screenshot-to-code
Self-Hosted AIConvert screenshots and designs into working code
Replaces Vercel v0
Best Lightweight Self-Hosted AI for self-hosting
The 31 best Lightweight self-hosted ai you can self-host, ranked by community traction.
Docling
Self-Hosted AIParse documents into AI-ready structured formats
Replaces Adobe Acrobat
GPT4Free
Self-Hosted AIUnified OpenAI-compatible API gateway for many models
Replaces OpenAI API
whisper.cpp
Self-Hosted AIFast C++ port of OpenAI Whisper speech recognition
Replaces Otter.ai, Google Speech-to-Text
Aider
Self-Hosted AIAI pair programming in your terminal
Replaces GitHub Copilot, Cursor
Fabric
Self-Hosted AIOpen-source framework for augmenting humans with AI
Replaces ChatGPT
Agno
Self-Hosted AILightweight framework for building multi-agent systems
Replaces LangChain
Marker
Self-Hosted AIConvert PDFs and documents to Markdown with AI
Replaces Adobe Acrobat
GraphRAG
Self-Hosted AIGraph-based retrieval-augmented generation system
Replaces OpenAI Assistants
Tabby
Self-Hosted AISelf-hosted AI coding assistant alternative to Copilot
Replaces GitHub Copilot
Chroma
Self-Hosted AIOpen-source embedding database for AI applications
Replaces Pinecone
promptfoo
Self-Hosted AITest, evaluate, and red-team LLM applications
Replaces LangSmith
faster-whisper
Self-Hosted AIReimplementation of Whisper for fast local transcription
Replaces Google Speech-to-Text
Rembg
Self-Hosted AIRemove image backgrounds automatically with AI
Replaces remove.bg
WhisperX
Self-Hosted AIFast Whisper transcription with word-level timestamps and diarization
Replaces Otter.ai, Rev
OpenUI
Self-Hosted AIDescribe UI with text and see it rendered live
Replaces Vercel v0
Guidance
Self-Hosted AIConstrained generation language for controlling LLMs
Replaces OpenAI API
Surya
Self-Hosted AIMultilingual document OCR and layout analysis toolkit
Replaces Google Cloud Vision, AWS Textract
DeepEval
Self-Hosted AIUnit testing framework for LLM outputs
Replaces LangSmith, Braintrust
Ragas
Self-Hosted AIEvaluation framework for retrieval-augmented generation
Replaces LangSmith
Outlines
Self-Hosted AIStructured text generation for language models
Replaces OpenAI Structured Outputs
LLMWare
Self-Hosted AIFramework for enterprise RAG with small language models
Replaces OpenAI Assistants
txtai
Self-Hosted AIAll-in-one embeddings database for semantic search and RAG
Replaces Pinecone
LanceDB
Self-Hosted AIEmbedded vector database for AI applications
Replaces Pinecone
AIChat
Self-Hosted AIAll-in-one LLM CLI tool with chat-REPL and API server
Replaces ChatGPT
Kokoro
Self-Hosted AICompact, high-quality open text-to-speech model
Replaces ElevenLabs, OpenAI TTS
ExLlamaV2
Self-Hosted AIFast inference library for quantized LLMs on consumer GPUs
Replaces OpenAI API
nano-graphrag
Self-Hosted AISimple, hackable GraphRAG implementation in Python
Replaces Microsoft GraphRAG
Whisper ASR Webservice
Self-Hosted AISelf-hosted speech-to-text API powered by OpenAI Whisper
Replaces Google Speech-to-Text, Otter.ai
ExLlama
Self-Hosted AIMemory-efficient inference library for quantized Llama models
Replaces OpenAI API
Wyoming faster-whisper
Self-Hosted AISpeech-to-text service for the Wyoming voice protocol
Replaces Google Speech-to-Text
No apps match these filters.
At a glance
| App | Deploy effort | Difficulty | Min. RAM | License | ★ Stars |
|---|---|---|---|---|---|
| screenshot-to-code | Under-an-hour setup | Easy | 1,024 MB | MIT | 80,328 |
| Docling | Under-an-hour setup | Medium | 4,096 MB | MIT | 68,836 |
| GPT4Free | ≈5-minute setup | Medium | 512 MB | GPL-3.0 | 66,839 |
| whisper.cpp | Read-the-docs project | Medium | 1,024 MB | MIT | 54,355 |
| Aider | Read-the-docs project | Easy | 512 MB | Apache-2.0 | 49,534 |
| Fabric | Read-the-docs project | Easy | 512 MB | MIT | 44,252 |
| Agno | Read-the-docs project | Medium | 1,024 MB | MPL-2.0 | 42,718 |
| Marker | Read-the-docs project | Medium | 4,096 MB | GPL-3.0 | 40,429 |
| GraphRAG | Read-the-docs project | Hard | 4,096 MB | MIT | 36,327 |
| Tabby | Under-an-hour setup | Medium | 8,192 MB | Apache-2.0 | 33,912 |
Quick picks
- Most adopted: screenshot-to-code (80,328 GitHub stars).
- Lightest footprint: AIChat (from 256 MB RAM, against 8,192 MB for the heaviest here).
- Fastest to deploy: GPT4Free (verified single-container compose).
- Licensing: 15 of 31 use MIT; check the rest individually if you plan to modify and redistribute.
Every option here is open-source and self-hostable, tagged Lightweight within the self-hosted ai category. Deploy effort is graded from published container recipes, not vendor claims — see the methodology.