Build an AI app
The open-source pieces for an AI app: retrieval, vector search, structured outputs, prompt programs, serving, fine-tuning and tracing, for $0.
7 free toolsTotal: $0
Tools in this stack
01Vector search
pgvector
Open-source Postgres extension for vector similarity search, keeping embeddings alongside the rest of your data.
Free forever, Free Postgres extension for vector similarity search
$0Try02Retrieval (RAG)
LightRAG
Open-source RAG framework that builds a knowledge graph from your documents for graph-plus-vector retrieval.
Free forever, Open source (MIT): free to self-host with your own LLM
$0Try03Structured outputs
Instructor
Open-source library that gets typed, validated structured data from LLMs, with automatic retries, for 15+ providers.
Free forever, Open source (MIT): free library; you bring your own LLM provider or local model
$0Try04Prompt programs
DSPy
Open-source Python framework for building LLM programs from typed signatures and modules, with prompt optimizers.
Free forever, Open source (MIT): free framework; you pay your LLM provider for calls
$0Try05Serve models
SGLang
Open-source engine that serves language, vision-language and diffusion models on your own hardware.
Free forever, Open source (Apache 2.0): free to self-host on your own hardware
$0Try06Fine-tuning
Unsloth
Open-source desktop app and library to run and fine-tune LLMs, image and audio models on your own computer.
Free forever, Open source (Apache 2.0, with the Studio UI under AGPL-3.0): free to run locally
$0Try07Tracing & costs
Pydantic Logfire
Observability for AI apps and agents: traces LLM calls, tool calls and costs next to the rest of your app's telemetry.
Free forever, 10M telemetry records per month · 3 projects
$0Try
Total
$0
Every tool here has a free plan or free credits we checked on its pricing page; the dates are on each tool page. Free plans have limits, and vendors change them, so confirm the current terms before you rely on one. How we check
