Packages
Creates and verifies cryptographically signed Trust Records that prove an AI agent ran under a specific policy in a verified hardware environment, touching classified data and invoking identified tools.
However, it is in Alpha (v0.2) and explicitly marked for developer preview—do not rely on it in production without reviewing the Limitations page.
Provides PyTorch implementations of BERT, GPT, Transformer-XL, and GPT-2 models with pre-trained weights and fine-tuning examples for natural language processing tasks.
However, consider whether a more actively maintained alternative (such as HuggingFace Transformers, which this package predates) better suits your project's long-term…
Perceptron is a Python SDK for building physical AI applications with multimodal perception tasks—detection, localization, OCR, and visual Q&A—routing requests to specialized models with structured outputs ready for robotics and edge deployment.
However, verify the license terms before production use—the metadata does not specify whether it is open-source, proprietary, or commercial.
dstack is a unified control plane for provisioning and orchestrating GPU workloads across cloud providers, Kubernetes, and on-premises clusters, supporting NVIDIA, AMD, Google TPU, and Tenstorrent accelerators.
Python client library for Azure's Question Answering service, enabling you to query knowledge bases and manage question-answer projects via cloud API.
Install it if you need to query or manage Azure Question Answering services from Python; it's the only supported way to do so.
Adds stop-sequence support to Huggingface Transformers text generation, allowing you to halt output when specified text patterns appear.
Tavily CLI provides command-line and programmatic access to Tavily's web search, content extraction, website crawling, and AI-powered research APIs, with both interactive REPL and structured JSON output modes.
Official Python client for Clarifai's AI platform, enabling computer vision tasks (classification, detection, segmentation) and natural language processing (classification, summarization, generation, Q&A) through a unified API.
However, you are committing to Clarifai's ecosystem and pricing model; this is not a general-purpose ML library but a client for a specific commercial platform.
Implements equivariant convolutional neural networks for 3D rotations, translations, and mirrors using JAX, with an IrrepsArray data structure that tracks irreducible representations alongside tensor data.
Install it if you need to build or experiment with E(3)-equivariant models; skip it if you don't work with 3D geometric data or prefer PyTorch.
CPEX is a plugin framework that lets you intercept and enforce policies on AI agent operations through composable hooks and plugins, without modifying core application logic.
However, it is early-stage (0.1.3, released 2026-03-05) with 12 GitHub stars—evaluate stability and API maturity for your use case before committing to production.
Megatron-FSDP implements Fully Sharded Data Parallelism (FSDP) in native PyTorch to distribute training and inference of large models across multiple GPUs with configurable memory-communication trade-offs.
Provides utilities for handling raw HTTP requests in Azure ML webservices, built on Flask.
However, the Alpha status and unclear proprietary license mean you should verify the license terms for your use case and be prepared for potential API changes.
Pocket TTS generates speech from text on CPU using a lightweight 100M-parameter model, with support for multiple languages, voice cloning, and audio streaming via Python API or CLI.
Zuko implements normalizing flows in PyTorch as trainable neural modules, enabling conditional probability distributions and transformations that can be sent to GPU and optimized with standard PyTorch workflows.
Install it if you need conditional probability distributions or flow-based generative models in PyTorch.
Monkey-patches PyTorch tensors to display human-readable summaries with statistics, histograms, and anomaly detection instead of raw numerical dumps.
Metrax provides standard evaluation metrics for machine learning models implemented in JAX, supporting classification, regression, recommendation, and language modeling tasks.
Install it if you're building JAX-based ML models and need standard evaluation metrics without implementing them yourself.
Routes LLM requests to appropriate models in a mixture-of-models setup based on semantic intent classification, with a CLI for local Docker deployment, Kubernetes support, and a web dashboard for monitoring and testing.
An MCP server that bridges AI agents to Microsoft Fabric Real-Time Intelligence services, exposing Kusto queries, Eventstreams, Activators, and Maps as tools through the Model Context Protocol.
Anam plugin for the LiveKit Agents framework that extends voice agent capabilities with additional speech processing or model integration features.
However, verify that anam's functionality aligns with your use case before committing, as its exact role in the agent pipeline is not clearly documented in the…
QONNX provides Python utilities to work with quantized neural networks in ONNX format, including custom operators for arbitrary-precision integer and minifloat quantization, model execution, shape inference, and inference cost analysis.
Provides a Python client library for interacting with Microsoft Azure's QnA Maker service, though the package is no longer maintained and has been superseded by azure-ai-language-questionanswering.
Integrates LanceDB as a vector store backend for LlamaIndex, enabling semantic search and retrieval-augmented generation workflows with LanceDB's vector database.
Implements the mRMR (minimum Redundancy - Maximum Relevance) feature selection algorithm to identify the smallest subset of relevant features for classification and regression tasks across Pandas, Polars, Spark, and BigQuery.
However, the last release was 2023-06-30 with no recent commits, so expect no active support for new dependency versions or bug fixes.