# reworkd/tarsier

Vision utilities for web interaction agents 👀

Repository: https://github.com/reworkd/tarsier
Canonical: https://ross.abutalabs.com/products/tarsier
Homepage: https://reworkd.ai
Language: Jupyter Notebook
License: MIT
License Family: permissive
Topics: ocr, playwright, selenium, webscraping, pypi-package, gpt4v, llms, python
Last push: 2024-11-25T07:50:59+00:00

## Health v2 (maintenance only)
Score: 17/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 73
- inputs: {"age_days": 1028, "days_push": 646, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1761, forks 121 (observed 2026-08-28T04:05:32.625437+00:00)

## What it is
Tarsier is a Python library providing vision utilities for LLM-driven web interaction agents. It visually tags interactable page elements with bracketed IDs and converts page screenshots into whitespace-structured text via OCR so even text-only LLMs can perceive and act on webpages.

## Use cases
- let an LLM agent click buttons and fill forms on a webpage
- convert a page screenshot into text an LLM can understand
- map LLM responses like CLICK [23] back to web elements
- give a text-only LLM awareness of a page's visual structure
- build browser automation agents with Playwright or Selenium
- tag visible links, buttons, and inputs for web agents

## When to choose
- you are building an LLM web agent and need page perception
- you want to feed webpage structure to text-only models instead of vision models
- you use Playwright or Selenium and need element-to-ID mappings

## When to avoid
- you need a general-purpose OCR tool unrelated to web agents
- you want a maintained product-backed solution (Reworkd is sunsetting its product)
- you need non-Python tooling

## Facets
- artifact type: library
- maturity: maintenance
- function: ocr, computer-vision, web-scraping, llm-inference, agent-framework
- domain: artificial-intelligence, web-development, crawlers
- platform: python, cross-platform
- tags: web-agents, browser-automation, playwright, selenium, gpt4v, page-perception, element-tagging, ai-agents, natural-language-processing

## Member repositories
- reworkd/tarsier (main) score 17

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:32.625437+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 2026-08-30T03:27:12.491002+00:00, confidence not recorded.
  - readme: https://github.com/reworkd/tarsier (fetched 2026-08-28T04:05:32.625437+00:00, sha 343c2a529310)
  - homepage: https://reworkd.ai (fetched 2026-08-29T11:05:35.996355+00:00, sha 5b857cca86fb)
  - site_page: https://www.reworkd.ai/docs (fetched 2026-08-29T11:05:36.009540+00:00, sha 382f64c93743)
  - site_page: https://www.reworkd.ai/pricing (fetched 2026-08-29T11:05:36.005794+00:00, sha 32614897157e)
- Data as of 2026-08-30T08:39:29.467469+00:00.