Packages
Python client library for IBM Watson Machine Learning service, enabling model training, testing, and deployment as APIs on IBM Cloud and IBM Cloud Pak for Data.
Evaluates image captions using multiple automatic metrics (BLEU, METEOR, ROUGE-L, CIDEr, SPICE) designed for the MS COCO dataset, providing Python 3 support for caption generation assessment.
However, verify the license terms first, and be aware that the package is dormant—no updates are expected, and you should confirm that all five metrics work correctly…
Trustcall helps LLMs reliably generate and update complex JSON structures by using JSON patch operations instead of full regeneration, with built-in retry logic for validation errors.
Integrates Anthropic's Claude models into LiveKit's agent framework, enabling voice agents to use Claude for natural language understanding and generation in real-time conversations.
Portkey is a Python SDK that wraps OpenAI-compatible APIs to add monitoring, failover, load balancing, caching, and observability without rewriting existing code.
kernels-data provides Python bindings for kernel data structures used by the Hugging Face kernels ecosystem, enabling dynamic loading and execution of optimized compute kernels from the Hub.
However, verify the unclear license terms before use in proprietary projects, and confirm your environment meets the stated requirements.
Client library for pyannoteAI's speaker diarization, speaker identification, and speech-to-text orchestration services via REST API.
However, verify the license terms before use, and be aware the package is aging with no visible repository or documentation—assess whether you're comfortable relying…
Trains interpretable machine learning models and explains blackbox predictions using techniques like Explainable Boosting Machines, SHAP, LIME, and decision trees.
Scrapling is a web scraping and crawling framework that handles single requests to full-scale crawls, with built-in anti-bot bypass, adaptive element relocation, proxy rotation, and concurrent spider support.
Install it if you need to scrape protected or dynamic websites at scale; skip it if you only need simple static HTML parsing.
libtpu is the runtime library that enables JAX, PyTorch, and TensorFlow to run models on Google Cloud TPUs, handling compilation, inter-chip communication, and execution.
Provides integration modules connecting Optuna hyperparameter optimization with third-party ML frameworks like PyTorch, scikit-learn, TensorFlow, XGBoost, LightGBM, and others.
Connects LangChain applications to Cohere's language models, providing chat, text embedding, retrieval-augmented generation, and reranking capabilities through a unified integration layer.
Install it if you are building LangChain applications that need Cohere's models or embeddings.
MLX-VLM runs vision language models and omni models (with audio and video support) for inference and fine-tuning on Mac hardware using MLX, with CLI, Python, FastAPI server, and Gradio UI interfaces.
pgmpy provides data structures and algorithms for causal discovery, causal inference, probabilistic inference, parameter learning, and model validation across Bayesian networks, DAGs, and structural equation models.
Arize is a Python client library for interacting with the Arize AI engineering platform, enabling you to log, trace, evaluate, and manage ML models, datasets, experiments, and LLM applications through a unified API.
However, it is tightly coupled to Arize's hosted service—you will need API credentials and an active account to use it meaningfully.
Model2Vec converts sentence transformers into small, fast static embedding models that generate vector representations of text for tasks like retrieval, classification, and clustering.
Streams large training datasets from cloud storage (AWS S3, GCS, Azure, OCI) as a PyTorch-compatible IterableDataset, handling images, text, video, and multimodal data with built-in compression and deterministic shuffling.
However, verify the license terms first (currently marked unclear in metadata), and confirm that the 18 runtime dependencies align with your environment.
Curated Transformers provides PyTorch implementations of state-of-the-art transformer models (BERT, Llama, Falcon, etc.) built from reusable components, with support for loading models from Hugging Face Hub and generation tasks.
Integrates Microsoft Foundry (Azure AI) capabilities into LangChain and LangGraph, providing access to chat models, agents, tools, vector stores, and tracing through a unified Python interface.
Install it if you are building LangChain or LangGraph applications on Azure infrastructure or need access to Azure AI services as tools.
DSPy is a framework for building and optimizing modular AI systems through declarative Python code rather than prompt engineering, supporting RAG pipelines and agent loops.
Install it if you're building anything beyond simple single-prompt use cases.
Converts machine learning models from multiple frameworks (TensorFlow, scikit-learn, Core ML, LightGBM, XGBoost, H2O, CatBoost, Spark ML, libsvm) into ONNX format for cross-platform inference.
Unified Python interface to call many LLM providers (OpenAI, Anthropic, Bedrock, Azure, VertexAI, Groq, etc.) in OpenAI format, plus an AI Gateway proxy server for centralized access with authentication, cost tracking, and routing.
pymoo implements single- and multi-objective optimization algorithms with visualization and decision-making tools for solving complex optimization problems in Python.
Scores automatic speech recognition or transcription output against a reference text, computing WER and CER with detailed alignment reports and diagnostic tools.
Install it if you regularly evaluate ASR or transcription systems.
Phoenix Evals provides composable building blocks for evaluating LLM applications, including pre-built evaluators for hallucination detection, relevance, toxicity, and other common assessment tasks.
However, the Elastic-2.0 license treatment is flagged as unclear—verify the license terms for your use case before production deployment.
NeMo Gym provides infrastructure for building, running, and scaling evaluation and training environments where agents interact with tasks, datasets, verifiers, and execution state to solve problems.
However, the strict Python 3.13.14+ requirement and heavy dependency footprint may conflict with existing projects.
Inspect Evals provides a repository of community-contributed LLM evaluations built on the Inspect AI framework, allowing developers to run standardized benchmarks against language models from multiple providers.
Integrates DeepSeek language models into LangChain applications, enabling use of DeepSeek's API through the LangChain framework.
Install it if you are already using LangChain and want to integrate DeepSeek models; it is a straightforward drop-in provider integration.
Provides helper functions to preprocess and integrate images, videos, and audio with Qwen multimodal language models for use in transformers pipelines.
Connects LangChain applications to Qdrant vector database for semantic search and retrieval-augmented generation workflows.
Integrates NVIDIA AI Foundation Models and chat endpoints into LangChain applications, providing access to models like Nemotron through the NVIDIA API Catalog or self-hosted NIM containers.
Enables PyTorch to run computations on Huawei Ascend NPU hardware, bridging PyTorch's tensor operations to Ascend AI Processors.
DyNet38 is a Python binding for DyNet, a C++ neural network library designed for efficient computation on CPU or GPU with support for dynamic network structures that vary per training instance.
However, the aging maintenance status (934 days since last release) and small repository star count (4) suggest limited active development and community; consider…
Integrates Tavily's web search and content extraction APIs into LangChain applications, enabling agents and tools to perform web searches and extract page content programmatically.
The main gotcha is the requirement for a Tavily API key and account setup, but that is external to the package itself.
Chainlit provides a web UI framework for building conversational AI applications in Python, handling message routing, async execution, and real-time communication between a backend and browser-based chat interface.
GLiNER is a lightweight framework for named entity recognition that can extract any entity type from text without labeled training data, and also supports relation extraction, PII detection, and token classification tasks.
Install it if you need flexible NER without task-specific training, or if you want to experiment with entity extraction on diverse entity types.
XGBoost CPU-only gradient boosting library for machine learning model training and prediction without GPU acceleration, designed for space-constrained environments.
Enables JAX to run numerical computations and machine learning workloads on NVIDIA GPUs with CUDA 12 support.
NumPyro is a probabilistic programming library that uses JAX for automatic differentiation and JIT compilation, enabling Bayesian inference with MCMC and variational inference algorithms on CPU, GPU, and TPU.
Not recommended if you require API stability or are new to probabilistic programming.
Integrates AssemblyAI speech-to-text into the LiveKit Agents framework for building real-time voice agents.