neuronlens 0.2.1

NeuronLens Python Package - Interpretability analysis functions

NeuronLens Python Package

A Python package for accessing NeuronLens interpretability analysis functions.

Publishing to PyPI

1. Update version in setup.py and __init__.py
2. Build package: cd sae_app_v3/pip && python -m build
3. Publish: export $(cat .env | xargs) && python -m twine upload dist/*

Note: PyPI API token in sae_app_v3/pip/.env (excluded from git).

Installation

pip install neuronlens

Development: cd sae_app_v3/pip && pip install -e .

Quick Start

API key required (via parameter or NEURONLENS_API_KEY env var):

import neuronlens
engine = neuronlens.Engine(api_key="nlive_your_api_key_here")

# Analyze agent query
result = engine.analyze_agent("Get Tesla stock price")
print(f"Alignment: {result['alignment_status']}")

# Analyze sentiment
result = engine.analyze_sentiment("Strong earnings", top_n=10)
print(f"Sentiment: {result['sentiment']['predicted_label']}")

Function Reference

High-Level: analyze_agent(), analyze_sentiment(), analyze_trading(), analyze_reasoning(), analyze_hallucination(), search_features(), steer_features()

Common Backend: extract_top_features(), model_cache_status(), unload_model(), unload_all_models()

Examples

# Sentiment
result = engine.analyze_sentiment("Strong quarterly earnings")
print(f"Label: {result['sentiment']['predicted_label']}")

# Search & Steer
features = engine.search_features("market volatility", k=10, layer=16)
steered = engine.steer_features(
    prompt="Market outlook?",
    features=[{"id": features[0]['feature_id'], "magnitude": 0.7}]
)

Architecture

• Direct mode: When sae_app_v3 available, imports functions directly
• HTTP mode: When sae_app_v3 not available, makes HTTP requests to frontend server (port 8001)

Server Requirements: Backend (port 8000) and Frontend (port 8001) must be running.

Configuration

API URL (priority): config.json → NEURONLENS_API_URL env → .ngrok_url file → Ngrok API → constructor param

API Key: NEURONLENS_API_KEY or API_KEY env var, .env file, or Engine(api_key="...")

Requirements

Python 3.8+, requests>=2.25.0, numpy>=1.20.0

License

MIT License