llama-index-hypergraph

LlamaIndex retriever and chat memory backed by the Temporal Hypergraph Engine.

pip install llama-index-hypergraph
docker run -d -p 50051:50051 ghcr.io/divyansh151005/temporal-hypergraph-server

For development from a clone: pip install -e integrations/python-client -e integrations/llamaindex.

from llama_index_hypergraph import TemporalHypergraph

memory = TemporalHypergraph(api_key="your_key")  # a TemporalHypergraphChatMemory on session "default"
memory = TemporalHypergraph(api_key="your_key", session_id="user-42")

The server address comes from endpoint= or $HYPERGRAPH_ENDPOINT (default localhost:50051).

Retriever

from llama_index.core.query_engine import RetrieverQueryEngine
from llama_index_hypergraph import TemporalHypergraphRetriever

retriever = TemporalHypergraphRetriever(
    known_entities=["org:acme", "org:beta"],  # default linker: label match in the query
    k_hops=3,
    as_of=1700000000,                        # facts valid at this instant (default now)
)
engine = RetrieverQueryEngine.from_args(retriever, llm=llm)

• Seeds come from entity_linker(query) (pass your own callable for NER or fuzzy matching), plus seed_entities and session:<session_id>.
• embed_model (a LlamaIndex BaseEmbedding or LangChain-style embedder) re-ranks results by similarity; with no seeds the retriever falls back to global similarity search.
• Each TextNode carries event_type, entities, valid_from, valid_to and hyperedge_id metadata.

Chat memory

TemporalHypergraphChatMemory is a BaseMemory. get(input=...) returns the recent history (history_limit=), preceded by a system message listing relevant facts reached within k_hops of the session. It uses the same storage scheme as langchain-hypergraph, so either framework can read a session written by the other.

Tests

cargo build --release -p grpc-server   # for the live tests; they skip otherwise
pytest integrations/llamaindex