jev-ultrafast README

The project's README from github.com/browser-use/jev-ultrafast. As one markdown file: README.md.

Jev Ultrafast · Browser Use × TypeSafe

Jev Ultrafast ⚡

[!IMPORTANT] The Browser Use Cloud waitlist is open. Get early access to ultrafast browser agents in the cloud. Join the waitlist →

A browser agent with a dynamic, indexed action space.

Give it one goal. TypeSafe's Jev picks an operation and an element. A small LLM writes text only when the operation is TYPE_TEXT.

Zürich → London on Google Flights in 7.1 seconds. One natural-language goal, actual text generation, and loading waits included.

A real Google Flights search at 1× speed, with generated city names and dynamic operation/target decisions

Watch the MP4 · Measurements · Read the loop

The action space

Every observation produces a new element table:

[1] button    Change ticket type · Round trip
[2] combobox  Where from?        · San Francisco
[3] combobox  Where to?          · empty
[4] textbox   Departure          · empty
...

The operations are CLICK, TYPE_TEXT, SELECT, SCROLL_UP, SCROLL_DOWN, WAIT, DONE, and BLOCKED. Only supported operations and targets are offered.

                      one TypeSafe request
                     ┌───────────────────────────┐
page → element table → operation                 │
                     │ click_target              │
                     │ type_text_target          │
                     │ select_target, if present │
                     └─────────────┬─────────────┘
                         use the matching target
                                   │
                    CLICK [7] ─────┤──→ browser
                TYPE_TEXT [3] ─────┘
                          ↓
                   small LLM → text → browser

Target questions are speculative. If the operation is CLICK, only click_target can execute. Two decisions, one network round trip. Each target head contains only compatible elements. Native dropdown choices carry an observed element/option index.

There are no site-specific action scripts or prepared field strings in the policy. The Flights example supplies a goal and independently verifies the outcome. The screenshot renderer adds labels afterward; it does not drive the browser.

Try it

git clone https://github.com/browser-use/jev-ultrafast.git
cd jev-ultrafast
uv sync
cp .env.example .env
# Add TYPESAFE_API_KEY and TEXT_MODEL_API_KEY.
uv run jev

Open http://127.0.0.1:8766 and click Start demo → Run automatically. The inspector shows numbered elements, operation probabilities, target probabilities, and executed actions. Choose next pauses before execution.

Chrome connects through Browser Harness, installed by uv sync. Run uv run browser-harness --doctor if it needs connecting. Allow remote debugging in Chrome when prompted.

TEXT_MODEL_API_KEY is an OpenRouter key in the example configuration. The current demo uses inception/mercury-2.5 with reasoning disabled. Gemini, GLM, and DeepSeek can also use the OpenAI-compatible text helper; configure the appropriate model, endpoint, and reasoning setting.

Use the library

from jev_ultrafast import Agent

with Agent(
    "https://www.google.com/travel/flights?hl=en",
    "Find one-way flights from Zurich to London on September 20, 2026, "
    "for one adult in economy. Stop when matching flight options are visible.",
) as agent:
    for state in agent.run():
        print(state["elapsed_ms"], state["status"])

Run with uv run --env-file .env python your_script.py. The same policy can run a different task:

uv run --env-file .env python examples/run.py \
  --url https://en.wikipedia.org/wiki/Main_Page \
  --goal 'Find and open the Wikipedia article about Gödel’s incompleteness theorems.'

uv run --env-file .env python examples/flights.py --keep-open performs the flight search, checks the actual route/date/results, and saves its trace. It does not select or book a flight.

Why it moves

Every executed target is resolved from an observed node. The executor rechecks page freshness and click occlusion. Model output never becomes selectors, coordinates, shell commands, or executable JavaScript. Text-helper output must parse as a small JSON object before typing.

Small enough to read

File Job
agent.py The complete loop and text-helper handoff
snapshot.js Atomic DOM snapshot, indexed controls, freshness guards
browser.py Browser connection, current geometry, execution
model.py Dynamic operation/target heads and text generation
questions.py Model instructions
demo.py Local inspector

Evidence and limits

The current video is a 7,073 ms Google Flights run. Timing starts after initial page observation and includes model calls, generated text, browser work, stale decisions, and loading waits. A fresh independent check verifies the one-way setting, Zürich, London, September 20, 2026, and visible flight options. The video plays at 1×, with no opening hold and a 0.5-second final hold.

In six alternating runs with identical models and settings, both versions passed 3/3. Median task time went from 9.450 s → 7.092 s, a 25% reduction; median browser protocol calls went from 1,092 → 101. This is three repeats of one task on one browser profile, not a general reliability benchmark.

The same policy opened the requested Wikipedia article in 2.798 s and passed a local hotel search/filter task in 1.896 s. Runs, failures, source hashes, and measurement boundaries are in performance.md.

A DONE choice still requires independent outcome verification. The DOM reader handles common HTML and ARIA controls, not the full accessible-name specification. Shadow roots, frames, canvas, uploads, pop-up tabs, nested scrolling, and arbitrary keyboard widgets remain outside this MVP. Owned tabs share the existing Chrome profile.

Development

uv run ruff check .
uv run pytest
node --check jev_ultrafast/static/app.js
node --check jev_ultrafast/snapshot.js
uv build

Tests are offline. uv run python scripts/check_guards.py checks real controls in a local browser without model calls. Live examples and recording scripts make paid API calls. scripts/record_flights.py <new-folder> captures original browser timestamps; scripts/render_demo.py <recording-folder> renders that verified run at 1× and crops out the Google account strip. Credentials and raw traces stay ignored.

Browser Use · Browser Harness · TypeSafe speculative fan-out

jev-ultrafast (browser-use/jev-ultrafast)

browser
Searching Database…

23.1K ·GitHub
Jev-ultrafast
What is this?A browser agent with a dynamic, indexed action space: give it one natural-language goal and it picks each operation and target element itself — for example, a Zuerich to London Google Flights search completed in about 7 seconds.
Is it a finished app, or a library I call from my own code?Both. A local demo inspector runs a full task at the click of a button and visualizes every decision, and the same policy can be driven from your own Python script with uv run.
How do I get it running?Install dependencies with uv sync (this also installs Browser Harness), open the demo and click Start demo → Run automatically, then allow remote debugging when Chrome prompts. If the connection drops, run uv run browser-harness --doctor.
Do I need a particular model or an API key?You need an OpenRouter key in TEXTMODELAPIKEY. The demo runs 'inception/mercury-2.5' with reasoning disabled, and Gemini, GLM and DeepSeek also work through the OpenAI-compatible text helper.
Are there sites or page features it can't handle yet?It's an MVP: the DOM reader handles common HTML and ARIA controls, not the full accessible-name specification. Shadow roots, frames, canvas, file uploads, pop-up tabs, nested scrolling and arbitrary keyboard widgets are out of scope.
Does it drive my own Chrome, with my existing logins?Yes. Tabs the agent owns share your existing Chrome profile, connecting through Browser Harness with remote debugging that you allow in Chrome.
Which actions can the agent perform on a page?Eight operations: CLICK, TYPETEXT, SELECT, SCROLLUP, SCROLLDOWN, WAIT, DONE and BLOCKED. Operation and target are chosen together in one network round trip, and a small LLM generates text only when the operation is TYPETEXT.
Why is it faster than typical screenshot-based browser agents?One model request per decision cycle, structured DOM state instead of screenshots, one browser call per atomic snapshot, waits capped at roughly 200 ms, and only visible text sent to the model. Focus emulation keeps background tabs rendering so Chrome doesn't throttle them.
Is the 7-second demo video real and unedited?Yes — it plays at 1x with no opening hold, and timing starts after the initial page observation, so model calls, generated text, browser work and loading waits are all included. A fresh independent check verifies the route, date and visible flight options.
Will it book or buy things on its own?No — the example flight task only performs the search, checks the actual route, date and results, and saves a trace; it does not select or book a flight. Even a DONE choice still requires independent outcome verification.
What are similar projects?SystemOneHarness, stagehand, BrowserKitten, scout, invisible_playwright_mcp, CUAWright, Agent-S, browser-use-pi
Which topics is it in?browser-agent, openrouter, chrome, browser-automation, openai-compatible, browser, benchmark, openai, performance, agent, automation, ai-agents
Creator
browser-use
Size
4.3 MB
Stars
23.1K
Updated
CommitSep 18
Age
1mo
Commits · 90 Days