nexrag 0.1.0

Framework-agnostic RAG pipeline SDK. Plug in any component, swap any stage, configure everything in YAML

NexRAG

Framework-agnostic RAG pipeline SDK. Plug in any component, swap any stage, configure everything in YAML.

---

What is NexRAG?

NexRAG is a production-grade RAG (Retrieval-Augmented Generation) pipeline SDK for Python.

NexRAG owns the pipeline shape. You own the components.

Every stage — loading, chunking, embedding, retrieval, generation — is a clean interface. NexRAG ships default implementations for each. You can swap any of them by implementing the interface and declaring it in YAML. No framework lock-in. No magic. No hidden behavior.

---

Quickstart

Note: NexRAG v1.0 is under active development. This section will be updated on first release.

from nexrag import NexRAG

pipeline = NexRAG.from_config("nexrag.yaml")

# Ingest documents
pipeline.ingest("docs/contracts/")

# Query
result = pipeline.query("What are the termination clauses?")
print(result.answer)
print(result.source_chunks)

# nexrag.yaml
nexrag:
  version: "1.0"

ingestion:
  chunker:
    strategy: recursive
    chunk_size: 512
  embedder:
    provider: openai
    model: text-embedding-3-small
  vector_db:
    provider: chroma
    default_collection: contracts

query:
  llm:
    provider: openai
    model: gpt-4o

---

Installation

# Core only
pip install nexrag

# With OpenAI support
pip install "nexrag[openai]"

# With everything
pip install "nexrag[all]"

---

Design Principles

Principle	What it means
Interface-first	Every stage is a contract. Implementation is secondary.
Config-driven	YAML configures the pipeline. Code defines the logic.
Zero lock-in	Core has no dependency on LangChain, LlamaIndex, or any AI SDK.
Explicit over implicit	No hidden defaults. Every behavior is declared or documented.
Extensible by design	New components plug in without touching core.

---

Architecture

NexRAG has two independent pipelines:

INGESTION  →  Loader → Sanitizer → Chunker → Embedder → VectorDB
QUERY      →  Embedder → Retriever → PromptBuilder → LLM → PipelineResult

See Architecture Documentation for full pipeline diagrams.

---

Supported Providers (V1)

Category	Providers
Embedders	OpenAI, HuggingFace, Ollama
Vector DBs	ChromaDB (local + remote)
LLMs	OpenAI, Anthropic, Ollama
Loaders	PDF, TXT/MD, Word, Excel, JSON, HTML, Code

---

Contributing

NexRAG is in early development. Contribution guidelines will be published with v1.0.

---

Changelog

See CHANGELOG.md.