Explore AI, agents, blockchain and crypto through official sources, concrete tasks and practical guides.
Selected profiles with a clear purpose, supported use cases and source-based guides. Selection is not an execution test or a ranking of project quality.
An open-source, self-hosted AI engine and API layer for running models locally across language, vision, voice, image, and video tasks.
Builder toolsOpenAI Agents SDKAn open-source, MIT-licensed Python framework for building and coordinating multi-agent LLM workflows. It supports configurable agents, tool use, handoffs and tracing.
InfrastructureLight ProtocolAn open-source Rust protocol for ZK compression on Solana. Its stated purpose is to reduce the cost or footprint of Solana state using zero-knowledge compression.
Builder toolsVelora SDKAn open-source TypeScript SDK for calling the Velora API and accessing aggregated decentralized-exchange liquidity and rates.
Selected for a documented reader task. Source review does not claim an execution test.
LAMBDA is a large-language-model-based data analysis agent that turns natural-language questions about uploaded datasets into executable analysis workflows, visualizations, summaries and report or notebook outputs.
Guide: Run LAMBDA locallyA TypeScript example repository and setup guide for implementing a basic LangChain retrieval-augmented generation pipeline in Supabase Edge Functions, including package import-map configuration and function deployment.
Guide: Set up and deploy a Supabase Edge Function with LangChain importsLangsearch is a Python library for building retrieval-augmented question-answering applications over private or otherwise non-public data.
Guide: Install Langsearch and inspect its documented QA exampleLanguagecrunch is a Dockerized NLP server that exposes HTTP endpoints for sentence parsing, sentiment analysis, entity and relation extraction, sentence-type detection, coreference resolution, and word lookup.
Guide: Start the Dockerized NLP serviceLegal Summarizer is a Python and Streamlit application that processes legal documents, retrieves relevant passages through BM25 ranking or semantic search, and supplies the selected context to an LLM for structured summaries.
Guide: Set up the Legal Summarizer applicationLemniscate.pytorch is a Python and PyTorch implementation of non-parametric instance discrimination for unsupervised image feature learning.
Guide: Train an instance-discrimination model on ImageNetAn open-source Rust protocol for ZK compression on Solana. Its stated purpose is to reduce the cost or footprint of Solana state using zero-knowledge compression.
Guide: Set Up and Build the Light Protocol Development EnvironmentLLM Augmentation is a public repository of tutorial Jupyter notebooks for experimenting with custom-data retrieval and function calling in LLM applications.
Guide: Set up API keys for the tutorial notebooksLLM Corpus is an open-source Python project for building a retrieval-augmented knowledge base from documents.
Guide: Create a persistent Qdrant knowledge baseA Python question-answering chatbot application that uses LangChain retrieval-augmented generation, a Streamlit web interface and a GGML-quantized Llama 2 7B model to answer questions from external documents.
Guide: Understand the chatbot’s RAG answer flowLLM Recommender is a Python and Streamlit hotel-recommendation application that combines Redis review search with OpenAI language-model steps.
Guide: Run the app with Docker ComposeLLMVoX is a Python-based, autoregressive streaming text-to-speech model that converts language-model text output into speech and includes a streaming server for voice-based conversations.
Guide: Start LLMVoX voice chat with a configured LLM checkpointAn open-source, self-hosted AI engine and API layer for running models locally across language, vision, voice, image, and video tasks.
Guide: Start LocalAI in a CPU-only Docker containerLogfire Insight Relay is a callback-oriented telemetry bridge for Hugging Face Transformers training pipelines.
Guide: Configure a Trainer event relay to a local JSONL filelogfire-observer is a callback-style bridge that collects Hugging Face Transformers training events, enriches them with contextual information and emits structured telemetry to Logfire using OpenTelemetry primitives.
Guide: Connect the observer to a Transformers training runLongRoPE is a Python research implementation of a method for extending large language model context windows beyond two million tokens through positional-embedding interpolation and progressive extension.
Guide: Extend a model's context with the README exampleLTX-2 is a Python package for inference with an audio-video generative model and for training LoRAs that adapt it.
Guide: Set up and generate with the distilled pipelineMACE is an official research implementation of a fine-tuning method that erases multiple concepts from text-to-image diffusion models.
Guide: Set up the core MACE environment