RAG

Build retrieval-augmented generation pipelines — ingest documents, split them, embed them, and retrieve them to ground LLM responses.

📄️ RAG Pipeline — SynapseKit Python LLM Framework

Build RAG pipelines in Python with SynapseKit. Async-native, hybrid BM25+vector search, 66 loaders, 22 vector stores. LangChain RAG alternative.

📄️ Document Loaders — SynapseKit RAG Python

66 document loaders for RAG in Python: PDF, Word, CSV, YouTube, S3, Notion, HubSpot, and more. Drop-in LangChain document loader alternative.

📄️ Text Splitters

Text splitters break documents into chunks for embedding and retrieval. SynapseKit provides eight splitters — all extend BaseSplitter and share the same split(text) → list[str] interface.

📄️ Retriever — Hybrid Search | SynapseKit RAG Python

Hybrid BM25 + vector search retriever in Python. MMR diversity, metadata filtering, and async retrieval for production RAG pipelines with SynapseKit.

📄️ Vector Stores — SynapseKit Python RAG Framework

22 vector store backends for Python RAG: Chroma, Pinecone, Weaviate, Qdrant, Redis, FAISS, and more. Unified async interface, MMR support.

📄️ Output Parsers

Parsers extract structured data from raw LLM text output.

📄️ Prompt Templates

Prompt templates help you build reusable, parameterised prompts and message lists.

📄️ PromptHub

PromptHub is a local filesystem prompt registry that lets you version, push, pull, and share prompt templates across your team — without any external service.

📄️ Knowledge Graph Retrieval

Multi-hop retrieval over an entity graph built from your documents. Entities and relationships are extracted using an LLM, stored in a graph backend, and then traversed at query time to surface non-obvious connections.

📄️ Property Graph RAG — SynapseKit Python RAG Framework

Vector search fused with property-graph traversal in Python. LLM + heuristic entity/relationship extraction, NetworkX and Neo4j backends, graph-backed agent memory.

📄️ WorldModelRAG — Temporal & Causal Graph RAG in Python

Temporal knowledge-graph RAG in Python: entities, relations, events and causal links with validity windows. Hybrid graph + vector retrieval, Mermaid export, optional Kuzu backend.

📄️ Open Knowledge Format (OKF)

Open Knowledge Format (OKF v0.1) is Google's vendor-neutral, low-tech format for the curated knowledge agents need: a directory tree of Markdown files, one per concept (a dataset, table, metric, ...). Each file is YAML frontmatter (type required; title, description, resource, tags, timestamp optional) plus a Markdown body, and concepts cross-link via ordinary relative Markdown links — so a bundle is effectively a knowledge graph on disk. SynapseKit reads OKF bundles with a loader, and — because the cross-links are already explicit and canonical — maps them straight onto a WorldModel graph with no lossy LLM/heuristic extraction.