openbrowse 1.8.1__tar.gz

openbrowse-1.8.1/.gitignore

@@ -0,0 +1,37 @@
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.egg-info/
+dist/
+build/
+*.egg
+
+# Virtual environment
+.venv/
+venv/
+ENV/
+
+# Environment variables
+.env
+
+# Data directory (keep .gitkeep)
+data/*
+!data/.gitkeep
+
+# IDE
+.idea/
+.vscode/
+*.swp
+*.swo
+
+# OS
+.DS_Store
+Thumbs.db
+.worktrees/
+.claude
+.playwright-mcp/
+uv.lock
+
+# Internal planning notes, never tracked
+docs/

openbrowse-1.8.1/CITATION.cff

@@ -0,0 +1,23 @@
+cff-version: 1.2.0
+message: "If you use OpenBrowse in your research or project, please cite it as below."
+type: software
+title: "OpenBrowse: Self-hosted AI browser agents"
+version: 1.8.1
+date-released: 2026-08-20
+license: MIT
+doi: "10.5281/zenodo.21986248"
+authors:
+  - family-names: "Johnston Kurilov"
+    given-names: "Lucas"
+    orcid: "https://orcid.org/0009-0004-3743-2349"
+repository-code: "https://github.com/lujstn/openbrowse"
+url: "https://openbrowse.co"
+abstract: "OpenBrowse is an open-source, self-hosted alternative to Browser Use Cloud: AI browser agents that run on your own hardware, driven through the v3 REST API, with schema-validated structured output, anti-hallucination grounding guards, and a live visual dashboard."
+keywords:
+  - artificial intelligence
+  - ai agents
+  - browser automation
+  - browser agents
+  - structured data extraction
+  - web scraping
+  - self-hosted

openbrowse-1.8.1/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Lucas Johnston Kurilov
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

openbrowse-1.8.1/PKG-INFO

@@ -0,0 +1,254 @@
+Metadata-Version: 2.5
+Name: openbrowse
+Version: 1.8.1
+Summary: Self-hosted AI browser automation server with a live dashboard, built for small machines like the Raspberry Pi
+Project-URL: Homepage, https://openbrowse.co
+Project-URL: Repository, https://github.com/lujstn/openbrowse
+Project-URL: Issues, https://github.com/lujstn/openbrowse/issues
+Author: Lucas Johnston Kurilov
+License: MIT License
+        
+        Copyright (c) 2026 Lucas Johnston Kurilov
+        
+        Permission is hereby granted, free of charge, to any person obtaining a copy
+        of this software and associated documentation files (the "Software"), to deal
+        in the Software without restriction, including without limitation the rights
+        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+        copies of the Software, and to permit persons to whom the Software is
+        furnished to do so, subject to the following conditions:
+        
+        The above copyright notice and this permission notice shall be included in all
+        copies or substantial portions of the Software.
+        
+        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+        SOFTWARE.
+License-File: LICENSE
+Keywords: ai-agent,automation,browser-use,raspberry-pi,self-hosted
+Classifier: Development Status :: 5 - Production/Stable
+Classifier: Environment :: Web Environment
+Classifier: Framework :: FastAPI
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: POSIX :: Linux
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Internet :: WWW/HTTP
+Requires-Python: >=3.11
+Requires-Dist: aiosqlite>=0.20.0
+Requires-Dist: browser-use<0.14,>=0.13.7
+Requires-Dist: cloakbrowser>=0.3.0
+Requires-Dist: fastapi>=0.115.0
+Requires-Dist: httpx>=0.28.0
+Requires-Dist: jinja2>=3.1.0
+Requires-Dist: pydantic>=2.0.0
+Requires-Dist: python-dotenv>=1.0.0
+Requires-Dist: python-multipart>=0.0.9
+Requires-Dist: sse-starlette>=2.0.0
+Requires-Dist: uvicorn[standard]>=0.34.0
+Requires-Dist: websockets>=12.0
+Provides-Extra: dev
+Requires-Dist: httpx; extra == 'dev'
+Requires-Dist: pytest; extra == 'dev'
+Requires-Dist: pytest-asyncio; extra == 'dev'
+Description-Content-Type: text/markdown
+
+<p align="center">
+  <picture>
+    <source media="(prefers-color-scheme: dark)" srcset=".github/assets/openbrowse-dark.gif">
+    <img src=".github/assets/openbrowse-light.gif" alt="OpenBrowse" width="432">
+  </picture>
+</p>
+
+# OpenBrowse
+
+**The open-source Browser Use Cloud alternative.** Self-host AI browser agents on a Raspberry Pi or any VPS, drive them through the same v3 REST API the `browser-use-sdk` already speaks, and watch every run live in a real browser. Built on top of the [Browser Use](https://github.com/browser-use/browser-use) SDK. It's cheaper, faster, and more reliable than BU Cloud.
+
+[openbrowse.co](https://openbrowse.co)
+
+<div align="left">
+  <a href="https://buildin.london"><img src="https://buildin.london/badge.svg" alt="Built in London" style="width: 200px;"></a>
+</div>
+
+---
+
+## Benchmarks
+
+Given the same real-world extraction task (a careers page with 14 records behind an embedded, cross-origin board, requiring full schema output):
+
+| Runtime | Model | Reasoning | Time | Tokens | LLM cost | Records |
+| --- | --- | --- | ---: | ---: | ---: | --- |
+| BU Cloud | claude-sonnet-5 | high | 2m 36s | 859k | $0.78 | 14/14<sup>1</sup> |
+| **OpenBrowse** | **gpt-5.6-terra** | **none** | **1m 47s** | **202k** | **$0.24** | **14/14** |
+| **OpenBrowse** | **claude-sonnet-5** | **high** | 4m 02s | **242k** | **$0.40** | **14/14** |
+
+<sub><i><sup>1</sup> Extracted successfully, though some returned fields (e.g. job seniority) were hallucinated when not shown on the page.</i></sub>
+
+## Why OpenBrowse over BU Cloud?
+
+| | BU Cloud | OpenBrowse |
+| --- | --- | --- |
+| Hosting | Managed, per-task pricing | Your hardware, you pay only LLM tokens |
+| How it works | Code-first: the agent scripts its way through pages | Visual-first: the agent opens real tabs you can watch live, like a human working |
+| Bulk page reads | One page at a time | `read_pages` opens whole listings in parallel tab waves, one step |
+| Structured output | Schema-validated | Schema-validated, plus a live answer store with a completeness gate the agent must pass before finishing |
+| Anti-hallucination | Often fills fields the page never shows | On-screen data first, enriched from the page's own structured data (JSON-LD, APIs), never guessed: values without evidence are refused at the store boundary |
+| Profiles | Cloud profiles | Import your BU Cloud profiles (cookies and localStorage) with one command |
+| Live view | Replay | Real-time VNC of the actual browser, a step feed with the model's reasoning, and an IDE-style code tab that streams the agent's sandbox scripts live as they're written |
+| API | v3 REST | The same v3 REST surface: point `browser-use-sdk` at your box and change nothing but `baseUrl` and `apiKey` |
+
+### ⚡ See it in action
+
+Here's a snippet of our benchmark run, with `claude-opus-5` taking agentic actions across parallel tabs while its code streams into a live IDE.
+
+<https://github.com/user-attachments/assets/c1330d77-67b6-4a7d-bd43-7cdfa230b9d1>
+
+## Model providers
+
+### Recommended models
+
+1. **For most use cases**, `gpt-5.6-terra { "reasoningEffort": "none" }`, `gpt-5.6-sol { "reasoningEffort": "none" }` and `claude-sonnet-5 { "reasoningEffort": "high" }` all strike a great balance of reliability, accuracy, and cost.
+
+2. **For intense workflows**, use `claude-opus-5 { "reasoningEffort": "medium" }` or `gpt-5.6-sol { "reasoningEffort": "none" }` — both are great options, but watch out for token burn.
+
+3. **On a budget?** Use `gpt-5.6-luna { "reasoningEffort": "max" }` with a tightly focused prompt. It might take a while, and it's more prone to hallucinations (especially with broad prompts), but the actual extractions are still great quality.
+
+### Comparisons
+
+The same real-world extraction task (a careers page with 14 records behind an embedded, cross-origin board, full schema output) run against BU Cloud and against OpenBrowse on a Raspberry Pi 5 (16GB) without concurrency, ordered cheapest to most expensive:
+
+| Runtime | Model | Reasoning | Steps | Time | Tokens | LLM cost | Records |
+| --- | --- | --- | --- | ---: | ---: | ---: | --- |
+| OpenBrowse | **gpt-5.6-luna** | **max** | 36 | 17m 03s | 1.08M | **$0.22** | 14/14 |
+| OpenBrowse | **gpt-5.6-terra** | **none** | 11 | **1m 47s** | **202k** | **$0.24** | 14/14 |
+| OpenBrowse | **claude-sonnet-5** | **high** | 10 | 4m 02s | **242k** | **$0.40** | 14/14 |
+| OpenBrowse | **gpt-5.6-sol** | **none** | **8** | **2m 03s** | **136k** | **$0.41** | 14/14 |
+| OpenBrowse | **claude-sonnet-5** | **none** | **9** | 5m 18s | **237k** | **$0.51** | 14/14 |
+| OpenBrowse | **gpt-5.6-terra** | **high** | 17 | 5m 05s | **434k** | **$0.66** | 14/14 |
+| BU Cloud | claude-sonnet-5 | high | 10 | 2m 36s | 859k | $0.78 | 14/14<sup>1</sup> |
+| OpenBrowse | gpt-5.6-sol | medium | 16 | 5m 24s | **339k** | $1.12 | 14/14 |
+| OpenBrowse | claude-opus-5 | medium | 15 | 3m 56s | **398k** | $1.32 | 14/14 |
+| OpenBrowse | claude-opus-5 | none | 17 | 4m 53s | **480k** | $1.62 | 14/14 |
+
+<sub><i><sup>1</sup> Extracted successfully, though some returned fields (e.g. job seniority) were hallucinated when not shown on the page.</i></sub>
+
+OpenAI and Anthropic models are generally at their best at the opposite ends of the reasoning dial. For example, OpenAI's GPT-5.6-Terra performs better with less reasoning, spending less time planning ahead and more time reacting to the page in front of it, while Anthropic's 5-series Claude models lean towards rabbit holes and need reasoning time to refocus on the goal.
+
+### What is "thinking" in OpenBrowse?
+
+OpenBrowse separates two kinds of reasoning. **Browser thinking** is our way of describing how the platform works in "steps" (the 👁️ see / 🛝 plan / ➡️ next / 💭 thinking cards in the live feed), so it can't be disabled.
+
+**Model reasoning** is different: it's the Chain-of-Thought reasoning provided by LLM providers (e.g. Anthropic's extended thinking, OpenAI's reasoning effort), and can be controlled per session by changing `reasoningEffort` in the API. Values are validated per model at runtime. Models will have different default reasoning levels depending on their provider, so it's a good idea to set this value explicitly.
+
+### All supported models
+
+- OpenAI: `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-5.6-luna`
+- Anthropic: `claude-mythos-5`, `claude-fable-5`, `claude-opus-5`, `claude-sonnet-5`, `claude-opus-4.8`, `claude-opus-4.8[1m]`, `claude-opus-4.7`, `claude-opus-4.7[1m]`, `claude-opus-4.6`, `claude-opus-4.6[1m]`, `claude-sonnet-4.6`, `claude-sonnet-4.6[1m]`
+- Google: ⚠️ Coming soon
+
+## Quick start
+
+```bash
+curl -LsSf https://astral.sh/uv/install.sh | sh
+uv tool install openbrowse
+openbrowse start
+```
+
+`openbrowse start` registers OpenBrowse as a systemd service, so it is running now **and starts automatically on every boot from here on** (`openbrowse stop --disable` undoes that). On a machine without systemd it falls back to running in the foreground, as does `openbrowse serve` anywhere.
+
+To run from source instead, clone the repo and use uv directly:
+
+```bash
+git clone git@github.com:lujstn/openbrowse.git
+cd openbrowse
+uv sync
+uv run openbrowse serve
+```
+
+Open `http://<your-host>:8420` in a browser. A fresh install serves a one-time **setup screen** that generates your API bearer key, takes your Anthropic / OpenAI / CapSolver keys, sets your dashboard password and concurrency limit, and writes `.env` for you.
+
+Then from any `browser-use-sdk` client:
+
+```ts
+import { BrowserUse } from "browser-use-sdk";
+
+const client = new BrowserUse({
+  apiKey: process.env.OPENBROWSE_API_KEY,
+  baseUrl: "http://<your-host>:8420/v3",
+});
+
+const task = await client.tasks.create({
+  task: "Find every product on this page and return the structured list.",
+  model: "claude-sonnet-5",
+  outputSchema: mySchema,
+});
+```
+
+Full installation (Raspberry Pi system packages, Xvfb + VNC live view, systemd service): see [GETTING_STARTED.md](GETTING_STARTED.md).
+
+## Updating
+
+The server checks PyPI for new releases in the background and shows an **Update available** badge in the dashboard when one exists; installing it is one click on the Settings page (the server restarts itself afterwards). From a shell, `openbrowse check-update` and `openbrowse update` do the same. Set `UPDATE_CHECK_HOURS=0` in `.env` to disable the background check.
+
+## Exposing it to the web
+
+OpenBrowse runs on plain Python + Chromium and can be easily port forwarded. To reach it from outside that box without opening ports, put it behind [Tailscale](https://tailscale.com/):
+
+```bash
+# private access from your own devices
+tailscale up
+
+# or expose the API publicly over TLS with Tailscale Funnel
+sudo tailscale funnel --bg 8420
+```
+
+## Features
+
+- **v3-compatible REST API**: sessions, structured output schemas, cost caps, live URLs.
+- **Visual, tab-based browsing**: parallel foreground tab waves for bulk reads; a code tab shows when the agent runs a script; everything visible over VNC.
+- **Schema answer store**: every write validated live against your JSON Schema, coverage tracked per field, a completeness gate before `done`, and mark-absent semantics for data a site genuinely does not publish.
+- **Grounding guards**: shell-read detection with automatic in-frame retry, evidence-checked enum writes, URL fields validated as absolute http(s) links at the store boundary, honest failure over invented data.
+- **Profile import**: bring BU Cloud profiles (cookies + localStorage) via CLI or the dashboard.
+- **Dashboard**: live session feed with model reasoning, per-step costs, JSON export (full / steps / output-only), profile management.
+- **CAPTCHA solving**: optional CapSolver integration.
+- **Multi-provider**: Anthropic and OpenAI models behind one API, with per-provider repair layers for each family's failure modes.
+
+## Citation
+
+This project is licensed under the MIT License. If you use OpenBrowse as part of your research or project, please cite:
+
+```bibtex
+@software{openbrowse2026,
+  author = {Johnston Kurilov, Lucas},
+  title = {{OpenBrowse}: Self-hosted AI browser agents},
+  year = {2026},
+  publisher = {Zenodo},
+  doi = {10.5281/zenodo.21986248},
+  url = {https://github.com/lujstn/openbrowse}
+}
+```
+
+[![DOI](https://zenodo.org/badge/1210478161.svg)](https://doi.org/10.5281/zenodo.21986248)
+
+<br><br>
+
+<table align="center">
+  <tr>
+    <th colspan="2">built with <з by @lujstn</th>
+  </tr>
+  <tr>
+    <td><img src=".github/assets/IMG_8874.jpg" alt="@lujstn" width="400"></td>
+    <td valign="middle">
+      <a href="https://x.com/intent/user?screen_name=lujstn"><img src="https://img.shields.io/twitter/follow/lujstn?style=social" alt="Twitter"></a>
+      <br>
+      <a href="https://www.instagram.com/lujstn/"><img src="https://img.shields.io/badge/Instagram-Follow-E4405F?style=social&logo=instagram" alt="Instagram"></a>
+      <br>
+      <a href="https://www.tiktok.com/@lujstn"><img src="https://img.shields.io/badge/TikTok-000000?style=flat&logo=tiktok&logoColor=white" alt="TikTok"></a>
+      <br>
+      <a href="https://lujstn.com"><img src="https://img.shields.io/badge/%F0%9F%94%97_lujstn.com-1a1a1a" alt="lujstn.com"></a>
+    </td>
+  </tr>
+</table>

openbrowse-1.8.1/README.md

@@ -0,0 +1,196 @@
+<p align="center">
+  <picture>
+    <source media="(prefers-color-scheme: dark)" srcset=".github/assets/openbrowse-dark.gif">
+    <img src=".github/assets/openbrowse-light.gif" alt="OpenBrowse" width="432">
+  </picture>
+</p>
+
+# OpenBrowse
+
+**The open-source Browser Use Cloud alternative.** Self-host AI browser agents on a Raspberry Pi or any VPS, drive them through the same v3 REST API the `browser-use-sdk` already speaks, and watch every run live in a real browser. Built on top of the [Browser Use](https://github.com/browser-use/browser-use) SDK. It's cheaper, faster, and more reliable than BU Cloud.
+
+[openbrowse.co](https://openbrowse.co)
+
+<div align="left">
+  <a href="https://buildin.london"><img src="https://buildin.london/badge.svg" alt="Built in London" style="width: 200px;"></a>
+</div>
+
+---
+
+## Benchmarks
+
+Given the same real-world extraction task (a careers page with 14 records behind an embedded, cross-origin board, requiring full schema output):
+
+| Runtime | Model | Reasoning | Time | Tokens | LLM cost | Records |
+| --- | --- | --- | ---: | ---: | ---: | --- |
+| BU Cloud | claude-sonnet-5 | high | 2m 36s | 859k | $0.78 | 14/14<sup>1</sup> |
+| **OpenBrowse** | **gpt-5.6-terra** | **none** | **1m 47s** | **202k** | **$0.24** | **14/14** |
+| **OpenBrowse** | **claude-sonnet-5** | **high** | 4m 02s | **242k** | **$0.40** | **14/14** |
+
+<sub><i><sup>1</sup> Extracted successfully, though some returned fields (e.g. job seniority) were hallucinated when not shown on the page.</i></sub>
+
+## Why OpenBrowse over BU Cloud?
+
+| | BU Cloud | OpenBrowse |
+| --- | --- | --- |
+| Hosting | Managed, per-task pricing | Your hardware, you pay only LLM tokens |
+| How it works | Code-first: the agent scripts its way through pages | Visual-first: the agent opens real tabs you can watch live, like a human working |
+| Bulk page reads | One page at a time | `read_pages` opens whole listings in parallel tab waves, one step |
+| Structured output | Schema-validated | Schema-validated, plus a live answer store with a completeness gate the agent must pass before finishing |
+| Anti-hallucination | Often fills fields the page never shows | On-screen data first, enriched from the page's own structured data (JSON-LD, APIs), never guessed: values without evidence are refused at the store boundary |
+| Profiles | Cloud profiles | Import your BU Cloud profiles (cookies and localStorage) with one command |
+| Live view | Replay | Real-time VNC of the actual browser, a step feed with the model's reasoning, and an IDE-style code tab that streams the agent's sandbox scripts live as they're written |
+| API | v3 REST | The same v3 REST surface: point `browser-use-sdk` at your box and change nothing but `baseUrl` and `apiKey` |
+
+### ⚡ See it in action
+
+Here's a snippet of our benchmark run, with `claude-opus-5` taking agentic actions across parallel tabs while its code streams into a live IDE.
+
+<https://github.com/user-attachments/assets/c1330d77-67b6-4a7d-bd43-7cdfa230b9d1>
+
+## Model providers
+
+### Recommended models
+
+1. **For most use cases**, `gpt-5.6-terra { "reasoningEffort": "none" }`, `gpt-5.6-sol { "reasoningEffort": "none" }` and `claude-sonnet-5 { "reasoningEffort": "high" }` all strike a great balance of reliability, accuracy, and cost.
+
+2. **For intense workflows**, use `claude-opus-5 { "reasoningEffort": "medium" }` or `gpt-5.6-sol { "reasoningEffort": "none" }` — both are great options, but watch out for token burn.
+
+3. **On a budget?** Use `gpt-5.6-luna { "reasoningEffort": "max" }` with a tightly focused prompt. It might take a while, and it's more prone to hallucinations (especially with broad prompts), but the actual extractions are still great quality.
+
+### Comparisons
+
+The same real-world extraction task (a careers page with 14 records behind an embedded, cross-origin board, full schema output) run against BU Cloud and against OpenBrowse on a Raspberry Pi 5 (16GB) without concurrency, ordered cheapest to most expensive:
+
+| Runtime | Model | Reasoning | Steps | Time | Tokens | LLM cost | Records |
+| --- | --- | --- | --- | ---: | ---: | ---: | --- |
+| OpenBrowse | **gpt-5.6-luna** | **max** | 36 | 17m 03s | 1.08M | **$0.22** | 14/14 |
+| OpenBrowse | **gpt-5.6-terra** | **none** | 11 | **1m 47s** | **202k** | **$0.24** | 14/14 |
+| OpenBrowse | **claude-sonnet-5** | **high** | 10 | 4m 02s | **242k** | **$0.40** | 14/14 |
+| OpenBrowse | **gpt-5.6-sol** | **none** | **8** | **2m 03s** | **136k** | **$0.41** | 14/14 |
+| OpenBrowse | **claude-sonnet-5** | **none** | **9** | 5m 18s | **237k** | **$0.51** | 14/14 |
+| OpenBrowse | **gpt-5.6-terra** | **high** | 17 | 5m 05s | **434k** | **$0.66** | 14/14 |
+| BU Cloud | claude-sonnet-5 | high | 10 | 2m 36s | 859k | $0.78 | 14/14<sup>1</sup> |
+| OpenBrowse | gpt-5.6-sol | medium | 16 | 5m 24s | **339k** | $1.12 | 14/14 |
+| OpenBrowse | claude-opus-5 | medium | 15 | 3m 56s | **398k** | $1.32 | 14/14 |
+| OpenBrowse | claude-opus-5 | none | 17 | 4m 53s | **480k** | $1.62 | 14/14 |
+
+<sub><i><sup>1</sup> Extracted successfully, though some returned fields (e.g. job seniority) were hallucinated when not shown on the page.</i></sub>
+
+OpenAI and Anthropic models are generally at their best at the opposite ends of the reasoning dial. For example, OpenAI's GPT-5.6-Terra performs better with less reasoning, spending less time planning ahead and more time reacting to the page in front of it, while Anthropic's 5-series Claude models lean towards rabbit holes and need reasoning time to refocus on the goal.
+
+### What is "thinking" in OpenBrowse?
+
+OpenBrowse separates two kinds of reasoning. **Browser thinking** is our way of describing how the platform works in "steps" (the 👁️ see / 🛝 plan / ➡️ next / 💭 thinking cards in the live feed), so it can't be disabled.
+
+**Model reasoning** is different: it's the Chain-of-Thought reasoning provided by LLM providers (e.g. Anthropic's extended thinking, OpenAI's reasoning effort), and can be controlled per session by changing `reasoningEffort` in the API. Values are validated per model at runtime. Models will have different default reasoning levels depending on their provider, so it's a good idea to set this value explicitly.
+
+### All supported models
+
+- OpenAI: `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-5.6-luna`
+- Anthropic: `claude-mythos-5`, `claude-fable-5`, `claude-opus-5`, `claude-sonnet-5`, `claude-opus-4.8`, `claude-opus-4.8[1m]`, `claude-opus-4.7`, `claude-opus-4.7[1m]`, `claude-opus-4.6`, `claude-opus-4.6[1m]`, `claude-sonnet-4.6`, `claude-sonnet-4.6[1m]`
+- Google: ⚠️ Coming soon
+
+## Quick start
+
+```bash
+curl -LsSf https://astral.sh/uv/install.sh | sh
+uv tool install openbrowse
+openbrowse start
+```
+
+`openbrowse start` registers OpenBrowse as a systemd service, so it is running now **and starts automatically on every boot from here on** (`openbrowse stop --disable` undoes that). On a machine without systemd it falls back to running in the foreground, as does `openbrowse serve` anywhere.
+
+To run from source instead, clone the repo and use uv directly:
+
+```bash
+git clone git@github.com:lujstn/openbrowse.git
+cd openbrowse
+uv sync
+uv run openbrowse serve
+```
+
+Open `http://<your-host>:8420` in a browser. A fresh install serves a one-time **setup screen** that generates your API bearer key, takes your Anthropic / OpenAI / CapSolver keys, sets your dashboard password and concurrency limit, and writes `.env` for you.
+
+Then from any `browser-use-sdk` client:
+
+```ts
+import { BrowserUse } from "browser-use-sdk";
+
+const client = new BrowserUse({
+  apiKey: process.env.OPENBROWSE_API_KEY,
+  baseUrl: "http://<your-host>:8420/v3",
+});
+
+const task = await client.tasks.create({
+  task: "Find every product on this page and return the structured list.",
+  model: "claude-sonnet-5",
+  outputSchema: mySchema,
+});
+```
+
+Full installation (Raspberry Pi system packages, Xvfb + VNC live view, systemd service): see [GETTING_STARTED.md](GETTING_STARTED.md).
+
+## Updating
+
+The server checks PyPI for new releases in the background and shows an **Update available** badge in the dashboard when one exists; installing it is one click on the Settings page (the server restarts itself afterwards). From a shell, `openbrowse check-update` and `openbrowse update` do the same. Set `UPDATE_CHECK_HOURS=0` in `.env` to disable the background check.
+
+## Exposing it to the web
+
+OpenBrowse runs on plain Python + Chromium and can be easily port forwarded. To reach it from outside that box without opening ports, put it behind [Tailscale](https://tailscale.com/):
+
+```bash
+# private access from your own devices
+tailscale up
+
+# or expose the API publicly over TLS with Tailscale Funnel
+sudo tailscale funnel --bg 8420
+```
+
+## Features
+
+- **v3-compatible REST API**: sessions, structured output schemas, cost caps, live URLs.
+- **Visual, tab-based browsing**: parallel foreground tab waves for bulk reads; a code tab shows when the agent runs a script; everything visible over VNC.
+- **Schema answer store**: every write validated live against your JSON Schema, coverage tracked per field, a completeness gate before `done`, and mark-absent semantics for data a site genuinely does not publish.
+- **Grounding guards**: shell-read detection with automatic in-frame retry, evidence-checked enum writes, URL fields validated as absolute http(s) links at the store boundary, honest failure over invented data.
+- **Profile import**: bring BU Cloud profiles (cookies + localStorage) via CLI or the dashboard.
+- **Dashboard**: live session feed with model reasoning, per-step costs, JSON export (full / steps / output-only), profile management.
+- **CAPTCHA solving**: optional CapSolver integration.
+- **Multi-provider**: Anthropic and OpenAI models behind one API, with per-provider repair layers for each family's failure modes.
+
+## Citation
+
+This project is licensed under the MIT License. If you use OpenBrowse as part of your research or project, please cite:
+
+```bibtex
+@software{openbrowse2026,
+  author = {Johnston Kurilov, Lucas},
+  title = {{OpenBrowse}: Self-hosted AI browser agents},
+  year = {2026},
+  publisher = {Zenodo},
+  doi = {10.5281/zenodo.21986248},
+  url = {https://github.com/lujstn/openbrowse}
+}
+```
+
+[![DOI](https://zenodo.org/badge/1210478161.svg)](https://doi.org/10.5281/zenodo.21986248)
+
+<br><br>
+
+<table align="center">
+  <tr>
+    <th colspan="2">built with <з by @lujstn</th>
+  </tr>
+  <tr>
+    <td><img src=".github/assets/IMG_8874.jpg" alt="@lujstn" width="400"></td>
+    <td valign="middle">
+      <a href="https://x.com/intent/user?screen_name=lujstn"><img src="https://img.shields.io/twitter/follow/lujstn?style=social" alt="Twitter"></a>
+      <br>
+      <a href="https://www.instagram.com/lujstn/"><img src="https://img.shields.io/badge/Instagram-Follow-E4405F?style=social&logo=instagram" alt="Instagram"></a>
+      <br>
+      <a href="https://www.tiktok.com/@lujstn"><img src="https://img.shields.io/badge/TikTok-000000?style=flat&logo=tiktok&logoColor=white" alt="TikTok"></a>
+      <br>
+      <a href="https://lujstn.com"><img src="https://img.shields.io/badge/%F0%9F%94%97_lujstn.com-1a1a1a" alt="lujstn.com"></a>
+    </td>
+  </tr>
+</table>

openbrowse-1.8.1/openbrowse/__init__.py

@@ -0,0 +1,26 @@
+"""OpenBrowse: self-hosted AI browser automation with a live dashboard."""
+
+from __future__ import annotations
+
+import re
+from importlib import metadata
+from pathlib import Path
+
+
+def _read_version() -> str:
+    # @nonobvious(must-hold): a git checkout's pyproject.toml outranks installed
+    # metadata, which goes stale between a version bump and the next `uv sync`.
+    pyproject = Path(__file__).resolve().parent.parent / "pyproject.toml"
+    try:
+        match = re.search(r'^version = "([^"]+)"$', pyproject.read_text(), re.MULTILINE)
+        if match:
+            return match.group(1)
+    except OSError:
+        pass
+    try:
+        return metadata.version("openbrowse")
+    except metadata.PackageNotFoundError:
+        return "0.0.0"
+
+
+__version__ = _read_version()

openbrowse-1.8.1/openbrowse/agent/activity.py

@@ -0,0 +1,96 @@
+"""Ephemeral per-session activity state for the live feed.
+
+Holds what a running session is doing *right now* (waiting for the model, acting,
+preparing the next step) plus when that phase started, so the dashboard can show a
+live indicator with a count-up timer. Also the run-lifecycle registry: which
+sessions are currently running and which profile each has claimed, letting tools
+scale contention pacing to real concurrency and letting the runner refuse two
+live sessions on one profile. In-memory only; never persisted.
+"""
+
+from __future__ import annotations
+
+from datetime import datetime, timezone
+
+
+_activity: dict[str, dict] = {}
+
+
+def set_activity(
+    session_id: str,
+    label: str,
+    step: int | None = None,
+    spin: bool = False,
+    stream: str | None = None,
+    seconds: float | None = None,
+    kind: str | None = None,
+) -> None:
+    """Record what a session is doing. ``stream`` carries the full accumulated
+    text of a token-by-token phase (model reasoning as it generates); it is
+    never a window or a tail slice, and a call that omits it clears any
+    previous stream rather than inheriting it, so a phase change (e.g. moving
+    to "Running actions") can't leak stale reasoning text into the next read.
+    """
+    prev = _activity.get(session_id) or {}
+    now = datetime.now(timezone.utc).isoformat()
+    # @nonobvious(must-hold): a streaming phase re-pushes several times a second,
+    # so the clock has to survive an unchanged label or it reads near zero forever
+    # and no phase can report how long it took.
+    started = prev.get("startedAt") if prev.get("label") == label else None
+    _activity[session_id] = {
+        "label": label,
+        "startedAt": started or now,
+        "step": step if step is not None else prev.get("step"),
+        "spin": spin,
+        "stream": stream,
+        # @nonobvious(forced-by): a phase whose label was interrupted and restored
+        # loses its clock, so a caller that measured the real elapsed time says so
+        # here rather than leaving the dashboard to infer it from startedAt.
+        "seconds": seconds,
+        # @nonobvious(means): "reasoning" marks a phase whose prose is the point.
+        # The dashboard shimmers those and spins the rest, so a phase that thinks
+        # and a phase that acts do not claim the same affordance.
+        "kind": kind,
+    }
+
+
+def get_activity(session_id: str) -> dict | None:
+    return _activity.get(session_id)
+
+
+def clear_activity(session_id: str) -> None:
+    _activity.pop(session_id, None)
+
+
+_running_sessions: set[str] = set()
+
+_claimed_profiles: dict[str, str] = {}
+
+
+def session_started(session_id: str) -> None:
+    _running_sessions.add(session_id)
+
+
+def session_ended(session_id: str) -> None:
+    _running_sessions.discard(session_id)
+
+
+def active_session_count() -> int:
+    return len(_running_sessions)
+
+
+def try_claim_profile(profile_id: str, session_id: str) -> str | None:
+    """Claim a profile for a session. Returns None on success, or the id of the
+    session already holding it. Check-and-set with no await between, so a single
+    event loop cannot interleave two claims.
+    """
+    holder = _claimed_profiles.get(profile_id)
+    if holder is not None and holder != session_id:
+        return holder
+    _claimed_profiles[profile_id] = session_id
+    return None
+
+
+def release_profile(profile_id: str, session_id: str) -> None:
+    if _claimed_profiles.get(profile_id) == session_id:
+        _claimed_profiles.pop(profile_id, None)