docs/content/api/python.mdx

---
title: Python API
description: Lux FHE Python API Reference
---

# Python API Reference

## Installation

```bash
pip install luxfhe

# With ML support
pip install luxfhe[ml]

# With agents support
pip install luxfhe[agents]

# Everything
pip install luxfhe[all]
```

## Core Classes

### Client

Manages private keys and encryption/decryption.

```python
from luxfhe import Client

client = Client()

# Encrypt values
ct = client.encrypt(42)
ct_bool = client.encrypt(True)
ct_array = client.encrypt_array([1, 2, 3, 4])
ct_json = client.encrypt_json({"key": "value"})

# Decrypt values
value = client.decrypt(ct)
bool_value = client.decrypt(ct_bool)
array = client.decrypt_array(ct_array)
data = client.decrypt_json(ct_json)

# Get evaluation key (safe to share)
eval_key = client.evaluation_key
```

### Server

Performs homomorphic operations (no private key).

```python
from luxfhe import Server

server = Server(client.evaluation_key)

# Arithmetic
ct_sum = server.add(ct_a, ct_b)
ct_diff = server.sub(ct_a, ct_b)
ct_prod = server.mul(ct_a, ct_b)
ct_neg = server.neg(ct_a)

# Comparisons
ct_lt = server.lt(ct_a, ct_b)   # a < b
ct_gt = server.gt(ct_a, ct_b)   # a > b
ct_eq = server.eq(ct_a, ct_b)   # a == b
ct_lte = server.lte(ct_a, ct_b) # a <= b
ct_gte = server.gte(ct_a, ct_b) # a >= b

# Bitwise
ct_and = server.and_(ct_a, ct_b)
ct_or = server.or_(ct_a, ct_b)
ct_xor = server.xor(ct_a, ct_b)
ct_not = server.not_(ct_a)

# Conditional
ct_result = server.select(ct_condition, ct_if_true, ct_if_false)
```

## ML Module

```python
from luxfhe.ml import EncryptedLinearRegression, EncryptedLogisticRegression

# Train on encrypted data
model = EncryptedLinearRegression()
model.fit(server, X_encrypted, y_encrypted)

# Predict on encrypted data
y_pred = model.predict(server, X_test_encrypted)
```

## Agents Module

```python
from luxfhe.agents import PrivateAgent

agent = PrivateAgent(
    model="gpt-4-turbo",
    fhe_server=server,
    system_prompt="You are a helpful assistant."
)

response = agent.run(
    query="What should I do?",
    encrypted_context=client.encrypt_json(sensitive_data)
)
```