Skip to content

CONFIGURATION.md

Latest commit

 

History

History
128 lines (103 loc) · 3.88 KB

CONFIGURATION.md

File metadata and controls

128 lines (103 loc) · 3.88 KB
Raw

Configuration

Configure rag-python via environment variables, .env files, or Python RAGConfig.

Environment variables

Copy .env.example to .env and set your keys.

API keys & providers

Variable Required when Description
OPENAI_API_KEY openai provider Default LLM and embeddings
ANTHROPIC_API_KEY anthropic LLM Claude models
GEMINI_API_KEY gemini LLM Gemini models
AZURE_OPENAI_ENDPOINT Azure Resource endpoint URL
AZURE_OPENAI_API_KEY Azure API key
OPENAI_API_VERSION Azure Default 2023-09-01-preview
OLLAMA_BASE_URL Ollama Default http://localhost:11434
LOCAL_EMBEDDING_MODEL local embeddings Default all-MiniLM-L6-v2

Paths

Variable Default Description
RAG_PYTHON_DATA_DIR ./data Default ingest directory
RAG_PYTHON_CHROMA_DIR ./chroma_db ChromaDB persistence path

Pipeline tuning

Variable Default Description
LLM_MODEL gpt-4o-mini Default chat model
EMBEDDING_MODEL text-embedding-3-small Default embedding model
CHUNK_STRATEGY recursive recursive, structure_aware, semantic
CHUNK_SIZE 512 Characters per chunk
CHUNK_OVERLAP 64 Overlap between chunks
TOP_K_RETRIEVE 20 Chunks retrieved before rerank
TOP_K_RERANK 5 Chunks kept after rerank
MULTI_QUERY_N 3 Rewritten queries for multi-query
GUARDRAILS_ENABLED true Prompt injection + hallucination checks
MAX_RETRIES 2 Self-correction retries
RERANKER_MODEL BAAI/bge-reranker-base Cross-encoder model

Python configuration

RAGConfig

from rag_python import RAG, RAGConfig, ChunkingConfig, SearchConfig, DocumentConfig, QueryConfig

rag = RAG(
    config=RAGConfig(
        chunking=ChunkingConfig(
            strategy="recursive",
            chunk_size=512,
            chunk_overlap=64,
        ),
        search=SearchConfig(
            retriever="hybrid",
            top_k_retrieve=20,
            top_k_rerank=5,
            multi_query_n=3,
            rerank_enabled=True,
            metadata_filter={"filename": "policy.pdf"},
        ),
        documents=DocumentConfig(
            extensions=(".txt", ".md", ".pdf", ".csv"),
            clean=True,
            copy_to_data_dir=True,
        ),
        query=QueryConfig(
            use_guardrails=True,
            use_retry=True,
            max_retries=2,
            eval_threshold=0.6,
        ),
    ),
)

Shorthand on RAG()

Pass these directly to RAG(...) without building RAGConfig:

Parameter Maps to
chunk_strategy ChunkingConfig.strategy
chunk_size ChunkingConfig.chunk_size
chunk_overlap ChunkingConfig.chunk_overlap
retriever SearchConfig.retriever
metadata_filter SearchConfig.metadata_filter
top_k_retrieve SearchConfig.top_k_retrieve
top_k_rerank SearchConfig.top_k_rerank
multi_query_n SearchConfig.multi_query_n
rerank_enabled SearchConfig.rerank_enabled
document_extensions DocumentConfig.extensions
data_dir Custom data directory
chroma_dir Custom Chroma persistence path

Per-query overrides

from rag_python import SearchConfig, QueryConfig

answer = rag.query(
    "annual leave",
    search=SearchConfig(retriever="vector", metadata_filter={"filename": "hr.txt"}),
    query_config=QueryConfig(use_retry=False),
)

Logging

import rag_python

rag_python.configure_logging()  # INFO to console
# or
import logging
logging.getLogger("rag_python").setLevel(logging.DEBUG)

Log events include ingest counts, retrieval mode, guardrail blocks, and evaluation scores.