CarvusTrain 1.0.0

CarvusTrain: A powerful AI development ecosystem — train, deploy, and serve AI models with code generation, grammar understanding, RAG, and agent training.

CarvusTrain

An AI Development Ecosystem — Train, Deploy, and Serve AI Models

Author: Aadil Fazal | GitHub | PyPI

CarvusTrain is a complete AI development ecosystem that goes beyond simple training. It features real code generation in 14+ programming languages, semantic knowledge bases with TF-IDF search, built-in English grammar understanding, REST API deployment, and goal-oriented agent training — all in pure Python with zero mandatory dependencies.

Quick Start

pip install -e .
python

from carvustrain import Model

# Create a model
model = Model(name="MyAI", architecture="transformer")
model.knowledge_base.load_grammar_knowledge()  # Load 62 grammar facts
model.learn("Python is a programming language.")
model.learn("JavaScript runs in browsers.")

# Ask questions
print(model.ask("What is Python?"))

# Generate real code
code = model.inference_engine.generate_code(
    "Write Dijkstra shortest path in Python"
)
print(code)

# Deploy as REST API
model.serve(port=8000)

Features

Core AI

• Model Builder — Architecture selection: transformer, neural_network, CNN, RNN, LSTM, custom
• AI Coding Modes — General, Programmer, Researcher, Teacher, Creative, Security, Data Scientist
• Auto AI Trainer — Automatic batch size, learning rate, and training strategy
• Memory Engine — Long-term conversation memory and context tracking
• Personality System — Configurable style, humor, and creativity traits
• Plugin System — Extensible architecture for custom abilities

Real Code Generation

• 50+ Algorithm Templates — DFS, BFS, Dijkstra, merge sort, linked list ops, tree traversals, DP patterns
• 14+ Programming Languages — Python, JavaScript, TypeScript, Java, C++, Rust, Go, Kotlin, Swift, Bash, Ruby, PHP, R, Scala
• Cross-Language Detection — Auto-detects language from code patterns
• Natural Language Aliases — "depth first search" → DFS, "shortest path" → Dijkstra

Knowledge & Learning

• Semantic Knowledge Base — TF-IDF search (pure Python), optional scikit-learn, optional sentence-transformers
• Learning Validator — Tracks accuracy, knowledge retention, language understanding, comprehension
• English Grammar — 62 grammar facts across 6 categories (parts of speech, tenses, rules, mistakes, style)
• Grammar-Aware Validation — Checks sentence structure and grammar term usage in responses

Deployment

• Model Server — REST API with POST / (ask) and GET / (info) endpoints
• Multi-Format Export — .ct, .json, .bin, .onnx, .gguf
• CLI — 15+ commands: create, init, train, chat, serve, deploy, export, evaluate, benchmark

Agent System

• Goal-Oriented Agents — Train agents for specific tasks
• Multi-Agent Teams — Add sub-agents and orchestrate complex workflows
• Custom Training — Agent-specific training with auto-optimization

CLI Reference

carvustrain create my-ai          # Create a new AI project
carvustrain train --data data.txt # Train a model
carvustrain chat --model my.ct    # Interactive chat
carvustrain serve --port 8000     # REST API server
carvustrain deploy --model my.ct  # Production deployment
carvustrain list                  # List .ct model files
carvustrain info --model my.ct    # Model details
carvustrain export --output out.onnx --format onnx
carvustrain evaluate --model my.ct --data test.txt
carvustrain benchmark --model my.ct --runs 50
carvustrain doctor                # System diagnostics

Algorithm Templates

Algorithm	Python	JavaScript	Rust	Go
Binary Search	✅	✅	✅	✅
Quick Sort	✅	✅	✅	✅
Merge Sort	✅	✅	✅	✅
DFS	✅	✅	✅	✅
BFS	✅	✅	✅	✅
Dijkstra	✅	✅	✅	✅
Reverse Linked List	✅	✅	✅	✅
Detect Cycle	✅	✅	—	✅
Tree Traversals	✅	✅	✅	✅
DP (LCS, Coin, Knap, Edit)	✅	—	—	—

Installation

# From source
git clone https://github.com/AadilFazal/CarvusTrain.git
cd CarvusTrain
pip install -e .

# Optional performance boosters
pip install scikit-learn        # Faster TF-IDF
pip install sentence-transformers  # Deep embeddings

Project Structure

CarvusTrain/
├── carvustrain/            # Core Package
│   ├── __init__.py         # Package entry point
│   ├── model.py            # Model, ChatModel, AgentModel, etc.
│   ├── memory.py           # KnowledgeBase, LearningValidator, Grammar
│   ├── inference.py        # CodeGenerator, QA, Text generation
│   ├── trainer.py          # Training loop with validation
│   ├── cli.py              # 15+ CLI commands
│   ├── exporter.py         # Multi-format export
│   ├── dataset.py          # Dataset & DataLoader
│   ├── tokenizer.py        # Tokenization suite
│   ├── ...                 # 20+ modules total
├── docs/                   # Documentation
├── tests/                  # Test suite
├── examples/               # Example scripts
├── train.ct                # Default training config
├── FULL_GUIDE.md           # Comprehensive guide
├── pyproject.toml          # Package config
└── README.md               # This file

License

MIT License — see LICENSE.