neuron-lang 1.0.1__tar.gz

neuron_lang-1.0.1/PKG-INFO

@@ -0,0 +1,252 @@
+Metadata-Version: 2.4
+Name: neuron-lang
+Version: 1.0.1
+Summary: AI-First Programming Language
+License: MIT
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+Requires-Dist: numpy>=1.21.0
+Provides-Extra: full
+Requires-Dist: matplotlib>=3.5.0; extra == "full"
+Requires-Dist: Pillow>=9.0.0; extra == "full"
+
+# Neuron Language 
+
+> The AI-First Programming Language. Built for humans, designed for Machine Learning.
+
+Neuron (`.nl`) is a high-level, Python-like, OOP-based programming language built from the ground up for Artificial Intelligence training, tensors, and neural networks. It features a clean, English-like syntax that replaces traditional programming keywords with intuitive verbs.
+
+---
+
+## Installation
+
+You can install Neuron directly from PyPI on any Windows, Mac, or Linux machine:
+
+```bash
+pip install neuron-lang
+```
+
+---
+
+## Running Code
+
+Once installed, you can use the `neuron` command to run files, compile them, or start the interactive REPL.
+
+```bash
+# Start the interactive coding environment
+neuron
+
+# Run a Neuron script
+neuron run my_script.nl
+
+# Compile for faster future execution
+neuron compile my_script.nl
+```
+
+---
+
+## Package Manager (Auto-Install!)
+
+Neuron has built-in integration with the entire Python ecosystem. Any Python library (NumPy, Pandas, TensorFlow, PyTorch) can be used in Neuron!
+
+If you `grab` a library that isn't installed, **Neuron will automatically install it for you in the background!**
+
+```neuron
+## If pandas isn't installed, Neuron auto-installs it and imports it!
+grab pandas
+
+## You can also explicitly manage packages via the CLI:
+# neuron get scikit-learn
+# neuron packages
+# neuron remove pandas
+```
+
+---
+
+## Language Syntax & Features
+
+Neuron uses natural language keywords to make coding more intuitive.
+
+### Variables & Printing
+```neuron
+## Variables don't need type declarations
+name = "AI Builder"
+version = 1.0
+
+## String interpolation uses {brackets}
+show("Welcome {name} to Neuron v{version}!")
+
+## Ask for user input
+user = ask("What is your name? ")
+```
+
+### Functions
+Use `do` to define a function, and `give` to return a value.
+
+```neuron
+do add_numbers(x, y):
+    result = x + y
+    give result
+
+show(add_numbers(10, 5))  ## Output: 15
+```
+
+### Conditions
+Neuron replaces `if/elif/else` with `when/orwhen/otherwise`.
+
+```neuron
+age = 20
+
+when age < 18:
+    show("Too young!")
+orwhen age == 18:
+    show("Just right!")
+otherwise:
+    show("Welcome!")
+```
+
+### Loops
+Neuron features a powerful `each ... within` loop construct.
+
+```neuron
+## Loop through a list
+each item within [1, 2, 3]:
+    show(item)
+
+## Loop through a range using upto()
+each i within upto(5):
+    show(i)  ## Outputs 0, 1, 2, 3, 4
+```
+
+### The Pipeline Operator `|>`
+Easily chain function calls together reading left-to-right!
+
+```neuron
+do double(x):
+    give x * 2
+
+do add_ten(x):
+    give x + 10
+
+## Passes 5 into double(), then passes the result into add_ten()
+result = 5 |> double |> add_ten
+show(result)  ## Output: 20
+```
+
+---
+
+## Object-Oriented Programming (OOP)
+
+Neuron uses `entity` instead of `class`, and `me` instead of `self` or `this`.
+
+```neuron
+entity Animal:
+    do create(me, name):
+        me.name = name
+
+    do speak(me):
+        show("{me.name} makes a sound.")
+
+## Inheritance using 'from'
+entity Dog from Animal:
+    do speak(me):
+        show("{me.name} says Woof!")
+
+buddy = Dog("Buddy")
+buddy.speak()  ## Output: Buddy says Woof!
+```
+
+---
+
+## AI-Native Features (`brain`)
+
+Neuron has a special class type called `brain` specifically for AI models. Brains automatically have `train()`, `predict()`, and `save()` functionalities built-in.
+
+```neuron
+grab neural
+
+brain MyNet:
+    do create(me):
+        me.add_layer(neural.Dense(128))
+        me.add_layer(neural.ReLU())
+        me.add_layer(neural.Dense(10))
+
+model = MyNet()
+show(model.status)  ## Output: NOT TRAINED
+model.train(data)
+```
+
+---
+
+## Tensors and Deep Learning
+
+Neuron ships with a built-in standard library for tensors (`tensor`) with automatic differentiation (Autograd) for building neural networks from scratch.
+
+### Compile-Time Shape Checks
+
+The compiler and interpreter check statically known tensor shapes before running a program. Use a typed tensor declaration to declare an expected shape:
+
+```neuron
+grab tensor
+
+tensor[100, 784] x = tensor.random(100, 784)
+tensor[784, 10] w = tensor.random(784, 10)
+tensor[100, 10] y = x.matmul(w)
+```
+
+The current analysis infers shapes from `tensor.random`, `tensor.rand`, `tensor.zeros`, and `tensor.ones`, propagates them through assignments, and checks rank-2 `matmul` and elementwise broadcasting. It reports a compile error when a known result conflicts with a declaration or matrix dimensions do not align. Shapes that cannot be inferred statically are left for runtime validation.
+
+### Linear Regression Example
+```neuron
+grab tensor
+
+## Generate Data
+x = tensor.random(100, 1)
+y = x * 3.0 + 2.0
+
+## Weights
+w = tensor.random(1, 1)
+b = tensor.zeros(1, 1)
+lr = 0.01
+
+## Training Loop
+each epoch within upto(100):
+    ## Forward pass
+    y_pred = x.matmul(w) + b
+    
+    ## MSE Loss
+    diff = y_pred - y
+    loss = (diff * diff).mean()
+    
+    ## Manual Gradients
+    grad_w = x.transpose().matmul(diff) * 0.02
+    grad_b = diff.mean() * 2.0
+    
+    ## Update
+    w = w - grad_w * lr
+    b = b - grad_b * lr
+
+show("Learned w: {w}, b: {b}")
+```
+
+---
+
+## Built-In Standard Libraries
+
+Neuron comes batteries-included with AI primitives:
+* `tensor` - Tensors and math operations
+* `neural` - Neural network layers (Dense, Conv2d, ReLU, Softmax)
+* `optim` - Optimizers (SGD, Adam)
+* `data` - Data loading and batching
+* `vision` / `nlp` - Image processing and text tokenization
+* `metrics` - Accuracy, MSE, MAE
+* `plot` - Easy data visualization
+
+## Compiler and Runtime Status
+
+Neuron currently uses a Python interpreter and a Python bytecode compiler. `.nlc` files are Python-pickled bytecode chunks; they are not yet a standalone executable format or a native VM. The shape-analysis pass is an initial compiler foundation, not a native tensor type system.
+
+Native CPU/CUDA memory management and GIL-free execution, a linear quantum type checker, hybrid parameter-shift training, OpenQASM 3 code generation, and IBM Quantum Runtime submission are not implemented yet. They require separate compiler/runtime modules and explicit circuit syntax and ownership rules; no claim of hardware support is made by this release.
+
+---
+*Created by the Neuron Developer Community*

neuron_lang-1.0.1/README.md

@@ -0,0 +1,240 @@
+# Neuron Language 
+
+> The AI-First Programming Language. Built for humans, designed for Machine Learning.
+
+Neuron (`.nl`) is a high-level, Python-like, OOP-based programming language built from the ground up for Artificial Intelligence training, tensors, and neural networks. It features a clean, English-like syntax that replaces traditional programming keywords with intuitive verbs.
+
+---
+
+## Installation
+
+You can install Neuron directly from PyPI on any Windows, Mac, or Linux machine:
+
+```bash
+pip install neuron-lang
+```
+
+---
+
+## Running Code
+
+Once installed, you can use the `neuron` command to run files, compile them, or start the interactive REPL.
+
+```bash
+# Start the interactive coding environment
+neuron
+
+# Run a Neuron script
+neuron run my_script.nl
+
+# Compile for faster future execution
+neuron compile my_script.nl
+```
+
+---
+
+## Package Manager (Auto-Install!)
+
+Neuron has built-in integration with the entire Python ecosystem. Any Python library (NumPy, Pandas, TensorFlow, PyTorch) can be used in Neuron!
+
+If you `grab` a library that isn't installed, **Neuron will automatically install it for you in the background!**
+
+```neuron
+## If pandas isn't installed, Neuron auto-installs it and imports it!
+grab pandas
+
+## You can also explicitly manage packages via the CLI:
+# neuron get scikit-learn
+# neuron packages
+# neuron remove pandas
+```
+
+---
+
+## Language Syntax & Features
+
+Neuron uses natural language keywords to make coding more intuitive.
+
+### Variables & Printing
+```neuron
+## Variables don't need type declarations
+name = "AI Builder"
+version = 1.0
+
+## String interpolation uses {brackets}
+show("Welcome {name} to Neuron v{version}!")
+
+## Ask for user input
+user = ask("What is your name? ")
+```
+
+### Functions
+Use `do` to define a function, and `give` to return a value.
+
+```neuron
+do add_numbers(x, y):
+    result = x + y
+    give result
+
+show(add_numbers(10, 5))  ## Output: 15
+```
+
+### Conditions
+Neuron replaces `if/elif/else` with `when/orwhen/otherwise`.
+
+```neuron
+age = 20
+
+when age < 18:
+    show("Too young!")
+orwhen age == 18:
+    show("Just right!")
+otherwise:
+    show("Welcome!")
+```
+
+### Loops
+Neuron features a powerful `each ... within` loop construct.
+
+```neuron
+## Loop through a list
+each item within [1, 2, 3]:
+    show(item)
+
+## Loop through a range using upto()
+each i within upto(5):
+    show(i)  ## Outputs 0, 1, 2, 3, 4
+```
+
+### The Pipeline Operator `|>`
+Easily chain function calls together reading left-to-right!
+
+```neuron
+do double(x):
+    give x * 2
+
+do add_ten(x):
+    give x + 10
+
+## Passes 5 into double(), then passes the result into add_ten()
+result = 5 |> double |> add_ten
+show(result)  ## Output: 20
+```
+
+---
+
+## Object-Oriented Programming (OOP)
+
+Neuron uses `entity` instead of `class`, and `me` instead of `self` or `this`.
+
+```neuron
+entity Animal:
+    do create(me, name):
+        me.name = name
+
+    do speak(me):
+        show("{me.name} makes a sound.")
+
+## Inheritance using 'from'
+entity Dog from Animal:
+    do speak(me):
+        show("{me.name} says Woof!")
+
+buddy = Dog("Buddy")
+buddy.speak()  ## Output: Buddy says Woof!
+```
+
+---
+
+## AI-Native Features (`brain`)
+
+Neuron has a special class type called `brain` specifically for AI models. Brains automatically have `train()`, `predict()`, and `save()` functionalities built-in.
+
+```neuron
+grab neural
+
+brain MyNet:
+    do create(me):
+        me.add_layer(neural.Dense(128))
+        me.add_layer(neural.ReLU())
+        me.add_layer(neural.Dense(10))
+
+model = MyNet()
+show(model.status)  ## Output: NOT TRAINED
+model.train(data)
+```
+
+---
+
+## Tensors and Deep Learning
+
+Neuron ships with a built-in standard library for tensors (`tensor`) with automatic differentiation (Autograd) for building neural networks from scratch.
+
+### Compile-Time Shape Checks
+
+The compiler and interpreter check statically known tensor shapes before running a program. Use a typed tensor declaration to declare an expected shape:
+
+```neuron
+grab tensor
+
+tensor[100, 784] x = tensor.random(100, 784)
+tensor[784, 10] w = tensor.random(784, 10)
+tensor[100, 10] y = x.matmul(w)
+```
+
+The current analysis infers shapes from `tensor.random`, `tensor.rand`, `tensor.zeros`, and `tensor.ones`, propagates them through assignments, and checks rank-2 `matmul` and elementwise broadcasting. It reports a compile error when a known result conflicts with a declaration or matrix dimensions do not align. Shapes that cannot be inferred statically are left for runtime validation.
+
+### Linear Regression Example
+```neuron
+grab tensor
+
+## Generate Data
+x = tensor.random(100, 1)
+y = x * 3.0 + 2.0
+
+## Weights
+w = tensor.random(1, 1)
+b = tensor.zeros(1, 1)
+lr = 0.01
+
+## Training Loop
+each epoch within upto(100):
+    ## Forward pass
+    y_pred = x.matmul(w) + b
+    
+    ## MSE Loss
+    diff = y_pred - y
+    loss = (diff * diff).mean()
+    
+    ## Manual Gradients
+    grad_w = x.transpose().matmul(diff) * 0.02
+    grad_b = diff.mean() * 2.0
+    
+    ## Update
+    w = w - grad_w * lr
+    b = b - grad_b * lr
+
+show("Learned w: {w}, b: {b}")
+```
+
+---
+
+## Built-In Standard Libraries
+
+Neuron comes batteries-included with AI primitives:
+* `tensor` - Tensors and math operations
+* `neural` - Neural network layers (Dense, Conv2d, ReLU, Softmax)
+* `optim` - Optimizers (SGD, Adam)
+* `data` - Data loading and batching
+* `vision` / `nlp` - Image processing and text tokenization
+* `metrics` - Accuracy, MSE, MAE
+* `plot` - Easy data visualization
+
+## Compiler and Runtime Status
+
+Neuron currently uses a Python interpreter and a Python bytecode compiler. `.nlc` files are Python-pickled bytecode chunks; they are not yet a standalone executable format or a native VM. The shape-analysis pass is an initial compiler foundation, not a native tensor type system.
+
+Native CPU/CUDA memory management and GIL-free execution, a linear quantum type checker, hybrid parameter-shift training, OpenQASM 3 code generation, and IBM Quantum Runtime submission are not implemented yet. They require separate compiler/runtime modules and explicit circuit syntax and ownership rules; no claim of hardware support is made by this release.
+
+---
+*Created by the Neuron Developer Community*

neuron_lang-1.0.1/neuron/__init__.py

@@ -0,0 +1,18 @@
+"""
+Neuron Programming Language
+===========================
+A Python-like, OOP-based programming language purpose-built for
+Artificial Intelligence development.
+
+File extension: .nl
+Website: https://neuron-lang.org (coming soon)
+
+Keywords: do, give, show, ask, when, orwhen, otherwise, each,
+          repeat, grab, entity, brain, me, create, yes, no,
+          nothing, also, either, nah, safe, rescue, stop, skip,
+          quick, within, drop, ensure, emit, shared, from, upto
+"""
+
+__version__ = "1.0.1"
+__author__ = "Neuron Language Team"
+__license__ = "MIT"