Run multi-agent swarm intelligence simulations that forecast future outcomes from seed data including news events, policy drafts, and financial signals.
Developer
AI & Machine Learning
Open source AI & Machine Learning tools 36
Connect 352 AI providers through one OpenAI-compatible endpoint, with automatic fallback, 19 routing strategies, and 15-95% token compression. MIT licensed.
Run local LLM training and inference with an open-source Studio UI and optimized Python library.
Build AI agents that learn over time using Hindsight, a self-hosted MIT-licensed memory system with multi-strategy retrieval and knowledge consolidation.
Build computer vision applications with a model-agnostic Python toolkit for object detection, tracking, annotation, and dataset management.
Run LLMs, voice, vision, and image models locally with an OpenAI-compatible API. MIT licensed, no GPU required.
Connect all your AI providers in one self-hosted chat interface, with agents, MCP, code execution, and enterprise authentication built in.
VoxCPM
Replaces ElevenLabs
Generate multilingual speech with an Apache-2.0 TTS model for voice cloning, voice design, and self-hosted 48kHz audio output.
Vane
Replaces Perplexity AI
Search the web privately with an AI answering engine that runs on your own server, cites sources, and supports local LLMs via Ollama or cloud providers.
Laya
Replaces Cohere Classify
Analyze text with typed decisions in a single 33 ms forward pass. Supports 100+ languages with a built-in router, Apache-2.0 licensed.
Build role-specific AI plugins for knowledge workers in Claude, bundling skills, slash commands, and connectors for 11 job functions.
Lingbot Map
Replaces Pix4D
Run streaming 3D reconstruction from video or image folders with a self-hosted feed-forward model that estimates camera poses and dense point clouds.
LMCache
Replaces Amazon SageMaker
Optimize LLM inference with a self-hosted KV cache layer that reuses context across vLLM deployments.
camofox-browser
Replaces Bright Data Scraping Browser
Browse the real web from AI agents using C++ fingerprint spoofing that bypasses Cloudflare and bot detection without JavaScript shims.
accelerate
Replaces Amazon SageMaker
Run PyTorch training scripts across any distributed hardware configuration with minimal code changes, adding support for multi-GPU, TPU, and mixed precision.
Kev
Replaces Jev
Run Jev-compatible decision models on your own hardware or Modal, with four sizes from a 0.8B laptop model to a 27B datacenter GPU. Apache-2.0.
Bifrost
Replaces Portkey
Connect any AI SDK to 23+ LLM providers through a unified gateway, with automatic failover, semantic caching, MCP support, and 11µs overhead at 5,000 RPS.
weathernext
Replaces AccuWeather for Business
Generate accurate global weather and cyclone forecasts using AI models trained by Google DeepMind on atmospheric data, with self-hosted inference on TPU or GPU.
ODS
Replaces Amazon SageMaker
Transform your PC into a private AI server. Bundles local LLM inference, Open WebUI, agents, workflows, RAG, and voice in one Apache-2.0 stack.
OpenResearch
Replaces Conductor
Run parallel research agents with reproducible experiments on your own machine. MIT licensed, supports Claude Code, Codex, and OpenCode.
VoiceInk
Replaces DictaFlow
Run accurate voice-to-text on any Mac app using on-device AI models that require no subscription, no cloud, and no privacy compromises.
open-science
Replaces Claude Science
Run a local-first, model-agnostic AI research workbench with scientific agents, Python/R execution, 24 data connectors, and traceable provenance.
hyperresearch
Replaces ChatGPT Pro
Run a 16-step deep research agent inside Claude Code that collects sources into a persistent, searchable vault that compounds across sessions.
sie
Replaces Baseten
Serve all the open-model inference your AI agents need from a self-hosted Kubernetes cluster, replacing per-token APIs with compute-cost-only pricing.