Unified Cognitive Architecture: 4096 CAM Operations • 144 Verbs • Quantum-Inspired Operators • 10K-bit Fingerprints
Crystal Lake is a unified cognitive database that treats everything as fingerprints (10,000-bit binary vectors). Unlike traditional databases that separate storage, query, and reasoning, Crystal Lake unifies them:
Traditional: SQL Database ←→ Graph Database ←→ Vector Store ←→ Reasoning Engine
(serialization, type conversion, impedance mismatch)
Crystal Lake: Everything is a Fingerprint
object ⊗ method ⊗ args → result
(same address space, no conversion, composable)
| Feature | Description |
|---|---|
| 4096 CAM Operations | Content-Addressable Methods - call operations by ID, name, or semantic description |
| 144 Verbs | Go-board topology for relationships (Structural, Causal, Temporal, Epistemic, Agentive, Experiential) |
| Quantum Operators | Linear mappings on fingerprint space with adjoints, composition, and measurement |
| Tree Addressing | 256-way hierarchical navigation (like LDAP Distinguished Names) |
| Cognitive Frameworks | Built-in NARS, ACT-R, RL, Causality, Qualia, Rung |
| Crystal LM | 5×5×5 compressed model (3.75 KB) achieving 140M× compression |
Add to your Cargo.toml:
[dependencies]
ladybug = "0.2"
# With all features
ladybug = { version = "0.2", features = ["full"] }
# Specific features
ladybug = { version = "0.2", features = ["lancedb", "neo4j", "quantum"] }# Clone
git clone https://github.com/AdaWorldAPI/ladybug-rs.git
cd ladybug-rs
# Build (release mode for SIMD optimizations)
cargo build --release
# Run tests
cargo test
# Run benchmarks
cargo bench- Rust: 1.75+ (for AVX-512 SIMD support)
- CPU: x86_64 with AVX2 (AVX-512 optional but recommended)
- Memory: 4GB minimum, 16GB recommended for large codebooks
- Storage: SSD recommended for LanceDB
| Category | Crate | Version | Purpose |
|---|---|---|---|
| Storage | arrow | 50 | Columnar data format |
| arrow-array | 50 | Arrow arrays | |
| arrow-schema | 50 | Schema definitions | |
| parquet | 50 | Parquet file format | |
| lance | 0.9 | Vector database (optional) | |
| Query | datafusion | 35 | SQL query engine |
| sqlparser | 0.43 | SQL parsing | |
| Graph | neo4rs | 0.7 | Neo4j driver (optional) |
| Cache | redis | 0.24 | Redis client (optional) |
| Async | tokio | 1.35 | Async runtime |
| futures | 0.3 | Futures utilities | |
| Parallel | rayon | 1.8 | Data parallelism |
| crossbeam | 0.8 | Concurrency primitives | |
| dashmap | 5.5 | Concurrent HashMap | |
| Serialization | serde | 1.0 | Serialization framework |
| serde_json | 1.0 | JSON support | |
| serde_yaml | 0.9 | YAML support | |
| bincode | 1.3 | Binary encoding | |
| Utilities | uuid | 1.6 | UUID generation |
| chrono | 0.4 | Date/time | |
| rand | 0.8 | Random numbers |
use ladybug::core::Fingerprint;
// Create fingerprints from content
let cat = Fingerprint::from_content("cat");
let dog = Fingerprint::from_content("dog");
// Similarity (Hamming-based)
let sim = cat.similarity(&dog);
// VSA Operations
let bound = cat.bind(&dog); // XOR binding
let recovered = bound.bind(&cat); // Recovers dog!use ladybug::graph::{CogGraph, CogNode, CogEdge, Verb, NodeType};
let mut graph = CogGraph::new();
// Create edge: cat IS_A animal
let edge = CogEdge::new(
cat_fp,
Verb::IsA, // One of 144 verbs
animal_fp
);
graph.add_edge(edge);use ladybug::learning::OpDictionary;
let dict = OpDictionary::new();
// Three ways to call:
dict.execute(0x310, &ctx, &[a, b]); // By ID
dict.execute_by_name("HAM_BIND", &ctx, &[a, b]); // By name
dict.execute_semantic("XOR bind", &ctx, &[a, b]); // Semantic!use ladybug::learning::{BindOp, PermuteOp, MeasureOp, QuantumOp};
// Operators are linear mappings
let op = BindOp::new(key);
let result = op.apply(&state);
// Adjoint reverses the operation
let adj = op.adjoint();
let recovered = adj.apply(&result);
// Measurement collapses to eigenstate
let measure = MeasureOp::new(eigenstates);
let collapsed = measure.apply(&superposition);0x000-0x0FF: LanceDB Core VectorSearch, Insert, Index
0x100-0x1FF: SQL SelectSimilar, SimilarJoin
0x200-0x2FF: Cypher/Neo4j MatchSimilar, PageRank
0x300-0x3FF: Hamming/VSA Bind, Bundle, MexicanHat
0x400-0x4FF: NARS Deduction, Induction, Abduction
0x500-0x5FF: Filesystem Read, Write, FindSimilar
0x600-0x6FF: Crystal AxisProject, Train, Infer
0x700-0x7FF: NSM 65 primes + Decompose
0x800-0x8FF: ACT-R Buffers, Chunks, Productions
0x900-0x9FF: RL Q-learning, Policy, Reward
0xA00-0xAFF: Causality do(), Counterfactual
0xB00-0xBFF: Qualia 8 channels, Blend, Shift
0xC00-0xCFF: Rung 10 levels, Ascend, Descend
0xD00-0xDFF: Meta Compose, Pipeline, Map
0xE00-0xEFF: Learning Moment, Resonance, Crystal
0xF00-0xFFF: User-Defined Extension space
Structural (24): IS_A, HAS_A, PART_OF, CONTAINS...
Causal (24): CAUSES, ENABLES, PREVENTS, TRIGGERS...
Temporal (24): BEFORE, AFTER, DURING, MEETS...
Epistemic (24): KNOWS, BELIEVES, INFERS, EXPECTS...
Agentive (24): DOES, WANTS, DECIDES, TRIES...
Experiential(24): SEES, FEELS, ENJOYS, FEARS...
| Framework | Purpose |
|---|---|
| NARS | Non-Axiomatic Reasoning: truth values, inference |
| ACT-R | Cognitive Architecture: buffers, chunks, activation |
| RL | Reinforcement Learning: Q-values, policy |
| Causality | Pearl's do-calculus: intervention, counterfactual |
| Qualia | 8 affect channels: arousal, valence, tension... |
| Rung | 10 abstraction levels: noise → transcendent |
ladybug-rs/ (26,919 lines)
├── src/
│ ├── core/ # Fingerprint operations
│ ├── graph/ # 144 verbs, CogGraph
│ │ └── cognitive.rs (955 lines)
│ ├── learning/ # Cognition (4,084 lines)
│ │ ├── cam_ops.rs (1,126 lines) - 4096 operations
│ │ ├── cognitive_frameworks.rs (934 lines) - NARS, ACT-R, RL...
│ │ ├── quantum_ops.rs (1,086 lines) - Operators + Tree
│ │ └── ...
│ └── extensions/
│ ├── cognitive_codebook.rs (1,122 lines)
│ └── crystal_lm.rs (817 lines)
├── Cargo.toml # Dependencies
└── README.md
| Operation | Throughput |
|---|---|
| Hamming Distance (AVX-512) | 400M ops/sec |
| Fingerprint Bind (XOR) | 200M ops/sec |
| Vector Search (10K in 1M) | 2ms |
| Crystal LM Compression | 140,000,000× |
Apache-2.0. See LICENSE for details.
Built with 🦋 for the Ada Consciousness Project