LangChain integration for qortex graph-enhanced retrieval.
pip install langchain-qortexThis package provides QortexVectorStore, a LangChain VectorStore backed by qortex's knowledge graph. It works anywhere LangChain expects a VectorStore: chains, agents, as_retriever(), similarity_search(), add_documents(), etc.
qortex augments standard vector search with graph structure, rules, and a feedback-driven learning loop. This package gives LangChain users access to those capabilities through the API they already know.
from langchain_qortex import QortexVectorStore
# Zero-config (like Chroma.from_texts)
vs = QortexVectorStore.from_texts(
texts=["OAuth2 is an authorization framework", "JWT carries signed claims"],
embedding=my_embedding,
domain="security",
)
# Standard VectorStore API
docs = vs.similarity_search("authentication", k=5)
retriever = vs.as_retriever()Beyond standard vector search, QortexVectorStore exposes three capabilities that flat vector stores cannot provide:
# Explore: traverse typed edges from any search result
explore = vs.explore(docs[0].metadata["node_id"])
for edge in explore.edges:
print(f"{edge.source_id} --{edge.relation_type}--> {edge.target_id}")
# Rules: get projected rules linked to retrieved concepts
rules = vs.rules(concept_ids=[d.metadata["node_id"] for d in docs])
for rule in rules.rules:
print(f"[{rule.category}] {rule.text}")
# Feedback: close the learning loop
vs.feedback({docs[0].id: "accepted", docs[-1].id: "rejected"})
# Future queries adjust PPR teleportation weights based on this signal- Full VectorStore guide (graph exploration, rules, feedback loop)
- qortex docs
- Querying guide
MIT