setup RAG (#6146)

engine/src/stirling/config/__init__.py

@@ -1,8 +1,10 @@
 """Configuration models and loaders for the Stirling AI service."""
 
-from .settings import AppSettings, load_settings
+from .settings import ENGINE_ROOT, AppSettings, RagBackend, load_settings
 
 __all__ = [
+    "ENGINE_ROOT",
     "AppSettings",
+    "RagBackend",
     "load_settings",
 ]

engine/src/stirling/config/settings.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import logging
 import logging.handlers
+from enum import StrEnum
 from functools import lru_cache
 from pathlib import Path
 
@@ -13,6 +14,11 @@ ENGINE_ROOT = Path(__file__).resolve().parents[3]
 ENV_FILE = ENGINE_ROOT / ".env"
 
 
+class RagBackend(StrEnum):
+    SQLITE = "sqlite"
+    PGVECTOR = "pgvector"
+
+
 class AppSettings(BaseSettings):
     model_config = SettingsConfigDict(env_file=ENV_FILE, extra="ignore", populate_by_name=True)
 
@@ -21,6 +27,15 @@ class AppSettings(BaseSettings):
     smart_model_max_tokens: int = Field(validation_alias="STIRLING_SMART_MODEL_MAX_TOKENS")
     fast_model_max_tokens: int = Field(validation_alias="STIRLING_FAST_MODEL_MAX_TOKENS")
 
+    # RAG settings — always on; the backend picks between embedded sqlite-vec and external pgvector.
+    rag_backend: RagBackend = Field(validation_alias="STIRLING_RAG_BACKEND")
+    rag_embedding_model: str = Field(validation_alias="STIRLING_RAG_EMBEDDING_MODEL")
+    rag_store_path: Path = Field(validation_alias="STIRLING_RAG_STORE_PATH")
+    rag_pgvector_dsn: str = Field(validation_alias="STIRLING_RAG_PGVECTOR_DSN")
+    rag_chunk_size: int = Field(validation_alias="STIRLING_RAG_CHUNK_SIZE")
+    rag_chunk_overlap: int = Field(validation_alias="STIRLING_RAG_CHUNK_OVERLAP")
+    rag_default_top_k: int = Field(validation_alias="STIRLING_RAG_TOP_K")
+
     log_level: str = Field(default="INFO", validation_alias="STIRLING_LOG_LEVEL")
     log_file: str = Field(default="", validation_alias="STIRLING_LOG_FILE")