# StarTrail-org/PixelRAG

The end of web parsing. The beginning of scalable pixel-native search. link: https://pixelrag.ai/

Repository: https://github.com/StarTrail-org/PixelRAG
Canonical: https://ross.abutalabs.com/products/pixelrag
Homepage: https://arxiv.org/pdf/2606.28344
Language: Python
License: Apache-2.0
License Family: permissive
Topics: agent, ai, memory, multimodal, rag, search, searchengine, vision, vlm
Last push: 2026-07-31T08:52:00+00:00

## Health v2 (maintenance only)
Score: 76/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 95, release rhythm 93, longevity 6
- inputs: {"age_days": 96, "days_push": 33, "days_rel": 48, "gap_med": 21, "n_releases_24m": 4}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 9742, forks 828 (observed 2026-08-28T04:10:37.031619+00:00)

## What it is
PixelRAG is a Python library and hosted service for visual retrieval-augmented generation: it renders web pages and documents into screenshot tiles, embeds them with vision-language models, and searches a visual index instead of parsed text. It ships a CLI (pixelshot) and a hosted API over a pre-built 8.28M-page Wikipedia visual index.

## Use cases
- search documents by how they look instead of parsing text
- build a RAG pipeline over web page screenshots
- query a Wikipedia index with an image instead of text
- render web pages to screenshot tiles for indexing
- avoid brittle HTML-to-text extraction for retrieval
- give LLM agents visual memory of web pages

## When to choose
- text extraction from complex or visually structured pages loses information
- you want multimodal (text or image) queries over a large document corpus
- you need a ready-made hosted visual search index with no setup

## When to avoid
- you need a fully self-hosted, offline solution with no external API dependency
- your corpus is plain text where traditional text retrieval is cheaper and sufficient
- you cannot afford the GPU/compute cost of rendering and embedding screenshots at scale

## Facets
- artifact type: library
- maturity: active
- function: rag, search-engine, image-processing, machine-learning, llm-inference, web-scraping
- domain: large-language-models, computer-vision, artificial-intelligence
- platform: python, cli, cloud
- tags: visual-search, screenshots, vision-language-models, embeddings, faiss, multimodal-retrieval, pixel-native, retrieval-augmented-generation, search, web-server

## Member repositories
- StarTrail-org/PixelRAG (main) score 76

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:37.031619+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:21:12.363693+00:00, confidence not recorded.
  - readme: https://github.com/StarTrail-org/PixelRAG (fetched 2026-08-28T04:10:37.031619+00:00, sha cd3fc764a539)
  - registry_pypi: https://pypi.org/pypi/pixelrag/json (fetched 2026-08-29T08:20:46.559353+00:00, sha 06c25311c8b6)
- Data as of 2026-08-30T08:39:29.467469+00:00.