Packages
Memori is a Python SDK that automatically captures and recalls structured memory from LLM interactions and agent conversations, persisting context across sessions without requiring code changes to existing LLM calls.
A Python SDK for discovering, downloading, loading, and running local AI models with OpenAI-compatible chat, embeddings, and audio transcription APIs — no cloud required.
ale-py provides a Python interface to the Atari 2600 emulator Stella, allowing you to develop and test reinforcement learning agents on classic Atari games with automatic score extraction and game-over detection.
Install it if you are developing RL agents and need a standard Atari benchmark environment.
Counts multiply-accumulate operations (MACs) in PyTorch models by tracing the computation graph, providing both aggregate and per-operator breakdowns.
Provides Python bindings for NVIDIA TensorRT 11.2.1.2 compiled for CUDA 12, enabling deep learning inference acceleration on x86_64 Linux and Windows systems.
A Python SDK for building video and audio calling into desktop and server applications, with support for joining meetings, managing participants, and sending/receiving media streams.
Install only if your application targets Python 3.8+ and your deployment environment meets the glibc requirement.
Python wrapper for OpenFst and Kaldi extensions, providing finite-state transducer operations like fstdeterminizestar without requiring Kaldi installation.
PyTorch implementation of the CREPE pitch tracker that estimates fundamental frequency and periodicity from audio signals using pre-trained neural network models.
Provides computer vision annotation and detection models (edge detection, pose estimation, depth mapping, segmentation) packaged as PyPI-installable processors for use with ControlNet.
A PyTorch quantization backend that reduces model size and memory by converting weights and activations to lower-precision integer or float8 formats while maintaining inference accuracy.
Echo Agent is a self-hosted, long-running AI agent runtime that maintains cross-session memory, evolves skills from execution traces, and integrates with multiple communication channels (CLI, webhooks, messaging platforms) while enforcing tool approval and credential security.
BasicSR is a PyTorch-based toolbox for image and video restoration tasks including super-resolution, denoising, deblurring, and JPEG artifact removal.
However, expect high setup friction, no active maintenance, and unresolved security vulnerabilities.
Provides experimental lab modules (GAIA, TAU2, Lightning) built on Microsoft Agent Framework for benchmarking, evaluating, and training agents with research prototypes and incubating features.
Generates natural-language prompts from images by combining CLIP and BLIP models to reverse-engineer what text would recreate a given image.
Install only if you're comfortable pinning versions and managing torch dependencies yourself, and if the last release from 2023-03-20 meets your needs.
LeRobot provides a unified PyTorch framework for training, deploying, and managing robotic policies and datasets across diverse hardware platforms, from low-cost arms to humanoids.
TimesFM is a pretrained time-series foundation model from Google Research that generates point and quantile forecasts for time-series data using a decoder-only transformer architecture.
Install it if you need a modern pretrained foundation model for time-series forecasting; the permissive Apache-2.0 license and optional backends (torch/Flax) make it…
GFPGAN restores and enhances degraded faces in images using a generative adversarial network trained on facial priors, producing clearer face details while optionally upscaling the full image.
However, expect no active maintenance or updates—use it for production only if you can tolerate a frozen codebase and are comfortable debugging against a 2022-era…
Provides timelock encryption for Bittensor commit-reveal weight commitments and general-purpose data using drand randomness, with Rust core and Python bindings.
However, the repository is archived and marked abandoned, meaning no active maintenance or bug fixes should be expected.
SAHI performs sliced inference on large images to detect and segment small objects, breaking images into overlapping tiles for processing by detection models and merging results.
Install it if you work with large images containing small objects or need a framework-agnostic wrapper for detection inference.
openWakeWord detects wake words and phrases in audio streams using pre-trained neural network models, enabling voice-activated applications to listen for specific spoken triggers in real-time.
However, the aging maintenance status and lack of recent updates warrant caution if you require ongoing support or compatibility with the latest Python ecosystem.
Prodigy is a PyTorch optimizer that adapts the learning rate automatically during training, eliminating the need to manually tune this critical hyperparameter.
However, verify that PyTorch is available in your environment, as it is not declared as a package dependency.
Bridges Composio's tool catalog to OpenAI's Agents framework, letting agents execute actions across 1000+ integrated apps through a unified interface.
TensorFlow Probability provides probabilistic modeling, statistical inference, and Bayesian machine learning tools integrated with TensorFlow and JAX, including distributions, bijectors, inference algorithms, and probabilistic neural network layers.
BentoML is a Python framework for building and deploying REST APIs that serve AI and ML model inference at scale, with support for dynamic batching, multi-model orchestration, and containerized deployment.
SNAC compresses audio into discrete hierarchical tokens at low bitrate, enabling efficient audio representation for neural codecs and language modeling approaches to audio generation.
However, verify the license before production use, and note that maintenance is dormant—expect no active updates, though the package is stable and has no known…
Integrates pretrained transformer models (BERT, RoBERTa, XLNet, GPT-2) into spaCy pipelines via Hugging Face's transformers library, enabling state-of-the-art neural NLP components with automatic tokenization alignment.
Mooncake Transfer Engine provides high-performance, zero-copy data transfer for distributed KV cache management in large-scale LLM inference and training, using RDMA and P2P mechanisms to move cache and model weights across GPU clusters.
Provides an Adam optimizer variant using atan2 for numerical stability and scale invariance, with optional plasticity features for continual learning.
Integrates xAI's Grok LLM with LiveKit's Agent Framework to enable voice agents that can understand and respond in real-time conversations.
Collects metrics, configs, and performance profiles from ML workloads running on Google Cloud TPUs and GPUs, integrating with Google Cloud Logging and XProf for centralized diagnostics.
Provides REST client classes for creating and managing HyperDrive hyperparameter tuning runs within the Azure ML SDK ecosystem.
However, clarify whether you need it directly or if a higher-level Azure ML package already includes it as a dependency—it is a specialized component, not a…
pykalman provides Kalman filtering, smoothing, and expectation-maximization algorithms for state estimation in time series and sequential data, supporting linear and nonlinear dynamical systems.
onnx-weekly provides a Python package for working with ONNX (Open Neural Network Exchange) models—an open format for representing AI models with a computation graph, built-in operators, and standard data types for model interoperability and inference.
However, if you need production stability, use the stable onnx package instead.
DoWhy is a Python library for causal inference that identifies and estimates causal effects, performs root cause analysis, and answers what-if questions using graphical causal models and potential outcomes frameworks.
Install it if you work with observational data and need to move beyond correlation to causal reasoning—whether for effect estimation, root cause analysis, or…
Strands Agent Builder is a terminal-based toolkit for building, testing, and extending custom AI agents and tools with a model-driven approach using Claude and Amazon Bedrock.
However, the repository is archived and no longer maintained as of the fact sheet date; you should verify that its dependencies remain viable and that you are…
SetFit fine-tunes Sentence Transformers for text classification using only a handful of labeled examples per class, without requiring handcrafted prompts or large language models.
Install it if you need to classify text with limited labeled examples.
Argilla is a Python SDK client for a server-based collaboration platform that helps teams build and curate high-quality datasets for AI projects, with support for labeling, filtering, and semantic search workflows.
However, be aware that maintenance is aging (522 days since last release), so expect slower updates; verify compatibility with your runtime dependencies before…
Integrates Daytona sandboxes with LangChain's Deep Agents framework, allowing agents to execute code in isolated sandbox environments.
However, verify that daytona and deepagents meet your infrastructure requirements before committing, as this package is young (first release 2026-02-05) and…
Boruta performs all-relevant feature selection by identifying all features carrying predictive information, using ensemble methods with a scikit-learn-compatible interface.
Install it if you need all-relevant feature selection with a scikit-learn-compatible interface and can accept aging maintenance (last release 2024-08-13).
Extracts Kaldi-compatible filterbank (fbank) audio features from waveforms in real-time without external dependencies, supporting online processing across multiple architectures and operating systems.
However, if your use case is offline batch processing or you already have Kaldi or another audio library integrated, the aging maintenance status (309 days since last…