v0.1.0-alphaSeptember 2026 (Target)

Alpha Preview

Property graph Schema and triple (Subject-Predicate-Object) definitions.
Automatic natural language triple extraction from agent history streams.
Embedded local storage adapters powered by Kùzu Graph Database.
Multi-hop Graph-RAG search with neighborhood graph context retrieval.

Launching in 60 Days (Q3 2026)Open SourceKnowledge Graph Layer

SolidStateGraph

The open-source Knowledge Graph Management Layer designed specifically for agentic AI runtimes and LLM reasoning. Seamlessly bridge unstructured conversation state with structured property graphs.

Notify Me at Launch Star on GitHub

Release Roadmap

SolidStateGraph is undergoing active development to bring production-grade Knowledge Graphs to AI agents.

Phase 1 — Completed

Core Specification

Defined state machine integration contracts, property graph abstractions, and triple schema standards.

Phase 2 — Active Dev

Embedded Triple Extractor

Implementing real-time LLM entity resolution and automatic (Subject)-[Predicate]->(Object) canonicalization.

Phase 3 — Next Month

Hybrid Graph-RAG Engine

Coupling dense vector similarity embeddings with multi-hop Graph traversal engines for contextual subgraph extraction.

Phase 4 — Launch Target

Public v0.1 Open Source

PyPI package release, Neo4j / Kùzu / FalkorDB adapters, and full integration with solidstate-core.

Key Capabilities

Solving the state isolation and memory fragmentation problem in modern agent frameworks.

Automatic Triple Extraction

Parses execution histories and tool outputs in real-time, converting unstructured text into structured entity-relational property graphs.

Hybrid Graph-RAG Retrieval

Combines dense vector search with n-hop graph neighborhood traversals, ensuring LLMs receive precise subgraphs rather than noisy chunks.

Persistent Agent Memory

Enables long-term episodic and semantic memory across agent sessions. Knowledge evolves dynamically without full graph re-indexing.

Pluggable Graph DB Adapters

Supports embedded in-process engines (Kùzu, NetworkX) for zero-dependency local runs, and distributed databases (Neo4j, Memgraph, Neptune).

Deterministic Policy Gates

Applies node-level and edge-level authorization rules. Restrict agent graph mutations using the PolicyEvaluator engine.

API Preview

Designed to seamlessly plug into your existing SolidState agent runtimes.

solidstategraph_preview.pyComing Soon

from solidstate import AgentRuntime
from solidstategraph import KnowledgeGraph, KuzuAdapter
from solidstategraph.extractors import TripleExtractor

# 1. Initialize embedded Knowledge Graph management layer
graph = KnowledgeGraph(adapter=KuzuAdapter("./agent_memory.db"))

# 2. Attach Knowledge Graph to runtime with automatic triple extraction
runtime = AgentRuntime(
    model=model,
    knowledge_graph=graph,
    extractor=TripleExtractor(auto_canonicalize=True)
)

# 3. Query hybrid Graph RAG subgraphs on reasoning turns
subgraph = await graph.query_subgraph("Find all compliance policies for User C123", max_hops=2)
print(f"Triples retrieved: {len(subgraph.triples)}")

Q3 2026 Public Beta

Be the First to Build with SolidStateGraph

Join the early access developer waitlist. We will notify you the moment the open-source repository and PyPI package go public.

60Days

00Hours

00Min

00Sec

Release Notes