AI/ML●●Solid
Kremis – Deterministic memory graph for AI agents (Rust)
Deterministic graph memory vs. embeddings is clever, but fragmentation of AI memory ecosystems is already painful.
Big BrainWizardry
M2Dr3g0n
103mo ago
Back to browse
GitHub Repository
A minimal graph engine for grounded AI — records, associates, and retrieves, but never invents. Written in Rust.
12 starsRust
Kremis – Graph-based memory for AI agents with no hidden state (Rust)
by M2Dr3g0n·Feb 15, 2026·1 point·0 comments
AI Analysis
●●SolidBig BrainNiche Gem
Deterministic graph memory that traces every result back to ingested data—no hallucination by design.
Strengths
- •Core insight is sound: determinism + traceability solves hallucination and opacity in agent memory
- •ACID persistence via redb + MCP server integration means real deployment path, not just prototype
- •No floating-point in core logic and ~277 tests show serious engineering discipline
Weaknesses
- •Experimental (v0.3.1) with incomplete features; graph-as-memory for agents is unsolved territory
- •Narrow niche: competes against vector DBs and RAG systems; unclear when deterministic graphs outperform embeddings
Category
Target Audience
AI agent developers, LLM app builders seeking grounded memory systems
Similar To
Qdrant · Weaviate · Milvus
Post Description
Hi HN — I built Kremis, a deterministic graph engine designed as a memory substrate for AI agents. Written in Rust, development was heavily AI-assisted.
The core idea: agent memory should be inspectable, deterministic, and honest.
- Same input → same output. No randomness, no floating-point in core. - Every query result traces back to a concrete graph path — no hidden state. - Zero pre-loaded knowledge. All structure emerges from ingested signals. - ACID transactions via redb. Crash-safe persistent storage.
It ships as a library (kremis-core, pure Rust, no async), an HTTP API + CLI, and an MCP server so AI assistants like Claude can query the graph directly.
Current state: v0.3.1, experimental, ~277 tests, CI on 3 OS, Docker image.
I'd value feedback on: - Does the deterministic graph approach make sense for agent memory? - API ergonomics — is the query model (lookup/traverse/path/intersect) intuitive? - What failure modes should I prioritize testing?
Similar Projects
Developer Tools●●Solid
Graph-flow – LangGraph-inspired AI agent workflows in Rust
LangGraph patterns in Rust with type safety, 300 stars and real production examples.
Big BrainShip It
alonagmon
201mo ago
AI/ML●●Solid
Unify memory across agents and improve context rot, written in Rust
Unifies siloed agent memory via MCP, but Mem0 and Zep already exist.
Niche GemSolve My ProblemShip It
feelingsonice
512mo ago
AI/ML●●Solid
Gliding Horse – a Rust Agent OS with CPU-Like Memory Architecture
MESI coherence for agent memory is clever, but LangGraph and AutoGen already own this space.
Big BrainBold Bet
doiito
1017d ago
AI/ML●●Solid
Kremis – Rust graph DB; every answer is fact, inference, or unknown
Classifies AI query results as fact, inference, or unknown without using embeddings.
Big BrainNiche Gem
TyKolt
313mo ago
Developer Tools●●●Banger
FlowScript – Agent memory where contradictions are features
Five queries vector stores can't answer: why(), tensions(), blocked(), whatIf(), alternatives().
Big BrainZero to OneWizardry
phillipclapham
212mo ago