fennec-community 0.1.0

Production-grade RAG framework with Arabic NLP support — chunking, embeddings, LLM interfaces, vector databases, routing, and more.

fennec-community

Production-grade RAG framework with Arabic NLP support

fennec-community is a modular Python library for building Retrieval-Augmented Generation (RAG) pipelines. It ships with first-class support for Arabic text and integrates with all major LLM providers, embedding models, and vector databases.

---

Features

• Smart Chunking — Semantic, adaptive, structure-aware, context-aware, and Arabic-specific chunkers
• Embeddings — OpenAI, Gemini, Mistral, Ollama, HuggingFace, and a dedicated Arabic embedder
• LLM Interfaces — Unified API over OpenAI, Anthropic, Gemini, Mistral, Groq, and Ollama
• Vector Databases — FAISS, ChromaDB, and Pinecone
• RAG Variants — Standard, Agentic, Conversational, Graph, Hybrid, Multi-Doc, Multi-Hop, Streaming, Domain-Specific, Self-Improving
• Prompt Engine — Dynamic context-aware prompt building with strategy support
• Router — Hierarchical semantic routing with caching, feedback, and observability
• Plugin System — Extensible plugin architecture with security controls
• Output Parsers — JSON, YAML, CSV, Pydantic, structured output with auto-fix and retry
• Document Loaders — PDF, DOCX, HTML, CSV, JSON, plain text, web, and directory loaders

---

Installation

pip install fennec-community

Install with optional extras for your provider of choice:

pip install fennec-community[openai]       # OpenAI LLM + embeddings
pip install fennec-community[anthropic]    # Anthropic Claude
pip install fennec-community[gemini]       # Google Gemini
pip install fennec-community[ollama]       # Local Ollama models
pip install fennec-community[faiss]        # FAISS vector store
pip install fennec-community[chroma]       # ChromaDB vector store
pip install fennec-community[pinecone]     # Pinecone vector store
pip install fennec-community[arabic]       # Arabic NLP tools
pip install fennec-community[all]          # Everything

---

Quick Start

Basic RAG Pipeline

from fennec_community.rag.core import RAGSystem, RAGConfig
from fennec_community.embeddings import OpenAIEmbedder
from fennec_community.vector_database import FAISSVectorDatabase

config = RAGConfig(top_k=5)
embedder = OpenAIEmbedder(api_key="sk-...")
db = FAISSVectorDatabase(embedder=embedder)

rag = RAGSystem(config=config, vector_db=db)
rag.load_documents(["path/to/docs/"])

answer = rag.query("What is the capital of France?")
print(answer)

Arabic Chunking

from fennec_community.chunks import ArabicTextChunker, ChunkConfig

chunker = ArabicTextChunker(config=ChunkConfig(chunk_size=512))
chunks = chunker.chunk("النص العربي هنا...")

for chunk in chunks:
    print(chunk.text)

Conversational RAG

from fennec_community.rag.types.conversational_rag import ConversationalRAG

crag = ConversationalRAG(rag_system=rag)
response = crag.chat("Tell me about the report.")
response2 = crag.chat("Can you summarize the key points?")

Semantic Router

from fennec_community.router import HierarchicalRouter, Route, RouteGroup

router = HierarchicalRouter()
router.add_group(
    RouteGroup(
        name="support",
        routes=[
            Route(name="billing", keywords=["invoice", "payment", "refund"]),
            Route(name="technical", keywords=["error", "bug", "crash"]),
        ]
    )
)

result = router.route("I need help with my invoice")
print(result.matched_route)  # "billing"

LLM Interface

from fennec_community.llm import OpenAIInterface

llm = OpenAIInterface(api_key="sk-...")
response = llm.chat("Explain RAG in one sentence.")
print(response)

Output Parsing

from fennec_community.output_parser import OutputParser

parser = OutputParser()
result = parser.parse('{"name": "Alice", "age": 30}', format="json")
print(result)  # {'name': 'Alice', 'age': 30}

---

Modules

Module	Description
fennec_community.chunks	Text chunking strategies including Arabic support
fennec_community.embeddings	Embedding models for multiple providers
fennec_community.llm	LLM provider interfaces
fennec_community.vector_database	Vector store backends (FAISS, Chroma, Pinecone)
fennec_community.rag	Full RAG system with multiple retrieval strategies
fennec_community.prompt	Prompt engineering and context management
fennec_community.router	Semantic routing with hierarchical matching
fennec_community.context	Context window management and retrieval
fennec_community.output_parser	Structured output parsing and validation
fennec_community.document_loaders	Document ingestion from various sources
fennec_community.plugins	Plugin system for extending functionality
fennec_community.chain	Sequential, parallel, and conditional chains

---

RAG Variants

Variant	Class	Use Case
Standard RAG	RAGSystem	General question answering
Agentic RAG	AgenticRAG	Multi-step reasoning with tool use
Conversational RAG	ConversationalRAG	Multi-turn chat with memory
Graph RAG	GraphRAG	Knowledge graph-based retrieval
Hybrid Search RAG	HybridSearchRAG	Dense + sparse retrieval
Multi-Doc RAG	MultiDocRAG	Cross-document reasoning
Multi-Hop RAG	MultiHopRAG	Complex multi-step queries
Streaming RAG	StreamingRAG	Token-by-token streaming responses
Domain RAG	DomainSpecificRAG	Domain-constrained retrieval
Self-Improving RAG	SelfImprovingRAG	Iterative refinement with HyDE
Federated RAG	FederatedRAG	Multi-source distributed retrieval

---

Requirements

• Python >= 3.9
• pydantic >= 2.0
• numpy >= 1.24
• tiktoken >= 0.5

All other dependencies are optional and installed via extras.

---

License

MIT License — see LICENSE for details.

---

Contributing

Contributions are welcome! Please open an issue or pull request on GitHub.