Mahdi Bazrafshan
5fd19c12d9
Why: - Operators need Cloud and Local Embedding Models with stable ids, dimensions, and defaults. Changes: - Add Embedding Model Registry; Ollama client; env defaults for model, Ollama host, and Neighbor Expansion knobs. Impact: - New installs default to text-embedding-3-large; OLLAMA_BASE_URL required for Local provider. Co-authored-by: Cursor <cursoragent@cursor.com>
49 lines
1.3 KiB
Python
49 lines
1.3 KiB
Python
"""Application configuration loaded from .env via pydantic-settings."""
|
|
|
|
from pydantic_settings import BaseSettings, SettingsConfigDict
|
|
|
|
|
|
class Settings(BaseSettings):
|
|
"""All environment variables with defaults and validation."""
|
|
|
|
model_config = SettingsConfigDict(env_file=".env", env_file_encoding="utf-8")
|
|
|
|
# OpenAI
|
|
openai_api_key: str
|
|
embedding_model: str = "text-embedding-3-large"
|
|
llm_model: str = "gpt-4o-mini"
|
|
|
|
# Local embeddings (Ollama) — Admin switches models; host stays in config
|
|
ollama_base_url: str = "http://192.168.10.10:11435"
|
|
|
|
# Qdrant
|
|
qdrant_url: str = "http://localhost:6333"
|
|
qdrant_api_key: str | None = None
|
|
|
|
# Retrieval
|
|
top_k: int = 5
|
|
# Neighbor Expansion for fixed_size (ADR-0023); 0/0 = off
|
|
neighbor_prev: int = 0
|
|
neighbor_next: int = 0
|
|
|
|
# LLM generation
|
|
temperature: float = 0.0
|
|
max_tokens: int = 1024
|
|
|
|
# Chunking defaults
|
|
chunk_size: int = 512
|
|
chunk_overlap: int = 50
|
|
|
|
# Semantic chunking
|
|
semantic_threshold: float = 0.3
|
|
semantic_min_chunk_size: int = 3
|
|
|
|
# Database
|
|
database_url: str = "sqlite:///./data/chunking_benchmark.db"
|
|
|
|
# Text PDF gate (reject Scanned PDFs with near-empty text layer)
|
|
pdf_min_total_chars: int = 100
|
|
pdf_min_median_chars_per_page: int = 40
|
|
|
|
|
|
settings = Settings() |