Enterprise-grade symbolic NLP reasoning system for edge environments.
100% Symbolic · 0% Neural · 100% Explainable · 100% Deterministic
🚀 Quick Start · 🧠 Architecture · 📊 Three Versions · 🌐 Multilingual · 🐛 Issues
Most NLP systems today require large language models, GPU infrastructure, and cloud connectivity. LRLRE proves none of that is necessary for high-quality language reasoning.
LRLRE is a 100% symbolic NLP reasoning engine — built on Robinson's unification, Forward/Backward Chaining inference, and NetworkX knowledge graphs. It runs on a Raspberry Pi. It supports 5 languages. It never hallucinates. Every inference is deterministic, explainable, and auditable.
100% Symbolic · 0% Neural · No LLMs · No GPU · No Cloud Required
| Metric | Value | Context |
|---|---|---|
| CRUD Operations | < 5ms | SQLite + SQLAlchemy ORM |
| Complex Reasoning Cycles | 50–100ms | Forward/Backward Chaining |
| Concurrent Users | 100+ verified | WebSocket + FastAPI |
| Knowledge Capacity | 10,000+ facts/rules | NetworkX graph storage |
| Memory Footprint | < 100MB | Edge deployment optimized |
| Languages Supported | 5 | EN · JA · KO · ZH · FR |
- 🧠 100% Symbolic Reasoning — No LLMs, no neural models, no probabilistic approximation. Every inference follows an explicit, auditable rule chain.
- 🔄 Deterministic Inference — Robinson's Unification Algorithm combined with Forward and Backward Chaining. Same input always produces the same output.
- 🌐 Multilingual Processing — English, Japanese (Janome/Sudachi), Korean, Chinese (Unicode character analysis), and French — all handled by a unified processing layer.
- 📊 Three Dashboard Versions — v7.0 analysis-first, v8.2 animation-first (Bento Grid), v10.0 combined ultimate edition. Choose your deployment style.
- ⚡ Edge-Optimized — < 100MB memory footprint. Runs on constrained hardware without cloud connectivity.
- 🕸️ Knowledge Graph Architecture — NetworkX graph structures with SQLite persistence. Supports 10,000+ facts and rules with < 5ms CRUD latency.
- 📡 Real-Time WebSocket — Live inference updates pushed to the dashboard without polling.
- 🎨 Enterprise UI — Interactive Bento Grid, flip card animations, multiple themes (Milky Way, Quantum Blue, Sunset), global mouse effects.
| Version | Codename | Entry Point | Port | Best For |
|---|---|---|---|---|
| v7.0 | 🧠 Analysis Edition | start_analytics_v7.py |
8007 | Detailed Unicode, logical, entity analysis |
| v8.2 | ✨ Bento Grid | start_analytics_v8_bento_grid.py |
8009 | Animations, flip cards, scroll effects |
| v10.0 | 💎 Ultimate Complete | ultimate_v10_fixed.py |
8013 | Everything combined ← Recommended |
Mermaid Diagrams — Paste at mermaid.live to Render & Export
💡 Copy any block → paste at mermaid.live → Export PNG/SVG instantly.
graph TD
subgraph UI["🖥️ User Interface Layer"]
V7["v7.0\nAnalysis Dashboard\nlocalhost:8007"]
V8["v8.2\nBento Grid Animations\nlocalhost:8009"]
V10["v10.0 ⭐\nUltimate Complete\nlocalhost:8013"]
end
subgraph API["⚡ FastAPI Server Layer"]
REST["REST API\nEndpoints"]
WS["Real-time\nWebSocket"]
VAL["Request\nValidation & Routing"]
end
subgraph ENGINE["🧠 Symbolic Reasoning Engine"]
UNI["Robinson's\nUnification Algorithm"]
FWD["Forward\nChaining"]
BWD["Backward\nChaining"]
CONF["Confidence Scoring\n0.5 – 1.0 scale"]
end
subgraph MULTI["🌐 Multilingual Processing Layer"]
EN["English\nTokenizer"]
JA["Japanese\nJanome / Sudachi"]
KO["Korean\nUnicode Analysis"]
ZH["Chinese\nCharacter Segmentation"]
FR["French\nProcessing"]
end
subgraph KG["🗄️ Knowledge Graph & Storage"]
SQLITE["SQLite\nSQLAlchemy ORM\n<5ms CRUD"]
NX["NetworkX\nGraph Structures"]
RULES["Rule\nPersistence\n10K+ facts"]
end
V7 --> REST
V8 --> REST
V10 --> REST
V10 --> WS
REST --> UNI
WS --> FWD
VAL --> BWD
UNI --> CONF
FWD --> CONF
BWD --> CONF
CONF --> EN
CONF --> JA
CONF --> KO
CONF --> ZH
CONF --> FR
EN --> SQLITE
JA --> NX
KO --> RULES
ZH --> SQLITE
FR --> NX
style UI fill:#0d1117,stroke:#ffc107,color:#c9d1d9
style API fill:#0d1117,stroke:#58a6ff,color:#c9d1d9
style ENGINE fill:#0d1117,stroke:#28a745,color:#c9d1d9
style MULTI fill:#0d1117,stroke:#e06c75,color:#c9d1d9
style KG fill:#0d1117,stroke:#6f42c1,color:#c9d1d9
flowchart LR
Q([User Query]) --> LANG{Language\nDetection}
LANG -- "EN" --> PARSE_EN[English\nParser]
LANG -- "JA" --> PARSE_JA[Janome\nTokenizer]
LANG -- "KO/ZH/FR" --> PARSE_MULTI[Unicode\nAnalyzer]
PARSE_EN --> UNIFY[Robinson's\nUnification]
PARSE_JA --> UNIFY
PARSE_MULTI --> UNIFY
UNIFY --> CHAIN{Chaining\nStrategy}
CHAIN -- "Known goal" --> BWD[Backward\nChaining]
CHAIN -- "New facts" --> FWD[Forward\nChaining]
FWD --> KG[Knowledge\nGraph Query\nNetworkX]
BWD --> KG
KG --> SCORE[Confidence\nScoring\n0.5–1.0]
SCORE --> AUDIT[Audit\nTrail\nGenerated]
AUDIT --> RESP([Deterministic\nResponse +\nExplanation])
style Q fill:#4A90D9,color:#fff
style RESP fill:#238636,color:#fff
style UNIFY fill:#ffc107,color:#000
style SCORE fill:#6f42c1,color:#fff
graph LR
subgraph V7["🧠 v7.0 — Analysis Edition\nPort 8007"]
A1[Detailed Unicode\nAnalysis]
A2[Logical Entity\nExtraction]
A3[Language Reference\nDatabase]
A4[Analytics\nDashboard]
end
subgraph V8["✨ v8.2 — Bento Grid\nPort 8009"]
B1[Bento Grid\nUI Layout]
B2[Flip Card\nAnimations]
B3[Scroll\nEffects]
B4[3 Themes\nMilky Way etc.]
end
subgraph V10["💎 v10.0 — Ultimate\nPort 8013 ⭐"]
C1[All v7.0\nFeatures]
C2[All v8.2\nFeatures]
C3[Global Mouse\nEffects]
C4[WebSocket\nReal-time]
end
V7 --> V10
V8 --> V10
style V7 fill:#0d1117,stroke:#58a6ff,color:#c9d1d9
style V8 fill:#0d1117,stroke:#ffc107,color:#c9d1d9
style V10 fill:#0d1117,stroke:#ffd700,color:#c9d1d9
xychart-beta
title "LRLRE Multilingual Support — Processing Latency (ms)"
x-axis ["English", "French", "Chinese", "Korean", "Japanese"]
y-axis "Processing Latency (ms)" 0 --> 120
bar [45, 48, 65, 70, 95]
💡 Run the PowerShell setup block first, then copy each script into
charts/and execute.
# Clone and enter repo
git clone https://github.com/Ariyan-Pro/lrlre.git
Set-Location lrlre
# Create and activate virtual environment
python -m venv venv
.\venv\Scripts\Activate.ps1
# Install dependencies
pip install -r requirements.txt
pip install matplotlib numpy
# Create charts directory
New-Item -ItemType Directory -Force -Path charts
# Verify
python -c "import matplotlib; print('Matplotlib:', matplotlib.__version__)"python charts/performance_benchmarks.py
Invoke-Item charts/performance_benchmarks.png# charts/performance_benchmarks.py
import matplotlib.pyplot as plt
import numpy as np
fig, axes = plt.subplots(1, 2, figsize=(13, 6))
fig.patch.set_facecolor('#0d1117')
# Left — Latency comparison
ax1 = axes[0]
ax1.set_facecolor('#161b22')
operations = ['CRUD\nOperations', 'Simple\nReasoning', 'Complex\nReasoning', 'Multilingual\nDetection']
latencies = [5, 20, 100, 70]
colors = ['#28a745', '#58a6ff', '#ffc107', '#e06c75']
bars = ax1.bar(operations, latencies, color=colors, width=0.55, zorder=3)
ax1.set_ylabel('Latency (ms)', color='#c9d1d9', fontsize=11)
ax1.set_title('LRLRE Operation Latencies\n(Lower is Better)', color='#c9d1d9', fontsize=12)
ax1.tick_params(colors='#c9d1d9')
ax1.spines[:].set_color('#30363d')
ax1.yaxis.grid(True, color='#30363d', alpha=0.5, zorder=0)
for bar, val in zip(bars, latencies):
label = f'<{val}ms' if val in [5, 20] else f'{val}ms'
ax1.text(bar.get_x() + bar.get_width() / 2, bar.get_height() + 1.5,
label, ha='center', color='white', fontsize=10, fontweight='bold')
# Right — Capacity metrics
ax2 = axes[1]
ax2.set_facecolor('#161b22')
metrics = ['Concurrent\nUsers', 'Knowledge\nCapacity\n(×100)', 'Confidence\nRange (%)']
values = [100, 100, 50] # 10K facts = 100×100, confidence 0.5-1.0 → 50%
colors2 = ['#ffd700', '#28a745', '#6f42c1']
bars2 = ax2.bar(metrics, values, color=colors2, width=0.45, zorder=3)
ax2.set_ylabel('Value', color='#c9d1d9', fontsize=11)
ax2.set_title('LRLRE Capacity Specifications\n(Higher is Better)', color='#c9d1d9', fontsize=12)
ax2.tick_params(colors='#c9d1d9')
ax2.spines[:].set_color('#30363d')
ax2.yaxis.grid(True, color='#30363d', alpha=0.5, zorder=0)
actual_labels = ['100+ users', '10,000+ facts', '0.5–1.0 scale']
for bar, label in zip(bars2, actual_labels):
ax2.text(bar.get_x() + bar.get_width() / 2, bar.get_height() + 1.5,
label, ha='center', color='white', fontsize=9, fontweight='bold')
plt.suptitle('LRLRE Enterprise Grid v13.0 — Performance Overview',
color='#c9d1d9', fontsize=13, y=1.01)
plt.tight_layout()
plt.savefig('charts/performance_benchmarks.png', dpi=150, bbox_inches='tight',
facecolor=fig.get_facecolor())
print("Saved: charts/performance_benchmarks.png")python charts/multilingual_radar.py
Invoke-Item charts/multilingual_radar.png# charts/multilingual_radar.py
import matplotlib.pyplot as plt
import numpy as np
languages = ['English', 'French', 'Chinese\n(ZH)', 'Korean\n(KO)', 'Japanese\n(JA)']
# Score = 100 - normalized_latency (higher = faster)
scores = [100, 98, 85, 82, 75]
N = len(languages)
angles = [n / float(N) * 2 * np.pi for n in range(N)]
angles += angles[:1]
scores_plot = scores + scores[:1]
fig, ax = plt.subplots(figsize=(8, 8), subplot_kw=dict(polar=True))
fig.patch.set_facecolor('#0d1117')
ax.set_facecolor('#161b22')
ax.plot(angles, scores_plot, 'o-', linewidth=2.5, color='#ffc107', zorder=3)
ax.fill(angles, scores_plot, alpha=0.2, color='#ffc107')
ax.set_xticks(angles[:-1])
ax.set_xticklabels(languages, color='#c9d1d9', fontsize=11)
ax.set_ylim(0, 100)
ax.set_yticks([25, 50, 75, 100])
ax.set_yticklabels(['25', '50', '75', '100'], color='#8b949e', fontsize=8)
ax.grid(color='#30363d', linewidth=0.8)
ax.spines['polar'].set_color('#30363d')
ax.set_title('LRLRE Multilingual Processing Speed\n(Higher = Faster Processing)',
color='#c9d1d9', fontsize=13, pad=22, y=1.08)
for angle, score, lang in zip(angles[:-1], scores, languages):
ax.annotate(f'{score}', xy=(angle, score), xytext=(angle, score + 7),
color='#ffd700', fontsize=10, fontweight='bold', ha='center')
plt.tight_layout()
plt.savefig('charts/multilingual_radar.png', dpi=150, bbox_inches='tight',
facecolor=fig.get_facecolor())
print("Saved: charts/multilingual_radar.png")python charts/version_comparison.py
Invoke-Item charts/version_comparison.png# charts/version_comparison.py
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors
import numpy as np
fig, ax = plt.subplots(figsize=(12, 5))
fig.patch.set_facecolor('#0d1117')
ax.set_facecolor('#161b22')
versions = ['v7.0\nAnalysis', 'v8.2\nBento Grid', 'v10.0\nUltimate ⭐']
features = ['Unicode\nAnalysis', 'Entity\nExtraction', 'Bento Grid\nUI', 'Flip Card\nAnimations',
'Multiple\nThemes', 'WebSocket\nReal-time', 'Mouse\nEffects', 'Analytics\nDashboard']
matrix = np.array([
[1.0, 1.0, 0.0, 0.0, 0.5, 0.5, 0.0, 1.0], # v7.0
[0.5, 0.5, 1.0, 1.0, 1.0, 0.5, 0.5, 0.5], # v8.2
[1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0], # v10.0
])
cmap = mcolors.LinearSegmentedColormap.from_list(
'lrlre', ['#161b22', '#1a4f1a', '#ffd700'], N=256)
im = ax.imshow(matrix, cmap=cmap, vmin=0, vmax=1, aspect='auto')
ax.set_xticks(range(len(features)))
ax.set_xticklabels(features, color='#c9d1d9', fontsize=9)
ax.set_yticks(range(len(versions)))
ax.set_yticklabels(versions, color='#c9d1d9', fontsize=11, fontweight='bold')
ax.set_title('LRLRE Dashboard Version Feature Matrix\nv10.0 Recommended — Complete Feature Set',
color='#c9d1d9', fontsize=13, pad=12)
ax.tick_params(colors='#c9d1d9')
labels_map = {1.0: '✅ Full', 0.5: '⚡ Partial', 0.0: '—'}
for i in range(len(versions)):
for j in range(len(features)):
ax.text(j, i, labels_map[matrix[i, j]], ha='center', va='center',
color='white', fontsize=8, fontweight='bold')
plt.colorbar(im, ax=ax, fraction=0.02, pad=0.02).set_label(
'Feature Coverage', color='#c9d1d9', fontsize=9)
plt.tight_layout()
plt.savefig('charts/version_comparison.png', dpi=150, bbox_inches='tight',
facecolor=fig.get_facecolor())
print("Saved: charts/version_comparison.png")python charts/tech_stack.py
Invoke-Item charts/tech_stack.png# charts/tech_stack.py
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
fig, ax = plt.subplots(figsize=(11, 6))
fig.patch.set_facecolor('#0d1117')
ax.set_facecolor('#161b22')
layers = ['Knowledge\nStorage', 'Language\nProcessing', 'Symbolic\nReasoning',
'API\nServer', 'UI\nDashboard']
tech = ['SQLite +\nNetworkX', 'Janome /\nSudachi', 'Robinson\nUnification +\nChaining',
'FastAPI +\nUvicorn', 'HTML5 /\nChart.js']
colors = ['#6f42c1', '#e06c75', '#28a745', '#58a6ff', '#ffc107']
sizes = [0.9, 0.85, 1.0, 0.95, 0.8] # relative importance
y_pos = range(len(layers))
bars = ax.barh(y_pos, sizes, color=colors, height=0.55, zorder=3)
ax.set_yticks(y_pos)
ax.set_yticklabels(layers, color='#c9d1d9', fontsize=10)
ax.set_xlim(0, 1.4)
ax.set_xlabel('Relative Role Weight', color='#c9d1d9', fontsize=11)
ax.set_title('LRLRE Technology Stack — Component Breakdown\nv13.0 Enterprise Grid',
color='#c9d1d9', fontsize=13, pad=12)
ax.tick_params(colors='#c9d1d9')
ax.spines[:].set_color('#30363d')
ax.xaxis.grid(True, color='#30363d', alpha=0.4, zorder=0)
ax.set_xticks([])
for bar, tech_name, color in zip(bars, tech, colors):
ax.text(bar.get_width() + 0.02, bar.get_y() + bar.get_height() / 2,
tech_name, va='center', color=color, fontsize=9, fontweight='bold')
plt.tight_layout()
plt.savefig('charts/tech_stack.png', dpi=150, bbox_inches='tight',
facecolor=fig.get_facecolor())
print("Saved: charts/tech_stack.png")# Clone repository
git clone https://github.com/Ariyan-Pro/lrlre.git
Set-Location lrlre
# Create virtual environment
python -m venv venv
.\venv\Scripts\Activate.ps1
# Install dependencies
pip install -r requirements.txt
# Initialize knowledge database
python -c "from lrlre.symbols.persistence import init_db; init_db()"# Option 1 — v7.0 Analysis Edition (port 8007)
python start_analytics_v7.py
Start-Process "http://localhost:8007"
# Option 2 — v8.2 Bento Grid Animations (port 8009)
python start_analytics_v8_bento_grid.py
Start-Process "http://localhost:8009"
# Option 3 — v10.0 Ultimate Complete (port 8013) ← RECOMMENDED
python ultimate_v10_fixed.py
Start-Process "http://localhost:8013"
# Or use the included batch launcher
.\LAUNCH_LRLRE.bat
# Or the PowerShell launcher
.\LAUNCH_LRLRE.ps1
# Or create a desktop shortcut
.\CREATE_DESKTOP_SHORTCUT.ps1# Test multilingual language detection
$body = @{ text = "Hello, how are you?" } | ConvertTo-Json
Invoke-RestMethod -Uri "http://localhost:8013/detect" `
-Method Post -ContentType "application/json" -Body $body
# Test symbolic reasoning
$body = @{ query = "Is Socrates mortal?" } | ConvertTo-Json
Invoke-RestMethod -Uri "http://localhost:8013/reason" `
-Method Post -ContentType "application/json" -Body $body
# Test Japanese input
$body = @{ text = "人工知能とは何ですか？" } | ConvertTo-Json
Invoke-RestMethod -Uri "http://localhost:8013/detect" `
-Method Post -ContentType "application/json" -Body $body
# Health check
Invoke-RestMethod -Uri "http://localhost:8013/health" -Method Get# Language detection tests
python -m pytest tests/test_language_detection.py -v
# Inference engine tests
python -m pytest tests/test_inference.py -v
# Full test suite
python -m pytest tests/ -v --tb=short# Deploy v7.0
gunicorn -w 4 -k uvicorn.workers.UvicornWorker start_analytics_v7:app
# Deploy v10.0 (recommended)
gunicorn -w 4 -k uvicorn.workers.UvicornWorker ultimate_v10_fixed:app --bind 0.0.0.0:8013# Build image
docker build -t lrlre-enterprise:v13.0 .
# Run v10.0 Ultimate
docker run -d -p 8013:8013 --name lrlre lrlre-enterprise:v13.0
# Or use Docker Compose
docker-compose up -d
# Check logs
docker logs -f lrlre
# Open dashboard
Start-Process "http://localhost:8013"The repo includes a pre-configured nginx.conf for production reverse proxy:
# Apply nginx config (Linux)
sudo cp nginx.conf /etc/nginx/sites-available/lrlre
sudo ln -s /etc/nginx/sites-available/lrlre /etc/nginx/sites-enabled/
sudo nginx -t && sudo systemctl reload nginx| Technology | Purpose | Version |
|---|---|---|
| FastAPI | Web framework & WebSocket server | 0.104.1 |
| SQLAlchemy | Database ORM (SQLite backend) | 2.0.23 |
| NetworkX | Knowledge graph structures | 3.6.1 |
| Pydantic | Data validation & schema enforcement | 2.12.5 |
| Uvicorn | ASGI server | 0.24.0 |
| Janome / Sudachi | Japanese morphological analysis | Latest |
| Chart.js | Dashboard visualizations | CDN |
| Font Awesome | UI icons | CDN |
| Language | Code | Tokenizer | Unicode Range | Notes |
|---|---|---|---|---|
| English | EN | Whitespace + regex | ASCII | Baseline, fastest |
| French | FR | Whitespace + diacritics | Latin Extended | Accent-aware |
| Chinese | ZH | Character segmentation | U+4E00–U+9FFF | CJK unified |
| Korean | KO | Unicode analysis | U+AC00–U+D7A3 | Hangul syllables |
| Japanese | JA | Janome + Sudachi | U+3040–U+30FF + Kanji | Most complex, full morphology |
lrlre/
├── logo.png # Project logo
├── start_analytics_v7.py # v7.0 — Analysis dashboard (port 8007)
├── start_analytics_v8_bento_grid.py # v8.2 — Bento Grid UI (port 8009)
├── ultimate_v10_fixed.py # v10.0 — Ultimate Complete (port 8013) ⭐
├── v10_wrapper.py # v10.0 wrapper utility
│
├── lrlre/ # Core engine modules
│ ├── multilingual/ # Language detection & processing
│ ├── inference/ # Robinson's unification + chaining
│ ├── symbols/ # Knowledge graph & SQLite persistence
│ └── syntax/ # Grammar parsing
│
├── configs/ # System configuration files
├── data/ # Knowledge database & rule files
├── benchmarks/ # Performance benchmark scripts
├── results/ # Benchmark output and reports
├── tests/ # Test suite
│ ├── test_language_detection.py
│ └── test_inference.py
│
├── bin/ # Binary utilities
├── LAUNCH_LRLRE.bat # Windows batch launcher
├── LAUNCH_LRLRE.ps1 # PowerShell launcher
├── CREATE_DESKTOP_SHORTCUT.ps1 # Desktop shortcut creator
├── start_lrlre.bat # Alternative batch launcher
├── Dockerfile # Docker image definition
├── docker-compose.yml # Docker Compose stack
├── nginx.conf # Production reverse proxy config
├── pyproject.toml
├── requirements.txt
└── LICENSE.md
- Model Type: 100% Symbolic (rule-based) — no neural weights, no training data, no probabilistic outputs
- Inference Method: Robinson's Unification + Forward/Backward Chaining — fully deterministic
- External APIs: None — fully local, no cloud connectivity required
- Hallucination Risk: Zero — symbolic systems only produce outputs derivable from the explicit knowledge base
- Explainability: Every inference produces a complete audit trail showing the exact rule chain used
- Confidence Scoring: 0.5–1.0 scale based on rule chain depth and knowledge base coverage
- Known Limitations: Requires manual knowledge base construction. Cannot generalize beyond explicitly defined rules and facts.
# Fork and create feature branch
git checkout -b feature/AmazingFeature
# Make changes, add tests
python -m pytest tests/ -v
# Commit and push
git commit -m "feat: Add AmazingFeature"
git push origin feature/AmazingFeature
# Open a Pull RequestMIT © 2026 Ariyan Pro
- FastAPI — High-performance async web framework
- SQLAlchemy — Robust database ORM
- NetworkX — Knowledge graph structures
- Janome / Sudachi — Japanese morphological analysis
- Chart.js — Dashboard visualizations
- Font Awesome — UI icons
100% Symbolic · 0% Neural · 100% Explainable
Built with ❤️ by the LRLRE Project Team
⭐ Star on GitHub · 🐛 Report Issue · 🚀 Latest Release