Tag

supervision

Provides reusable computer-vision utilities for dataset loading/conversion, visualization/annotation of detections and segmentation, and connectors to popular detection frameworks—aimed at quick prototyping, dataset work, and visualization.

nanoGPT

Reproduces GPT-2 (124M) from scratch on OpenWebText in ~4 days on an 8xA100 node, with the whole stack kept to two ~300-line files: train.py for the loop and model.py for the architecture. A char-level Shakespeare run finishes in ~3 minutes on one GPU.

PaperQA2

Runs an agentic RAG loop over scientific papers: searches literature, gathers and re-ranks evidence chunks, then answers with in-text citations. Adds metadata-aware embeddings, retraction checks, and contradiction detection across full PDFs.

faster-whisper

Reimplements OpenAI's Whisper speech-to-text on the CTranslate2 inference engine, running up to 4x faster at the same accuracy while using less memory. Adds a batched pipeline, 8-bit quantization, VAD filtering, and word-level timestamps.

Introduction to Semantic Kernel

Maps your existing C#, Python, or Java functions into a form AI models can invoke, then translates model requests into real function calls and feeds results back. Model-agnostic middleware: swap in newer models without rewriting your app.

CAMEL-AI

Framework for building multi-agent systems where LLM agents take roles and converse to complete tasks via inception prompting, with no human in the loop after the initial brief. Used to auto-generate instruction data and run large-scale agent simulations.

llama.cpp

Runs large language models entirely in C/C++ with no external dependencies, using 1.5-to-8-bit integer quantization and CPU+GPU hybrid inference to fit models larger than available VRAM. Backs Ollama, LM Studio, and most local-inference tooling.

Outlines

Ensures LLM outputs match precise, schema-defined structures by enforcing Python types, Pydantic models, JSON schemas or grammars at generation time. Provider-agnostic integrations (OpenAI, transformers, vLLM, Ollama, etc.) and function-call style mapping reduce brittle post-processing and make structured generation reliable for production pipelines.

LLM

Run prompts against OpenAI, Claude, Gemini, and dozens of local or remote models from one terminal command, logging every prompt and response to SQLite. Plugins add new providers, tools, and embeddings; supports schema extraction and function calling.

PandasAI

Translates plain-English questions into pandas/SQL code over CSV, Parquet, and SQL databases, returning tables and charts. Combines LLMs with RAG and a semantic layer so non-coders query data; a Docker sandbox isolates generated code.

Ragas

Evaluates and tests LLM apps — RAG pipelines, agents, and workflows — using objective metrics that mix LLM-as-judge scoring with deterministic measures. Auto-generates synthetic test datasets and integrates with LangChain and tracing tools.

AI SDK (Vercel / vercel/ai)

Unifies access to OpenAI, Anthropic, Google and other LLM providers behind one TypeScript API — swap models by changing a string. Adds streaming UI hooks for React, Next.js, Svelte and Vue, plus a tool-calling loop for agentic workflows.