RAG on managed Apache Solr — one pip install
Opensolr is now a native LangChain vector store. Server-side GPU embeddings, hybrid BM25 + kNN search, and a managed Solr 9 index behind every retriever — with zero embedding models to configure.
The whole tutorial
This is not a teaser — this is the entire integration. No embedding model, no API key juggling, no schema design.
from langchain_opensolr import OpensolrVectorStore vs = OpensolrVectorStore( index="mysite__dense", # vector-enabled Opensolr index email="you@example.com", api_key="YOUR_OPENSOLR_API_KEY", create_if_missing=True, # provisions the index on first use ) vs.add_texts(["Hybrid search fuses BM25 with vector similarity", "Cats sleep sixteen hours a day"]) docs = vs.similarity_search("how do lexical and semantic search combine?")
No embedding model was configured, because embedding happens server-side — on Opensolr's GPU infrastructure, at both index and query time.
What happens under the hood
Your app talks to one package. The package talks to a managed pipeline that already exists.
Why this is different
Most vector stores in the LangChain directory make you bring your own embedding model and run your own database. This one doesn't.
Zero embedding config
The shortest constructor in the directory: api_key + index. Texts and queries are embedded server-side on GPU — no OpenAI key, no local model, no sentence-transformers install.
True hybrid search
Pure vector search fails on exact identifiers; pure BM25 fails on meaning. hybrid=True fuses both scores per document, with four modes and a tunable alpha balance.
Lossless metadata + filters
Metadata round-trips exactly as you stored it, and filters work both ways: filter={"category": "docs"} or any raw Solr fq expression when you need real power.
Auto index provisioning
create_if_missing=True provisions a vector-enabled Solr 9 index on first use — pick us, de or fi. No servers, no schema files, no ZooKeeper.
Drops into any chain
vs.as_retriever() and it's a standard LangChain retriever — RAG tutorials, agents, LangGraph, LCEL pipelines. Everything that accepts a retriever accepts this.
Plain Apache Solr underneath
Every index is also a real Solr core with the native /select API — facets, highlighting, spellcheck, stats. When you outgrow the vector-store interface, nothing is locked away.
Two lines you'll actually use
Hybrid retrieval with metadata filters, and the retriever that plugs into every RAG example ever written.
Hybrid search, filtered
docs = vs.similarity_search(
"affordable restaurants",
k=5,
hybrid=True, # BM25 + kNN, fused
mode="union", # or keywords_required,
# meaning_required,
# intersection
alpha=0.5, # semantic ↔ lexical
filter={"city": "Cluj"},
)As a retriever, in any chain
retriever = vs.as_retriever(
search_kwargs={"k": 5, "hybrid": True}
)
# now use it anywhere LangChain
# expects a retriever:
chain = (
{"context": retriever, "question": ...}
| prompt | llm
)Build your first RAG pipeline tonight
Free 15-day trial, no credit card. The included AI quota comfortably covers the whole tutorial — index, embed, search, and retrieve.
Germany · Solr 9.0
Finland · Solr 9.6
USA · Solr 9.6
Need a vector-enabled environment in another region? We deploy them on request — dedicated, in the region you choose (paid add-on).