backend/app/core/constants.py

"""Application constants and enums"""

from enum import Enum


class VectorDBType(str, Enum):
    """Supported vector database types"""
    CHROMA = "chroma"
    PINECONE = "pinecone"
    WEAVIATE = "weaviate"
    QDRANT = "qdrant"


class EmbeddingProvider(str, Enum):
    """Supported embedding providers"""
    HUGGINGFACE = "huggingface"
    GROQ = "groq"


class TokenizerType(str, Enum):
    """Supported tokenizer types"""
    TIKTOKEN = "tiktoken"
    HUGGINGFACE = "huggingface"


class DocumentType(str, Enum):
    """Supported document types"""
    PDF = "pdf"
    CSV = "csv"
    TEXT = "text"


class GroqModel(str, Enum):
    """Available Groq models"""
    LLAMA_31_8B = "llama-3.1-8b-instant"
    LLAMA_33_70B = "llama-3.3-70b-versatile"
    GPT_OSS_120B = "openai/gpt-oss-120b"
    GPT_OSS_20B = "openai/gpt-oss-20b"


GROQ_MODELS = [
    {"id": "llama-3.1-8b-instant", "name": "Llama 3.1 8B (Fast)", "context_window": 8000},
    {"id": "llama-3.3-70b-versatile", "name": "Llama 3.3 70B (Versatile)", "context_window": 8000},
    {"id": "openai/gpt-oss-120b", "name": "OpenAI GPT-OSS 120B", "context_window": 4000},
    {"id": "openai/gpt-oss-20b", "name": "OpenAI GPT-OSS 20B", "context_window": 4000},
]

EMBEDDING_MODELS = [
    {"id": "all-MiniLM-L6-v2", "name": "MiniLM (Small, Fast)", "dimension": 384},
    {"id": "all-mpnet-base-v2", "name": "MPNet (Large, Accurate)", "dimension": 768},
]

VECTOR_DBS = [
    {"id": "chroma", "name": "Chroma (Local)", "type": "local"},
    {"id": "pinecone", "name": "Pinecone (Cloud)", "type": "cloud"},
    {"id": "weaviate", "name": "Weaviate (Self-hosted)", "type": "hybrid"},
    {"id": "qdrant", "name": "Qdrant (Self-hosted)", "type": "local"},
]

TOKENIZERS = [
    {"id": "tiktoken", "name": "TikToken (OpenAI Standard)"},
    {"id": "huggingface", "name": "HuggingFace (Model-specific)"},
]

# RAG Configuration Defaults
DEFAULT_CHUNK_SIZE = 512
DEFAULT_CHUNK_OVERLAP = 50
DEFAULT_RETRIEVAL_TOP_K = 5
DEFAULT_GENERATION_TEMPERATURE = 0.7

# File upload limits
MAX_FILE_SIZE = 52428800  # 50 MB
ALLOWED_EXTENSIONS = {".pdf", ".csv", ".txt"}

# RAG Pipeline Steps
RAG_STEPS = [
    "document_retrieval",
    "context_assembly",
    "prompt_construction",
    "llm_generation",
    "post_processing",
]

# RAG Modes
class RAGMode(str, Enum):
    """RAG operation modes"""
    SIMPLE = "simple"      # Direct retrieval + generation
    AGENTIC = "agentic"    # Multi-step agent with tool use
    GRAPH = "graph"        # Graph-based knowledge representation


RAG_MODES = [
    {"id": "simple", "name": "Simple RAG", "description": "Direct retrieval and generation", "icon": "📄"},
    {"id": "agentic", "name": "Agentic RAG", "description": "Multi-step reasoning with tools", "icon": "🤖"},
    {"id": "graph", "name": "Graph RAG", "description": "Knowledge graph-based retrieval", "icon": "🕸️"},
]