposes:

# Main LLM for responses
LLM__MODEL_NAME=llama3.2
LLM__MODEL_PROVIDER=ollama

# Evaluation model
EVALUATOR__MODEL_NAME=gpt-4
EVALUATOR__MODEL_PROVIDER=openai

# Embedding model
EMBEDDING__MODEL_NAME=mxbai-embed-large
EMBEDDING__MODEL_PROVIDER=ollama

Reranker Configuration

RERANKER__ENABLED=true
RERANKER__MODEL_NAME=rerank-2
RERANKER__MODEL_PROVIDER=voyage
RERANKER__TOP_K=4

Analytics Integration

CHAT_ANALYTICS__ENABLED=true
CHAT_ANALYTICS__PROVIDER=mixpanel
CHAT_ANALYTICS__API_KEY=your_api_key

Spaces:

gabykim
/

KnowLang_Transformers_Demo

Sleeping