Not quite a neural network. More than a graph.

A Context Engine
for AI

Store structured knowledge as graphs. Retrieve the right context at the right time. Domain-agnostic. Configurable. Production-ready.

Coming Soon

◈

Context Spaces

Isolated knowledge graphs for any domain. Memories, knowledge bases, profiles — each space has its own schema, ingestion rules, and retrieval config.

⬡

Graph Intelligence

Community detection, link prediction, centrality scoring, and personalized PageRank. The graph does heavy lifting without LLM calls.

⚡

Fast Hydration

Six concurrent retrieval channels. Entity matching, vector search, graph expansion — all in under 200ms. No LLM in the hot path.

⚙

Fully Configurable

Define new space types with YAML. Custom schemas, extraction prompts, scoring weights, lifecycle rules. No code changes needed.

📦

One Binary

Single Go binary. Docker Compose for the full stack. Production-grade from day one. Scales to hundreds of thousands of users.

🐍

Python SDK

pip install neuralgraph — five lines to persistent context. Typed models, async support, clean API.

How it works

from neuralgraph import NeuralGraph

ng = NeuralGraph(url="http://localhost:8080")

# Ingest a conversation
ng.ingest(
    space_id="memories",
    user_id="david",
    messages=[
        {"role": "user", "content": "My dog Max has been limping"},
    ],
)

# Retrieve relevant context
ctx = ng.hydrate(
    user_id="david",
    messages=[{"role": "user", "content": "Should I take him to the vet?"}],
)

# ctx.nodes → [Node("pets.max", score=0.94)]
# ctx.system_prompt → full prompt with context injected
            

A Context Enginefor AI