Merge pull request #21 from j93es/feat/j93es

Feat/j93es
[Refactor] 리팩터링
2026-06-04 04:51:52 +09:00 · 2025-06-27 10:26:42 +09:00 · 2025-06-27 10:18:24 +09:00 · 2025-06-27 09:50:12 +09:00 · 2025-06-27 09:45:50 +09:00 · 2025-06-26 23:35:58 +09:00
49 changed files with 1727 additions and 5216 deletions
--- a/.env.example
+++ b/.env.example
@ -1,33 +1,30 @@
 ANONYMIZED_TELEMETRY=false
 # ========== LLM ========== 
 GOOGLE_API_KEY=
 # 권장 (다른 모델로 교체 가능) [다른 모델로 교체시 성능 보장 불가]
 GOOGLE_MODEL=gemini-2.5-flash-preview-05-20
 GOOGLE_PLANNER_MODEL=gemini-2.5-flash-preview-05-20
 # min(INITIAL_BACKOFF * (2 ** try_cnt), MAX_BACKOFF)만큼 API가 실패시 대기합니다.
 INITIAL_BACKOFF=60
 MAX_BACKOFF=600
 # ========== Monitoring ==========
 # 선택
 PROXY_HOST=127.0.0.1
 PROXY_PORT=11080
 BACKEND_URL=http://localhost:11081
-# https://docs.browser-use.com/development/observability
+# provider 계정 (본인이 사용하지 않는 계정 권장) (Github, apple, kakao등 다른 계정 추가 가능)
-# Lmnr 계정이 필요합니다.
+GOOGLE_ID=
-# https://lmnr.ai/
+GOOGLE_PASSWORD=
 LMNR_PROJECT_API_KEY=
-# 브라우저 언어 설정
+NAVER_ID=
-LANG=en_US
+NAVER_PASSWORD=
-# ========= Account ==========
+FACEBOOK_ID=
 FACEBOOK_PASSWORD=
-# 필수 뒤에 있는 이메일 주소는 Google 계정의 로그인 힌트로 사용됩니다.
+GITGUB_ID=
-# 이메일의 전체를 입력해주세요
+GITHUB_PASSWORD=
-GOOGLE_ID=whs.imnya.ng@gmail.com
+
 LinkedIn_ID=
 LinkedIn_PASSWORD=
 Microsoft_ID=
 Microsoft_PASSWORD=
--- a/.github/instructions/agent-settings.instructions.md
+++ b/.github/instructions/agent-settings.instructions.md
@ -1,345 +0,0 @@
 ---
 description: "Learn how to configure the agent"
 applyTo: '**'
 ---
 ## Overview
 The `Agent` class is the core component of Browser Use that handles browser automation. Here are the main configuration options you can use when initializing an agent.
 ## Basic Settings
 ```python
 from browser_use import Agent
 from langchain_openai import ChatOpenAI
 agent = Agent(
    task="Search for latest news about AI",
    llm=ChatOpenAI(model="gpt-4o"),
 )
 ```
 ### Required Parameters
 - `task`: The instruction for the agent to execute
 - `llm`: A LangChain chat model instance. See <a href="/customize/supported-models">LangChain Models</a> for supported models.
 ## Agent Behavior
 Control how the agent operates:
 ```python
 agent = Agent(
    task="your task",
    llm=llm,
    controller=custom_controller,  # For custom tool calling
    use_vision=True,              # Enable vision capabilities
    save_conversation_path="logs/conversation"  # Save chat logs
 )
 ```
 ### Behavior Parameters
 - `controller`: Registry of functions the agent can call. Defaults to base Controller. See <a href="/customize/custom-functions">Custom Functions</a> for details.
 - `use_vision`: Enable/disable vision capabilities. Defaults to `True`.
  - When enabled, the model processes visual information from web pages
  - Disable to reduce costs or use models without vision support
  - For GPT-4o, image processing costs approximately 800-1000 tokens (~$0.002 USD) per image (but this depends on the defined screen size)
 - `save_conversation_path`: Path to save the complete conversation history. Useful for debugging.
 - `override_system_message`: Completely replace the default system prompt with a custom one.
 - `extend_system_message`: Add additional instructions to the default system prompt.
 <Note>
  Vision capabilities are recommended for better web interaction understanding,
  but can be disabled to reduce costs or when using models without vision
  support.
 </Note>
 ### Reuse Existing Browser Context
 By default browser-use launches its own builtin browser using playwright chromium.
 You can also connect to a remote browser or pass any of the following 
 existing playwright objects to the Agent: `page`, `browser_context`, `browser`, `browser_session`, or `browser_profile`.
 These all get passed down to create a `BrowserSession` for the `Agent`:
 ```python
 agent = Agent(
    task='book a flight to fiji',
    llm=llm,
    browser_profile=browser_profile,  # use this profile to create a BrowserSession
    browser_session=BrowserSession(   # use an existing BrowserSession
      cdp_url=...,                      # remote CDP browser to connect to
      # or
      wss_url=...,                      # remote wss playwright server provider
      # or
      browser_pid=...                   # pid of a locally running browser process to attach to
      # or
      executable_path=...               # provide a custom chrome binary path
      # or
      channel=...                       # specify chrome, chromium, ms-edge, etc.
      # or
      page=page,                        # use an existing playwright Page object
      # or
      browser_context=browser_context,  # use an existing playwright BrowserContext object
      # or
      browser=browser,                  # use an existing playwright Browser object
    ),
 )
 ```
 For example, to connect to an existing browser over CDP you could do:
 ```python
 agent = Agent(
    ...
    browser_session=BrowserSession(cdp_url='http://localhost:9222'),
 )
 ```
 For example, to connect to a local running chrome instance you can do:
 ```python
 agent = Agent(
    ...
    browser_session=BrowserSession(browser_pid=1234),
 )
 ```
 See <a href="/customize/real-browser">Connect to your Browser</a> for more info.
 <Note>
  You can reuse the same `BrowserSession` after an agent has completed running. If you do nothing, the
  browser will be automatically closed on `run()` completion only if it was launched by us.
 </Note>
 ## Running the Agent
 The agent is executed using the async `run()` method:
 - `max_steps` (default: `100`)
  Maximum number of steps the agent can take during execution. This prevents infinite loops and helps control execution time.
 ## Agent History
 The method returns an `AgentHistoryList` object containing the complete execution history. This history is invaluable for debugging, analysis, and creating reproducible scripts.
 ```python
 # Example of accessing history
 history = await agent.run()
 # Access (some) useful information
 history.urls()              # List of visited URLs
 history.screenshots()       # List of screenshot paths
 history.action_names()      # Names of executed actions
 history.extracted_content() # Content extracted during execution
 history.errors()           # Any errors that occurred
 history.model_actions()     # All actions with their parameters
 ```
 The `AgentHistoryList` provides many helper methods to analyze the execution:
 - `final_result()`: Get the final extracted content
 - `is_done()`: Check if the agent completed successfully
 - `has_errors()`: Check if any errors occurred
 - `model_thoughts()`: Get the agent's reasoning process
 - `action_results()`: Get results of all actions
 <Note>
  For a complete list of helper methods and detailed history analysis
  capabilities, refer to the [AgentHistoryList source
  code](https://github.com/browser-use/browser-use/blob/main/browser_use/agent/views.py#L111).
 </Note>
 ## Run initial actions without LLM
 With [this example](https://github.com/browser-use/browser-use/blob/main/examples/features/initial_actions.py) you can run initial actions without the LLM.
 Specify the action as a dictionary where the key is the action name and the value is the action parameters. You can find all our actions in the [Controller](https://github.com/browser-use/browser-use/blob/main/browser_use/controller/service.py) source code.
 ```python
 initial_actions = [
 	{'open_tab': {'url': 'https://www.google.com'}},
 	{'open_tab': {'url': 'https://en.wikipedia.org/wiki/Randomness'}},
 	{'scroll_down': {'amount': 1000}},
 ]
 agent = Agent(
 	task='What theories are displayed on the page?',
 	initial_actions=initial_actions,
 	llm=llm,
 )
 ```
 ## Run with message context
 You can configure the agent and provide a separate message to help the LLM understand the task better.
 ```python
 from langchain_openai import ChatOpenAI
 agent = Agent(
    task="your task",
    message_context="Additional information about the task",
    llm = ChatOpenAI(model='gpt-4o')
 )
 ```
 ## Run with planner model
 You can configure the agent to use a separate planner model for high-level task planning:
 ```python
 from langchain_openai import ChatOpenAI
 # Initialize models
 llm = ChatOpenAI(model='gpt-4o')
 planner_llm = ChatOpenAI(model='o3-mini')
 agent = Agent(
    task="your task",
    llm=llm,
    planner_llm=planner_llm,           # Separate model for planning
    use_vision_for_planner=False,      # Disable vision for planner
    planner_interval=4                 # Plan every 4 steps
 )
 ```
 ### Planner Parameters
 - `planner_llm`: A LangChain chat model instance used for high-level task planning. Can be a smaller/cheaper model than the main LLM.
 - `use_vision_for_planner`: Enable/disable vision capabilities for the planner model. Defaults to `True`.
 - `planner_interval`: Number of steps between planning phases. Defaults to `1`.
 Using a separate planner model can help:
 - Reduce costs by using a smaller model for high-level planning
 - Improve task decomposition and strategic thinking
 - Better handle complex, multi-step tasks
 <Note>
  The planner model is optional. If not specified, the agent will not use the planner model.
 </Note>
 ### Optional Parameters
 - `message_context`: Additional information about the task to help the LLM understand the task better.
 - `initial_actions`: List of initial actions to run before the main task.
 - `max_actions_per_step`: Maximum number of actions to run in a step. Defaults to `10`.
 - `max_failures`: Maximum number of failures before giving up. Defaults to `3`.
 - `retry_delay`: Time to wait between retries in seconds when rate limited. Defaults to `10`.
 - `generate_gif`: Enable/disable GIF generation. Defaults to `False`. Set to `True` or a string path to save the GIF.
 ## Memory Management
 Browser Use includes a procedural memory system using [Mem0](https://mem0.ai) that automatically summarizes the agent's conversation history at regular intervals to optimize context window usage during long tasks.
 ```python
 from browser_use.agent.memory import MemoryConfig
 agent = Agent(
    task="your task",
    llm=llm,
    enable_memory=True,
    memory_config=MemoryConfig( # Ensure llm_instance is passed if not using default LLM config
        llm_instance=llm,      # Important: Pass the agent's LLM instance here
        agent_id="my_custom_agent",
        memory_interval=15
    )
 )
 ```
 ### Memory Parameters
 - `enable_memory`: Enable/disable the procedural memory system. Defaults to `True`.
 - `memory_config`: A `MemoryConfig` Pydantic model instance (required if `enable_memory` is `True`). Dictionary format is not supported.
 ### Using MemoryConfig
 You must configure the memory system using the `MemoryConfig` Pydantic model for a type-safe approach:
 ```python
 from browser_use.agent.memory import MemoryConfig
 from langchain_openai import ChatOpenAI # Assuming llm is an instance of ChatOpenAI
 llm_for_agent = ChatOpenAI(model="gpt-4o")
 agent = Agent(
    task=task_description,
    llm=llm_for_agent,
    enable_memory=True, # This is True by default
    memory_config=MemoryConfig(
        llm_instance=llm_for_agent, # Pass the LLM instance for Mem0
        agent_id="my_agent",
        memory_interval=15, # Summarize every 15 steps
        embedder_provider="openai",
        embedder_model="text-embedding-3-large",
        embedder_dims=1536,
        # --- Vector Store Customization ---
        vector_store_provider="qdrant", # e.g., Qdrant, Pinecone, Chroma, etc.
        vector_store_collection_name="my_browser_use_memories", # Optional: custom collection name
        vector_store_config_override={ # Provider-specific config
            "host": "localhost",
            "port": 6333
            # Add other Qdrant specific configs here if needed, e.g., api_key for cloud
        }
    )
 )
 ```
 The `MemoryConfig` model provides these configuration options:
 #### Memory Settings
 - `agent_id`: Unique identifier for the agent (default: `"browser_use_agent"`). Essential for persistent memory sessions if using a persistent vector store.
 - `memory_interval`: Number of steps between memory summarization (default: `10`)
 #### LLM Settings (for Mem0's internal operations)
 - `llm_instance`: The LangChain `BaseChatModel` instance that Mem0 will use for its internal summarization and processing. You must pass the same LLM instance used by the main agent, or another compatible one, here.
 #### Embedder Settings
 - `embedder_provider`: Provider for embeddings (`'openai'`, `'gemini'`, `'ollama'`, or `'huggingface'`)
 - `embedder_model`: Model name for the embedder
 - `embedder_dims`: Dimensions for the embeddings
 #### Vector Store Settings
 - `vector_store_provider`: Choose the vector store backend. Supported options include:
  `'faiss'` (default), `'qdrant'`, `'pinecone'`, `'supabase'`, `'elasticsearch'`, `'chroma'`, `'weaviate'`, `'milvus'`, `'pgvector'`, `'upstash_vector'`, `'vertex_ai_vector_search'`, `'azure_ai_search'`, `'lancedb'`, `'mongodb'`, `'redis'`, `'memory'` (in-memory, non-persistent).
 - `vector_store_collection_name`: (Optional) Specify a custom name for the collection or index in your vector store. If not provided, a default name is generated (especially for local stores like FAISS/Chroma) or used by Mem0.
 - `vector_store_base_path`: Path for local vector stores like FAISS or Chroma (e.g., `/tmp/mem0`). Default is `/tmp/mem0`.
 - `vector_store_config_override`: (Optional) A dictionary to provide or override specific configuration parameters required by Mem0 for the chosen `vector_store_provider`. This is where you'd put connection details like `host`, `port`, `api_key`, `url`, `environment`, etc., for cloud-based or server-based vector stores.
 The model automatically sets appropriate defaults based on the LLM being used:
 - For `ChatOpenAI`: Uses OpenAI's `text-embedding-3-small` embeddings
 - For `ChatGoogleGenerativeAI`: Uses Gemini's `models/text-embedding-004` embeddings
 - For `ChatOllama`: Uses Ollama's `nomic-embed-text` embeddings
 - Default: Uses Hugging Face's `all-MiniLM-L6-v2` embeddings
 <Note>
  **Important:**
  - Always pass a properly constructed `MemoryConfig` object to the `memory_config` parameter.
  - Ensure the `llm_instance` is provided to `MemoryConfig` so Mem0 can perform its operations.
  - For persistent memory across agent runs or for shared memory, choose a scalable vector store provider (like Qdrant, Pinecone, etc.) and configure it correctly using `vector_store_provider` and `vector_store_config_override`. The default 'faiss' provider stores data locally in `vector_store_base_path`.
 </Note>
 ### How Memory Works
 When enabled, the agent periodically compresses its conversation history into concise summaries:
 1. Every `memory_interval` steps, the agent reviews its recent interactions.
 2. It uses Mem0 (configured with your chosen LLM and vector store) to create a procedural memory summary.
 3. The original messages in the agent's active context are replaced with this summary, reducing token usage.
 4. This process helps maintain important context while freeing up the context window for new information.
 ### Disabling Memory
 If you want to disable the memory system (for debugging or for shorter tasks), set `enable_memory` to `False`:
 ```python
 agent = Agent(
    task="your task",
    llm=llm,
    enable_memory=False
 )
 ```
 <Note>
  Disabling memory may be useful for debugging or short tasks, but for longer
  tasks, it can lead to context window overflow as the conversation history
  grows. The memory system helps maintain performance during extended sessions.
 </Note>
--- a/.github/instructions/browser-settings.instructions.md
+++ b/.github/instructions/browser-settings.instructions.md
@ -1,968 +0,0 @@
 ---
 description: "Launch or connect to an existing browser and configure it to your needs."
 applyTo: '**'
 ---
 Browser Use uses [playwright](https://playwright.dev/python/docs/api/class-browsertype#browser-type-launch-persistent-context) (or [patchright](https://github.com/Kaliiiiiiiiii-Vinyzu/patchright)) to manage its connection with a real browser.
 ---
 **To launch or connect to a browser**, pass any playwright / browser-use configuration arguments you want to `BrowserSession(...)`:
 ```python
 from browser_use import BrowserSession, Agent
 browser_session = BrowserSession(
    headless=True,
    viewport={'width': 964, 'height': 647},
    user_data_dir='~/.config/browseruse/profiles/default',
 )
 agent = Agent('fill out the form on this page', browser_session=browser_session)
 ```
 <Note>
  The new `BrowserSession` & `BrowserProfile` accept all the same arguments that Playwright's [`launch_persistent_context(...)`](https://playwright.dev/python/docs/api/class-browsertype#browser-type-launch-persistent-context) takes, giving you full control over browser settings at launch. (see below for the full list)
 </Note>
 ---
 ## `BrowserSession`
 - 🎭 `BrowserSession(**params)` is Browser Use's object that tracks a playwright connection to a running browser. It sets up:
  - the `playwright` library, `browser` and/or `browser_context`, and `page` objects and tracks which tabs the agent & human are focused on
  - methods to interact with the browser window, apply config needed by the Agent, and run the `DOMService` for element detection
  - it can take a `browser_profile=BrowserProfile(...)` template containing some config defaults, and `**kwargs` session-specific config overrides
 ### Browser Connection Parameters
 Provide any one of these options to connect to an existing browser. These options are session-specific and cannot be stored in a `BrowserProfile(...)` template.
 #### `wss_url`
 ```python
 wss_url: str | None = None
 ```
 WSS URL of the playwright-protocol browser server to connect to. See here for [WSS connection instructions](https://docs.browser-use.com/customize/real-browser#method-d%3A-connect-to-remote-playwright-node-js-browser-server-via-wss-url).
 #### `cdp_url`
 ```python
 cdp_url: str | None = None
 ```
 CDP URL of the browser to connect to (e.g. `http://localhost:9222`). See here for [CDP connection instructions](https://docs.browser-use.com/customize/real-browser#method-e%3A-connect-to-remote-browser-via-cdp-url).
 #### `browser_pid`
 ```python
 browser_pid: int | None = None
 ```
 PID of a running chromium-based browser process to connect to on localhost. See here for [connection via pid](https://docs.browser-use.com/customize/real-browser#method-c%3A-connect-to-local-browser-using-browser-pid) instructions.
 <Note>
  For web scraping tasks on sites that restrict automated access, we recommend
  using [our cloud](https://browser-use.com) or an external browser provider for better reliability. 
  See the [Connect to your Browser](real-browser) guide for detailed connection instructions.
 </Note>
 ### Session-Specific Parameters
 #### `browser_profile`
 ```python
 browser_profile: BrowserProfile = BrowserProfile()
 ```
 Optional `BrowserProfile` template containing default config to use for the `BrowserSession`. (see below for more info)
 #### `playwright`
 ```python
 playwright: Playwright | None = None
 ```
 Optional playwright or patchright API client handle to use, the result of `(await async_playwright().start())` or `(await async_patchright().start())`, which spawns a node.js child subprocess that relays commands to the browser over CDP.
 See here for [more detailed usage instructions](https://docs.browser-use.com/customize/real-browser#method-b%3A-connect-using-existing-playwright-objects).
 #### `browser`
 ```python
 browser: Browser | None = None
 ```
 Playwright Browser object to use (optional). See here for [more detailed usage instructions](https://docs.browser-use.com/customize/real-browser#method-b%3A-connect-using-existing-playwright-objects).
 #### `browser_context`
 ```python
 browser_context: BrowserContext | None = None
 ```
 Playwright BrowserContext object to use (optional). See here for [more detailed usage instructions](https://docs.browser-use.com/customize/real-browser#method-b%3A-connect-using-existing-playwright-objects).
 #### `page` *aka* `agent_current_page`
 <a name="page"></a><a name="agent-current-page"></a>
 ```python
 page: Page | None = None
 ```
 Foreground Page that the agent is focused on, can also be passed as `page=...` as a shortcut. See here for [more detailed usage instructions](https://docs.browser-use.com/customize/real-browser#method-b%3A-connect-using-existing-playwright-objects).
 #### `human_current_page`
 ```python
 human_current_page: Page | None = None
 ```
 Foreground Page that the human is focused on to start, not necessary to set manually.
 #### `initialized`
 ```python
 initialized: bool = False
 ```
 Mark BrowserSession as already initialized, skips launch/connection (not recommended)
 #### `**kwargs`
 `BrowserSession` can also accept *all* of the parameters [below](#browserprofile).
 (the parameters *above* this point are specific to `BrowserSession` and cannot be stored in a `BrowserProfile` template)
 Extra `**kwargs` passed to `BrowserSession(...)` act as session-specific overrides to the `BrowserProfile(...)` template.
 ```python
 base_iphone13 = BrowserProfile(
    storage_state='/tmp/auth.json',     # share cookies between parallel browsers
    **playwright.devices['iPhone 13'],
    timezone_id='UTC',
 )
 usa_phone = BrowserSession(
    browser_profile=base_iphone13,
    timezone_id='America/New_York',     # kwargs override values in base_iphone13
 )
 eu_phone = BrowserSession(
    browser_profile=base_iphone13,
    timezone_id='Europe/Paris',
 )
 usa_agent = Agent(task='show me todays schedule...', browser_session=usa_phone)
 eu_agent = Agent(task='show me todays schedule...', browser_session=eu_phone)
 await asyncio.gather(agent1.run(), agent2.run())
 ```
 ---
 ## `BrowserProfile`
 A `BrowserProfile` is a 📋 config template for a 🎭 `BrowserSession(...)`.
 It's basically just a typed + validated version of a `dict` to hold config.
 When you find yourself storing or re-using many browser configs, you can upgrade from:
 ```diff
 - config = {key: val, key: val, ...}
 - BrowserSession(**config)
 ```
 To this instead:
 ```diff
 + config = BrowserProfile(key=val, key=val, ...)
 + BrowserSession(browser_profile=config)
 ```
 <Tip>
 You don't ever *need* to use a `BrowserProfile`, you can always pass config parameters directly to `BrowserSession`:
 ```python
 session = BrowserSession(headless=True, storage_state='auth.json', viewport={...}, ...)
 ```
 </Tip>
 `BrowserProfile` is optional, but it provides a number of benefits over a normal `dict` for holding config:
 - has type hints and pydantic field descriptions that show up in your IDE
 - validates config at runtime quickly without having to start a browser
 - provides helper methods to autodetect screen size, set up local paths, save/load config as json, and more...
 <Tip>
 `BrowserProfiles`s are designed to easily be given 🆔 `uuid`s and put in a database + made editable by users. 
 `BrowserSession`s get their own 🆔 `uuid`s and be linked by 🖇 foreign key to whatever `BrowserProfiles` they use.  
 This cleanly separates the per-connection rows from the bulky re-usable config and avoids wasting space in your db.
 This is useful because a user may only have 2 or 3 profiles, but they could have 100k+ sessions within a few months.
 </Tip>
 `BrowserProfile` and `BrowserSession` can both take any of the:
 - [Playwright parameters](#playwright)
 - [Browser-Use parameters](#browser-use-parameters) (extra options we provide on top of `playwright`)
 The only parameters `BrowserProfile` can NOT take are the session-specific connection parameters and live playwright objects:  
 `cdp_url`, `wss_url`, `browser_pid`, `page`, `browser`, `browser_context`, `playwright`, etc.
 ### Basic Example
 ```python
 from browser_use.browser import BrowserProfile
 profile = BrowserProfile(
    stealth=True,
    storage_state='/tmp/google_docs_cookies.json',
    allowed_domains=['docs.google.com', 'https://accounts.google.com'],
    viewport={'width': 396, 'height': 774},
    # ... playwright args / browser-use config args ...
 )
 phone1 = BrowserSession(browser_profile=profile, device_scale_factor=1)
 phone2 = BrowserSession(browser_profile=profile, device_scale_factor=2)
 phone3 = BrowserSession(browser_profile=profile, device_scale_factor=3)
 ```
 ### Browser-Use Parameters
 These parameters control Browser Use-specific features, and are outside the standard playwright set. They can be passed to `BrowserSession(...)` and/or stored in a `BrowserProfile` template.
 #### `keep_alive`
 ```python
 keep_alive: bool | None = None
 ```
 If `True` it wont close the browser after the first `agent.run()` ends. Useful for running multiple tasks with the same browser instance. If this is left as `None` and the Agent launched its own browser, the default is to close the browser after the agent completes. If the agent connected to an existing browser then it will leave it open.
 #### `stealth`
 ```python
 stealth: bool = False
 ```
 Set to `True` to use [`patchright`](https://github.com/Kaliiiiiiiiii-Vinyzu/patchright) to avoid bot-blocking. (Might cause issues with some sites, requires manual testing.)
 <a name="restrict-urls"></a>
 #### `allowed_domains`
 ```python
 allowed_domains: list[str] | None = None
 ```
 List of allowed domains for navigation. If None, all domains are allowed.
 Example: `['google.com', '*.wikipedia.org']` - Here the agent will only be able to access `google.com` exactly and `wikipedia.org` + `*.wikipedia.org`.
 Glob patterns are supported:
 - `['example.com']` ✅ will match only `https://example.com/*` exactly, subdomains will not be allowed.
  It's always the most secure to list all the domains you want to give the access to explicitly w/ schemes e.g.
  `['https://google.com', 'http*://www.google.com', 'https://myaccount.google.com', 'https://mail.google.com', 'https://docs.google.com']`
 - `['*.example.com']` ⚠️ **CAUTION** this will match `https://example.com` and *all* its subdomains.
  Make sure *all* the subdomains are safe for the agent! `abc.example.com`, `def.example.com`, ..., `useruploads.example.com`, `admin.example.com`
 #### `disable_security`
 ```python
 disable_security: bool = False
 ```
 Completely disables all basic browser security features. Allows interacting across cross-site iFrames boundaries, but 
 <Warning>
 This option is very INSECURE and is only for niche use cases. DO NOT LET YOUR AGENT visit untrusted URLs or give it real cookies when `disable_security=True`.
 Visiting a single malicious site in this mode can trivially compromise *all* the cookies in the browser profile in under 1 second.
 </Warning>
 #### `deterministic_rendering`
 ```python
 deterministic_rendering: bool = False
 ```
 Attempt to forced more deterministic rendering for consistent screenshots across different host operating systems and hardware.
 Disables OS-specific font hints, aliasing, GPU-accelerated rendering, normalizes DPI, and sets a specific JS random seed to try to avoid nondeterministic JS.
 <Warning>
 This flag is for niche use cases (e.g. screenshot diffing) where pixel-perfect rendering across different server operating systems is more important than stability.
 It makes the agent more likely to be blocked as a bot and triggers some glitchy behavior in chrome occasionally, it's not recommended unless you know you need it.
 </Warning>
 #### `highlight_elements`
 ```python
 highlight_elements: bool = True
 ```
 Highlight interactive elements on the screen with colorful bounding boxes.
 #### `viewport_expansion`
 ```python
 viewport_expansion: int = 500
 ```
 Viewport expansion in pixels. With this you can control how much of the page is included in the context of the LLM:
 - `-1`: All elements from the entire page will be included, regardless of visibility (highest token usage but most complete).
 - `0`: Only elements which are currently visible in the viewport will be included.
 - `500` (default): Elements in the viewport plus an additional 500 pixels in each direction will be included, providing a balance between context and token usage.
 #### `include_dynamic_attributes`
 ```python
 include_dynamic_attributes: bool = True
 ```
 Include dynamic attributes in selectors for better element targeting.
 #### `minimum_wait_page_load_time`
 ```python
 minimum_wait_page_load_time: float = 0.25
 ```
 Minimum time to wait before capturing page state for LLM input.
 #### `wait_for_network_idle_page_load_time`
 ```python
 wait_for_network_idle_page_load_time: float = 0.5
 ```
 Time to wait for network activity to cease. Increase to 3-5s for slower websites. This tracks essential content loading, not dynamic elements like videos.
 #### `maximum_wait_page_load_time`
 ```python
 maximum_wait_page_load_time: float = 5.0
 ```
 Maximum time to wait for page load before proceeding.
 #### `wait_between_actions`
 ```python
 wait_between_actions: float = 0.5
 ```
 Time to wait between agent actions.
 #### `cookies_file`
 ```python
 cookies_file: str | None = None
 ```
 JSON file path to save cookies to.
 <Warning>
 This option is DEPRECATED. Use [`storage_state`](#storage-state) instead, it's the standard playwright format and also supports `localStorage` and `indexedDB`!
 The library will automatically save a new `storage_state.json` next to any `cookies_file` path you provide, just use `storage_state='path/to/storage_state.json' to switch to the new format:
 `cookies_file.json`: `[{cookie}, {cookie}, {cookie}]`  
 ⬇️
 `storage_state.json`: `{"cookies": [{cookie}, {cookie}, {cookie}], "origins": {... optional localstorage state ...}}`
 Or run `playwright open https://example.com/ --save-storage=storage_state.json` and log into any sites you need to generate a fresh storage state file.
 </Warning>
 #### `profile_directory`
 ```python
 profile_directory: str = 'Default'
 ```
 Chrome profile subdirectory name inside of your `user_data_dir` (e.g. `Default`, `Profile 1`, `Work`, etc.).
 No need to set this unless you have multiple profiles set up in a single `user_data_dir` and need to use a specific one.
 #### `window_position`
 ```python
 window_position: dict | None = {"width": 0, "height": 0}
 ```
 Window position from top-left.
 ---
 <a name="playwright-parameters"></a><a name="playwright"></a>
 ### Playwright Launch Options
 All the parameters below are standard playwright parameters and can be passed to both `BrowserSession` and `BrowserProfile`.
 They are defined in `browser_use/browser/profile.py`. See here for the [official Playwright documentation](https://playwright.dev/python/docs/api/class-browsertype#browser-type-launch-persistent-context) for all of these options.
 #### `headless`
 ```python
 headless: bool | None = None
 ```
 Runs the browser without a visible UI. If None, auto-detects based on display availability. If you set `headless=False` on a server with no monitor attached, the browser will fail to launch (use `xvfb` + vnc to give a headless server a virtual display you can remote control). 
 `headless=False` is recommended for maximum stealth and is required for human-in-the-loop workflows.
 #### `channel`
 ```python
 channel: BrowserChannel = 'chromium'
 ```
 Browser channel: `['chromium']` (default when `stealth=False`), `'chrome'` (default when `stealth=True`), `'chrome-beta'`, `'chrome-dev'`, `'chrome-canary'`, `'msedge'`, `'msedge-beta'`, `'msedge-dev'`, `'msedge-canary'`
 Don't worry, other chromium-based browsers not in this list (e.g. `brave`) are still supported if you provide your own [`executable_path`](#executable_path), just set it to `chromium` for those.
 #### `executable_path`
 ```python
 executable_path: str | Path | None = None
 ```
 Path to browser executable for custom installations.
 #### `user_data_dir`
 ```python
 user_data_dir: str | Path | None = '~/.config/browseruse/profiles/default'
 ```
 Directory for browser profile data. Set to `None` to use an ephemeral temporary profile (aka incognito mode).
 Multiple running browsers **cannot share a single `user_data_dir` at the same time**. You must set it to `None` or
 provide a unique `user_data_dir` per-session if you plan to run multiple browsers.
 The browser version run must always be equal to or greater than the version used to create the `user_data_dir`.
 If you see errors like `Failed to parse Extensions` or similar and failures when launching, you're attempting to run an older browser with an incompatible `user_data_dir` that's already been migrated to a newer schema version.
 #### `args`
 ```python
 args: list[str] = []
 ```
 Additional command-line arguments to pass to the browser. See here for the [full list of available chrome launch options](https://peter.sh/experiments/chromium-command-line-switches/).
 #### `ignore_default_args`
 ```python
 ignore_default_args: list[str] | bool = ['--enable-automation', '--disable-extensions']
 ```
 List of default CLI args to stop playwright from including when launching chrome. Set it to `True` to disable *all* default options (not recommended).
 #### `env`
 ```python
 env: dict[str, str] = {}
 ```
 Extra environment variables to set when launching browser. e.g. `{'DISPLAY': '1'}` to use a specific X11 display.
 #### `chromium_sandbox`
 ```python
 chromium_sandbox: bool = not IN_DOCKER
 ```
 Whether to enable Chromium sandboxing (recommended for security). Should always be `False` when running inside Docker
 because Docker provides its own sandboxing can conflict with Chrome's.
 #### `devtools`
 ```python
 devtools: bool = False
 ```
 Whether to open DevTools panel automatically (only works when `headless=False`).
 #### `slow_mo`
 ```python
 slow_mo: float = 0
 ```
 Slow down actions by this many milliseconds.
 #### `timeout`
 ```python
 timeout: float = 30000
 ```
 Default timeout in milliseconds for connecting to a remote browser.
 #### `accept_downloads`
 ```python
 accept_downloads: bool = True
 ```
 Whether to automatically accept all downloads.
 #### `proxy`
 ```python
 proxy: dict | None = None
 ```
 Proxy settings. Example: `{"server": "http://proxy.com:8080", "username": "user", "password": "pass"}`. 
 #### `permissions`
 ```python
 permissions: list[str] = ['clipboard-read', 'clipboard-write', 'notifications']
 ```
 Browser permissions to grant. See here for the [full list of available permission](https://playwright.dev/python/docs/api/class-browsercontext#browser-context-grant-permissions).
 #### `storage_state`
 ```python
 storage_state: str | Path | dict | None = None
 ```
 Browser storage state (cookies, localStorage). Can be file path or dict. See here for the [Playwright `storage_state` documentation](https://playwright.dev/python/docs/api/class-browsercontext#browser-context-storage-state) on how to use it.
 This option is only applied when launching a new browser using the default builtin playwright chromium and `user_data_dir=None` is set.
 ```bash
 # to create a storage state file, run the following and log into the sites you need once the browser opens:
 playwright open https://example.com/ --save-storage=./storage_state.json
 # then setup a BrowserSession with storage_state='./storage_state.json' and user_data_dir=None to use it
 ```
 ### Playwright Timing Settings
 These control how the browser waits for CDP API calls to complete and pages to load.
 #### `default_timeout`
 ```python
 default_timeout: float | None = None
 ```
 Default timeout for Playwright operations in milliseconds.
 #### `default_navigation_timeout`
 ```python
 default_navigation_timeout: float | None = None
 ```
 Default timeout for page navigation in milliseconds.
 ### Playwright Viewport Options
 Configure browser window size, viewport, and display properties:
 #### `user_agent`
 ```python
 user_agent: str | None = None
 ```
 Specific user agent to use in this context.
 #### `is_mobile`
 ```python
 is_mobile: bool = False
 ```
 Whether the meta viewport tag is taken into account and touch events are enabled.
 #### `has_touch`
 ```python
 has_touch: bool = False
 ```
 Specifies if viewport supports touch events.
 #### `geolocation`
 ```python
 geolocation: dict | None = None
 ```
 Geolocation coordinates. Example: `{"latitude": 59.95, "longitude": 30.31667}`
 #### `locale`
 ```python
 locale: str | None = None
 ```
 Specify user locale, for example en-GB, de-DE, etc. Locale will affect the navigator.language value, Accept-Language request header value as well as number and date formatting rules.
 #### `timezone_id`
 ```python
 timezone_id: str | None = None
 ```
 Timezone identifier (e.g., 'America/New_York').
 #### `window_size`
 ```python
 window_size: dict | None = None
 ```
 Browser window size for headful mode. Example: `{"width": 1920, "height": 1080}`
 #### `viewport`
 ```python
 viewport: dict | None = None
 ```
 Viewport size with `width` and `height`. Example: `{"width": 1280, "height": 720}`
 #### `no_viewport`
 ```python
 no_viewport: bool | None = not headless
 ```
 Disable fixed viewport. Content will resize with window.
 *Tip:* don't use this parameter, it's a playwright standard parameter but it's redundant and only serves to override the `viewport` setting above.
 A viewport is *always* used in headless mode regardless of this setting, and is *never* used in headful mode unless you pass `viewport={width, height}` explicitly.
 #### `device_scale_factor`
 ```python
 device_scale_factor: float | None = None
 ```
 Device scale factor (DPI). Useful for high-resolution screenshots (set it to 2).
 #### `screen`
 ```python
 screen: dict | None = None
 ```
 Screen size available to browser. Auto-detected if not specified.
 #### `color_scheme`
 ```python
 color_scheme: ColorScheme = 'light'
 ```
 Preferred color scheme: `'light'`, `'dark'`, `'no-preference'`
 #### `contrast`
 ```python
 contrast: Contrast = 'no-preference'
 ```
 Contrast preference: `'no-preference'`, `'more'`, `'null'`
 #### `reduced_motion`
 ```python
 reduced_motion: ReducedMotion = 'no-preference'
 ```
 Reduced motion preference: `'reduce'`, `'no-preference'`, `'null'`
 #### `forced_colors`
 ```python
 forced_colors: ForcedColors = 'none'
 ```
 Forced colors mode: `'active'`, `'none'`, `'null'`
 #### `**playwright.devices[...]`
 Playwright provides launch & context arg presets to [emulate common device fingerprints](https://playwright.dev/python/docs/emulation).
 ```python
 BrowserProfile(
    ...
    **playwright.devices['iPhone 13'],    # playwright = await async_playwright().start()
 )
 ```
 Because `BrowserSession` and `BrowserProfile` take all the standard playwright args, we are able to support these device presets as well.
 ### Playwright Security Options
 > See `allowed_domains` above too!
 #### `offline`
 ```python
 offline: bool = False
 ```
 Emulate network being offline.
 #### `http_credentials`
 ```python
 http_credentials: dict | None = None
 ```
 Credentials for HTTP authentication.
 #### `extra_http_headers`
 ```python
 extra_http_headers: dict[str, str] = {}
 ```
 Additional HTTP headers to be sent with every request.
 #### `ignore_https_errors`
 ```python
 ignore_https_errors: bool = False
 ```
 Whether to ignore HTTPS errors when sending network requests.
 #### `bypass_csp`
 ```python
 bypass_csp: bool = False
 ```
 Toggles bypassing Content-Security-Policy.
 #### `java_script_enabled`
 ```python
 java_script_enabled: bool = True
 ```
 Whether or not to enable JavaScript in the context.
 #### `service_workers`
 ```python
 service_workers: ServiceWorkers = 'allow'
 ```
 Whether to allow sites to register Service workers: `'allow'`, `'block'`
 #### `base_url`
 ```python
 base_url: str | None = None
 ```
 Base URL to be used in `page.goto()` and similar operations.
 #### `strict_selectors`
 ```python
 strict_selectors: bool = False
 ```
 If true, selector passed to Playwright methods will throw if more than one element matches.
 #### `client_certificates`
 ```python
 client_certificates: list[ClientCertificate] = []
 ```
 Client certificates to be used with requests.
 ### Playwright Recording Options
 Note: Browser Use also provides some of our own recording-related options not listed below (see above).
 #### `record_video_dir`
 <a name="record-video-dir"></a>
 <a name="save-recording-path"></a>
 ```python
 record_video_dir: str | Path | None = None
 ```
 Directory to save `.webm` video recordings. [Playwright Docs: `record_video_dir`](https://playwright.dev/python/docs/api/class-browsertype#browser-type-launch-persistent-context-option-record-video-dir)
 <Note>
 This parameter also has an alias `save_recording_path` for backwards compatibility with past versions, but we recommend using the standard Playwright name `record_video_dir` going forward.
 </Note>
 #### `record_video_size`
 ```python
 record_video_size: dict | None = None. [Playwright Docs: `record_video_size`](https://playwright.dev/python/docs/api/class-browsertype#browser-type-launch-persistent-context-option-record-video-size)
 ```
 Video size. Example: `{"width": 1280, "height": 720}`
 #### `record_har_path`
 <a name="record-har-path"></a>
 <a name="save-har-path"></a>
 ```python
 record_har_path: str | Path | None = None
 ```
 Path to save `.har` network trace files. [Playwright Docs: `record_har_path`](https://playwright.dev/python/docs/api/class-browsertype#browser-type-launch-persistent-context-option-record-har-path)
 <Note>
 This parameter also has an alias `save_har_path` for backwards compatibility with past versions, but we recommend using the standard Playwright name `record_har_path` going forward.
 </Note>
 #### `record_har_content`
 ```python
 record_har_content: RecordHarContent = 'embed'
 ```
 How to persist HAR content: `'omit'`, `'embed'`, `'attach'`
 #### `record_har_mode`
 ```python
 record_har_mode: RecordHarMode = 'full'
 ```
 HAR recording mode: `'full'`, `'minimal'`
 #### `record_har_omit_content`
 ```python
 record_har_omit_content: bool = False
 ```
 Whether to omit request content from the HAR.
 #### `record_har_url_filter`
 ```python
 record_har_url_filter: str | Pattern | None = None
 ```
 URL filter for HAR recording.
 #### `downloads_path`
 ```python
 downloads_path: str | Path | None = '~/.config/browseruse/downloads'
 ```
 (aliases: `downloads_dir`, `save_downloads_path`)
 Local filesystem directory to save browser file downloads to.
 #### `traces_dir`
 <a name="traces-dir"></a>
 <a name="trace-path"></a>
 ```python
 traces_dir: str | Path | None = None
 ```
 Directory to save all-in-one trace files. Files are automatically named as `{traces_dir}/{context_id}.zip`. [Playwright Docs: `traces_dir`](https://playwright.dev/python/docs/api/class-browsertype#browser-type-launch-persistent-context-option-traces-dir)
 <Note>
 This parameter also has an alias `trace_path` for backwards compatibility with past versions, but we recommend using the standard Playwright name `traces_dir` going forward.
 </Note>
 #### `handle_sighup`
 ```python
 handle_sighup: bool = True
 ```
 Whether playwright should swallow SIGHUP signals and kill the browser.
 #### `handle_sigint`
 ```python
 handle_sigint: bool = False
 ```
 Whether playwright should swallow SIGINT signals and kill the browser.
 #### `handle_sigterm`
 ```python
 handle_sigterm: bool = False
 ```
 Whether playwright should swallow SIGTERM signals and kill the browser.
 ---
 ## Full Example
 ```python
 from browser_use import BrowserSession, BrowserProfile, Agent
 browser_profile = BrowserProfile(
    headless=False,
    storage_state="path/to/storage_state.json",
    wait_for_network_idle_page_load_time=3.0,
    viewport={"width": 1280, "height": 1100},
    locale='en-US',
    user_agent='Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/85.0.4183.102 Safari/537.36',
    highlight_elements=True,
    viewport_expansion=500,
    allowed_domains=['*.google.com', 'http*://*.wikipedia.org'],
    user_data_dir=None,
 )
 browser_session = BrowserSession(
    browser_profile=browser_profile,
    headless=True,                          # extra kwargs to the session override the defaults in the profile
 )
 # you can drive a session without the agent / reuse it between agents
 await browser_session.start()
 page = await browser_session.get_current_page()
 await page.goto('https://example.com/first/page')
 async def run_search():
    agent = Agent(
        task='Your task',
        llm=llm,
        page=page,                        # optional: pass a specific playwright page to start on
        browser_session=browser_session,  # optional: pass an existing browser session to an agent
    )
 ```
 ---
 ## Summary
 - **BrowserSession** (defined in `browser_use/browser/session.py`) handles the live browser connection and runtime state
 - **BrowserProfile** (defined in `browser_use/browser/profile.py`) is a template that can store default config parameters for a `BrowserSession(...)`
 Configuration parameters defined in both scopes consumed by these calls depending on whether we're connecting/launching:
 - `BrowserConnectArgs` - args for `playwright.BrowserType.connect_over_cdp(...)`
 - `BrowserLaunchArgs` - args for `playwright.BrowserType.launch(...)`
 - `BrowserNewContextArgs` - args for `playwright.BrowserType.new_context(...)`
 - `BrowserLaunchPersistentContextArgs` - args for `playwright.BrowserType.launch_persistent_context(...)`
 - Browser Use's own internal methods
 For more details on Playwright's browser context options, see their [launch args documentation](https://playwright.dev/python/docs/api/class-browsertype#browser-type-launch-persistent-context).
 ---
--- a/.github/instructions/browser-use.instructions.md
+++ b/.github/instructions/browser-use.instructions.md
@ -1,82 +0,0 @@
 ---
 applyTo: '**'
 ---
 ## 🧠 General Guidelines for Contributing to `browser-use`
 **Browser-Use** is an AI agent that autonomously interacts with the web. It takes a user-defined task, navigates web pages using Chromium via Playwright, processes HTML, and repeatedly queries a language model (like `gpt-4o`) to decide the next action—until the task is completed.
 ### 🗂️ File Documentation
 When you create a **new file**:
 * **For humans**: At the top of the file, include a docstring in natural language explaining:
  * What this file does.
  * How it fits into the browser-use system.
  * If it introduces a new abstraction or replaces an old one.
 * **For LLMs/AI**: Include structured metadata using standardized comments such as:
  ```python
  # @file purpose: Defines <purpose>
  ```
 ---
 ### 🧰 Development Rules
 * ✅ **Always use [`uv`](mdc:https:/github.com/astral-sh/uv) instead of `pip`**
  For deterministic and fast dependency installs.
 ```bash
 uv venv --python 3.11
 source .venv/bin/activate
 uv sync
 ```
 * ✅ **Use real model names**
  Do **not** replace `gpt-4o` with `gpt-4`. The model `gpt-4o` is a distinct release and supported.
 * ✅ **Type-safe coding**
  Use **Pydantic v2 models** for all internal action schemas, task inputs/outputs, and controller I/O. This ensures robust validation and LLM-call integrity.
 ---
 ## ⚙️ Adding New Actions
 To add a new action that your browser agent can execute:
 ```python
 from playwright.async_api import Page
 from browser_use.core.controller import Controller, ActionResult
 controller = Controller()
@controller.registry.action("Search the web for a specific query")
 async def search_web(query: str, page: Page):
    # Implement your logic here, e.g., query a search engine and return results
    result = ...
    return ActionResult(extracted_content=result, include_in_memory=True)
 ```
 ### Notes:
 * Use descriptive names and docstrings for each action.
 * Prefer returning `ActionResult` with structured content to help the agent reason better.
 ---
 ## 🧠 Creating and Running an Agent
 To define a task and run a browser-use agent:
 ```python
 from browser_use import Agent
 from langchain.chat_models import ChatOpenAI
 task = "Find the CEO of OpenAI and return their name"
 model = ChatOpenAI(model="gpt-4o")
 agent = Agent(task=task, llm=model, controller=controller)
 history = await agent.run()
 ```
--- a/.github/instructions/custom-functions.instructions.md
+++ b/.github/instructions/custom-functions.instructions.md
@ -1,249 +0,0 @@
 ---
 description: "Extend default agent and write custom action functions to do certain tasks"
 applyTo: '**'
 ---
 Custom actions are functions *you* provide, that are added to our [default actions](https://github.com/browser-use/browser-use/blob/main/browser_use/controller/service.py) the agent can use to accomplish tasks.
 Action functions can request [arbitrary parameters](#action-parameters-via-pydantic-model) that the LLM has to come up with + a fixed set of [framework-provided arguments](#framework-provided-parameters) for browser APIs / `Agent(context=...)` / etc.
 <Note>
  Our default set of actions is already quite powerful, the built-in `Controller` provides basics like `open_tab`, `scroll_down`, `extract_content`, [and more](https://github.com/browser-use/browser-use/blob/main/browser_use/controller/service.py).
 </Note>
 It's easy to add your own actions to implement additional custom behaviors, integrations with other apps, or performance optimizations.
 For examples of custom actions (e.g. uploading files, asking a human-in-the-loop for help, drawing a polygon with the mouse, and more), see [examples/custom-functions](https://github.com/browser-use/browser-use/tree/main/examples/custom-functions).
 ## Action Function Registration
 To register your own custom functions (which can be `sync` or `async`), decorate them with the `@controller.action(...)` decorator. This saves them into the `controller.registry`.
 ```python
 from browser_use import Controller, ActionResult
 controller = Controller()
@controller.action('Ask human for help with a question', domains=['example.com'])   # pass allowed_domains= or page_filter= to limit actions to certain pages
 def ask_human(question: str) -> ActionResult:
    answer = input(f'{question} > ')
    return ActionResult(extracted_content=f'The human responded with: {answer}', include_in_memory=True)
 ```
 ```python
 # Then pass your controller to the agent to use it
 agent = Agent(
    task='...',
    llm=llm,
    controller=controller,
 )
 ```
 <Note>
  Keep your action function names and descriptions short and concise:
  - The LLM chooses between actions to run solely based on the function name and description
  - The LLM decides how to fill action params based on their names, type hints, & defaults
 </Note>
 ---
 ## Action Parameters
 Browser Use supports two patterns for defining action parameters: normal function arguments, or a Pydantic model.
 ### Function Arguments
 For simple actions that don't need default values, you can define the action parameters directly as arguments to the function. This one takes a single string argument, `css_selector`.
 When the LLM calls an action, it sees its argument names & types, and will provide values that fit.
 ```python
@controller.action('Click element')
 def click_element(css_selector: str, page: Page) -> ActionResult:
    # css_selector is an action param the LLM must provide when calling
    # page is a special framework-provided param to access the browser APIs (see below)
    await page.locator(css_selector).click()
    return ActionResult(extracted_content=f"Clicked element {css_selector}")
 ```
 ### Pydantic Model
 You can define a pydantic model for the parameters your action expects by setting a `@controller.action(..., param_model=MyParams)`.
 This allows you to use optional parameters, default values, `Annotated[...]` types with custom validation, field descriptions, and other features offered by pydantic.
 When the agent calls calls your agent function, an instance of your model with the values filled by the LLM will be passed as the argument named `params` to your action function.
 Using a pydantic model is helpful because it allows more flexibility and power to enforce the schema of the values the LLM should provide.
 The LLM gets the entire pydantic JSON schema for your `param_model`, it will see the function name & description + individual field names, types, descriptions, and default values.
 ```python
 from typing import Annotated
 from pydantic import BaseModel, AfterValidator
 from browser_use import ActionResult
 class MyParams(BaseModel):
    field1: int
    field2: str = 'default value'
    field3: Annotated[str, AfterValidator(lambda s: s.lower())]  # example: enforce always lowercase
    field4: str = Field(default='abc', description='Detailed description for the LLM')
@controller.action('My action', param_model=MyParams)
 def my_action(params: MyParams, page: Page) -> ActionResult:
    await page.keyboard.type(params.field2)
    return ActionResult(extracted_content=f"Inputted {params} on {page.url}")
 ```
 Any special framework-provided arguments (e.g. `page`) will be passed as separate positional arguments after `params`.
 <Important>
 To use a `BaseModel` the arg *must* be called `params`. Action function args are matched and filled like named arguments; arg order doesn't matter but names and types do.
 </Important>
 ### Framework-Provided Parameters
 These special action parameters are injected by the `Controller` and are passed as extra args to any actions that expect them.
 For example, actions that need to run playwright code to interact with the browser should take the argument `page` or `browser_session`.
 - `page: Page` - The current Playwright page (shortcut for `browser_session.get_current_page()`)
 - `browser_session: BrowserSession` - The current browser session (and playwright context via `browser_session.browser_context`)
 - `context: AgentContext` - Any optional top-level context object passed to the Agent, e.g. `Agent(context=user_provided_obj)`
 - `page_extraction_llm: BaseChatModel` - LLM instance used for page content extraction
 - `available_file_paths: list[str]` - List of available file paths for upload / processing
 - `has_sensitive_data: bool` - Whether the action content contains sensitive data markers (check this to avoid logging sensitive data to terminal by accident)
 #### Example: Action uses the current `page`
 ```python
 from playwright.async_api import Page
 from browser_use import Controller, ActionResult
 controller = Controller()
@controller.action('Type keyboard input into a page')
 async def input_text_into_page(text: str, page: Page) -> ActionResult:
    await page.keyboard.type(text)
    return ActionResult(extracted_content='Website opened')
 ```
 #### Example: Action uses the `browser_context`
 ```python
 from browser_use import BrowserSession, Controller, ActionResult
 controller = Controller()
@controller.action('Open website')
 async def open_website(url: str, browser_session: BrowserSession) -> ActionResult:
    # find matching existing tab by looking through all pages in playwright browser_context
    all_tabs = await browser_session.browser_context.pages
    for tab in all_tabs:
        if tab.url == url:
            await tab.bring_to_foreground()
            return ActionResult(extracted_content=f'Switched to tab with url {url}')
    # otherwise, create a new tab
    new_tab = await browser_session.browser_context.new_page()
    await new_tab.goto(url)
    return ActionResult(extracted_content=f'Opened new tab with url {url}')
 ```
 ---
 ## Important Rules
 1. **Return an [`ActionResult`](https://github.com/search?q=repo%3Abrowser-use%2Fbrowser-use+%22class+ActionResult%28BaseModel%29%22&type=code)**: All actions should return an `ActionResult | str | None`. The stringified version of the result is passed back to the LLM, and optionally persisted in the long-term memory when `ActionResult(..., include_in_memory=True)`.
 2. **Type hints on arguments are required**: They are used to verify that action params don't conflict with special arguments injected by the controller (e.g. `page`)
 3. **Actions functions called directly must be passed kwargs**: When calling actions from other actions or python code, you must **pass all parameters as kwargs only**, even though the actions are usually defined using positional args (for the same reasons as [pluggy](https://pluggy.readthedocs.io/en/stable/index.html#calling-hooks)).
    Action arguments are always matched by name and type, **not** positional order, so this helps prevent ambiguity / reordering issues while keeping action signatures short.
    ```python
    @controller.action('Fill in the country form field')
    def input_country_field(country: str, page: Page) -> ActionResult:
        await some_action(123, page=page)                                # ❌ not allowed: positional args, use kwarg syntax when calling
        await some_action(abc=123, page=page)                            # ✅ allowed: action params & special kwargs
        await some_other_action(params=OtherAction(abc=123), page=page)  # ✅ allowed: params=model & special kwargs
    ```
 ```python
 # Using Pydantic Model to define action params (recommended)
 class PinCodeParams(BaseModel):
    code: int
    retries: int = 3                                               # ✅ supports optional/defaults
@controller.action('...', param_model=PinCodeParams)
 async def input_pin_code(params: PinCodeParams, page: Page): ...   # ✅ special params at the end
 # Using function arguments to define action params
 async def input_pin_code(code: int, retries: int, page: Page): ... # ✅ params first, special params second, no defaults
 async def input_pin_code(code: int, retries: int=3): ...           # ✅ defaults ok only if no special params needed
 async def input_pin_code(code: int, retries: int=3, page: Page): ... # ❌ Python SyntaxError! not allowed
 ```
 ---
 ## Reusing Custom Actions Across Agents
 You can use the same controller for multiple agents.
 ```python
 controller = Controller()
 # ... register actions to the controller
 agent = Agent(
    task="Go to website X and find the latest news",
    llm=llm,
    controller=controller
 )
 # Run the agent
 await agent.run()
 agent2 = Agent(
    task="Go to website Y and find the latest news",
    llm=llm,
    controller=controller
 )
 await agent2.run()
 ```
 <Note>
  The controller is stateless and can be used to register multiple actions and
  multiple agents.
 </Note>
 ## Exclude functions
 If you want to exclude some registered actions and make them unavailable to the agent, you can do:
 ```python
 controller = Controller(exclude_actions=['open_tab', 'search_google'])
 agent = Agent(controller=controller, ...)
 ```
 If you want actions to only be available on certain pages, and to not tell the LLM about them on other pages,
 you can use the `allowed_domains` and `page_filter`:
 ```python
 from pydantic import BaseModel
 from browser_use import Controller, ActionResult
 controller = Controller()
 async def is_ai_allowed(page: Page):
    if api.some_service.check_url(page.url):
        logger.warning('Allowing AI agent to visit url:', page.url)
        return True
    return False
@controller.action('Fill out secret_form', allowed_domains=['https://*.example.com'], page_filter=is_ai_allowed)
 def fill_out_form(...) -> ActionResult:
    ... will only be runnable by LLM on pages that match https://*.example.com *AND* where is_ai_allowed(page) returns True
 ```
--- a/.github/instructions/hooks.instructions.md
+++ b/.github/instructions/hooks.instructions.md
@ -1,381 +0,0 @@
 ---
 description: "Customize agent behavior with lifecycle hooks"
 applyTo: '**'
 ---
 Browser-Use provides lifecycle hooks that allow you to execute custom code at specific points during the agent's execution.
 Hook functions can be used to read and modify agent state while running, implement custom logic, change configuration, integrate the Agent with external applications.
 ## Available Hooks
 Currently, Browser-Use provides the following hooks:
 | Hook | Description | When it's called |
 | ---- | ----------- | ---------------- |
 | `on_step_start` | Executed at the beginning of each agent step | Before the agent processes the current state and decides on the next action |
 | `on_step_end` | Executed at the end of each agent step | After the agent has executed all the actions for the current step, before it starts the next step |
 ```python
 await agent.run(on_step_start=..., on_step_end=...)
 ```
 Each hook should be an `async` callable function that accepts the `agent` instance as its only parameter.
 ### Basic Example
 ```python
 from browser_use import Agent
 from langchain_openai import ChatOpenAI
 async def my_step_hook(agent: Agent):
    # inside a hook you can access all the state and methods under the Agent object:
    #   agent.settings, agent.state, agent.task
    #   agent.controller, agent.llm, agent.browser_session
    #   agent.pause(), agent.resume(), agent.add_new_task(...), etc.
    # You also have direct access to the playwright Page and Browser Context
    page = await agent.browser_session.get_current_page()
    #   https://playwright.dev/python/docs/api/class-page
    current_url = page.url
    visit_log = agent.state.history.urls()
    previous_url = visit_log[-2] if len(visit_log) >= 2 else None
    print(f"Agent was last on URL: {previous_url} and is now on {current_url}")
    # Example: listen for events on the page, interact with the DOM, run JS directly, etc.
    await page.on('domcontentloaded', lambda: print('page navigated to a new url...'))
    await page.locator("css=form > input[type=submit]").click()
    await page.evaluate('() => alert(1)')
    await page.browser.new_tab
    await agent.browser_session.session.context.add_init_script('/* some JS to run on every page */')
    # Example: monitor or intercept all network requests
    async def handle_request(route):
 		# Print, modify, block, etc. do anything to the requests here
        #   https://playwright.dev/python/docs/network#handle-requests
 		print(route.request, route.request.headers)
 		await route.continue_(headers=route.request.headers)
 	await page.route("**/*", handle_route)
    # Example: pause agent execution and resume it based on some custom code
    if '/completed' in current_url:
        agent.pause()
        Path('result.txt').write_text(await page.content()) 
        input('Saved "completed" page content to result.txt, press [Enter] to resume...')
        agent.resume()
 agent = Agent(
    task="Search for the latest news about AI",
    llm=ChatOpenAI(model="gpt-4o"),
 )
 await agent.run(
    on_step_start=my_step_hook,
    # on_step_end=...
    max_steps=10
 )
 ```
 ## Data Available in Hooks
 When working with agent hooks, you have access to the entire `Agent` instance. Here are some useful data points you can access:
 - `agent.task` lets you see what the main task is, `agent.add_new_task(...)` lets you queue up a new one
 - `agent.controller` give access to the `Controller()` object and `Registry()` containing the available actions
    - `agent.controller.registry.execute_action('click_element_by_index', {'index': 123}, browser_session=agent.browser_session)`
 - `agent.context` lets you access any user-provided context object passed in to `Agent(context=...)`
 - `agent.sensitive_data` contains the sensitive data dict, which can be updated in-place to add/remove/modify items
 - `agent.settings` contains all the configuration options passed to the `Agent(...)` at init time
 - `agent.llm` gives direct access to the main LLM object (e.g. `ChatOpenAI`)
 - `agent.state` gives access to lots of internal state, including agent thoughts, outputs, actions, etc.
    - `agent.state.history.model_thoughts()`: Reasoning from Browser Use's model.
    - `agent.state.history.model_outputs()`: Raw outputs from the Browsre Use's model.
    - `agent.state.history.model_actions()`: Actions taken by the agent
    - `agent.state.history.extracted_content()`: Content extracted from web pages
    - `agent.state.history.urls()`: URLs visited by the agent
 - `agent.browser_session` gives direct access to the `BrowserSession()` and playwright objects
    - `agent.browser_session.get_current_page()`: Get the current playwright `Page` object the agent is focused on
    - `agent.browser_session.browser_context`: Get the current playwright `BrowserContext` object
    - `agent.browser_session.browser_context.pages`: Get all the tabs currently open in the context
    - `agent.browser_session.get_page_html()`: Current page HTML
    - `agent.browser_session.take_screenshot()`: Screenshot of the current page
 ## Tips for Using Hooks
 - **Avoid blocking operations**: Since hooks run in the same execution thread as the agent, try to keep them efficient or use asynchronous patterns.
 - **Handle exceptions**: Make sure your hook functions handle exceptions gracefully to prevent interrupting the agent's main flow.
 - **Use custom actions instead**: hooks are fairly advanced, most things can be implemented with [custom action functions](/customize/custom-functions) instead
 ---
 ## Complex Example: Agent Activity Recording System
 This comprehensive example demonstrates a complete implementation for recording and saving Browser-Use agent activity, consisting of both server and client components.
 ### Setup Instructions
 To use this example, you'll need to:
 1. Set up the required dependencies:
   ```bash
   pip install fastapi uvicorn prettyprinter pyobjtojson dotenv browser-use langchain-openai
   ```
 2. Create two separate Python files:
   - `api.py` - The FastAPI server component
   - `client.py` - The Browser-Use agent with recording hook
 3. Run both components:
   - Start the API server first: `python api.py`
   - Then run the client: `python client.py`
 ### Server Component (api.py)
 The server component handles receiving and storing the agent's activity data:
 ```python
 #!/usr/bin/env python3
 #
 # FastAPI API to record and save Browser-Use activity data.
 # Save this code to api.py and run with `python api.py`
 # 
 import json
 import base64
 from pathlib import Path
 from fastapi import FastAPI, Request
 import prettyprinter
 import uvicorn
 prettyprinter.install_extras()
 # Utility function to save screenshots
 def b64_to_png(b64_string: str, output_file):
    """
    Convert a Base64-encoded string to a PNG file.
    :param b64_string: A string containing Base64-encoded data
    :param output_file: The path to the output PNG file
    """
    with open(output_file, "wb") as f:
        f.write(base64.b64decode(b64_string))
 # Initialize FastAPI app
 app = FastAPI()
@app.post("/post_agent_history_step")
 async def post_agent_history_step(request: Request):
    data = await request.json()
    prettyprinter.cpprint(data)
    # Ensure the "recordings" folder exists using pathlib
    recordings_folder = Path("recordings")
    recordings_folder.mkdir(exist_ok=True)
    # Determine the next file number by examining existing .json files
    existing_numbers = []
    for item in recordings_folder.iterdir():
        if item.is_file() and item.suffix == ".json":
            try:
                file_num = int(item.stem)
                existing_numbers.append(file_num)
            except ValueError:
                # In case the file name isn't just a number
                pass
    if existing_numbers:
        next_number = max(existing_numbers) + 1
    else:
        next_number = 1
    # Construct the file path
    file_path = recordings_folder / f"{next_number}.json"
    # Save the JSON data to the file
    with file_path.open("w") as f:
        json.dump(data, f, indent=2)
    # Optionally save screenshot if needed
    # if "website_screenshot" in data and data["website_screenshot"]:
    #     screenshot_folder = Path("screenshots")
    #     screenshot_folder.mkdir(exist_ok=True)
    #     b64_to_png(data["website_screenshot"], screenshot_folder / f"{next_number}.png")
    return {"status": "ok", "message": f"Saved to {file_path}"}
 if __name__ == "__main__":
    print("Starting Browser-Use recording API on http://0.0.0.0:9000")
    uvicorn.run(app, host="0.0.0.0", port=9000)
 ```
 ### Client Component (client.py)
 The client component runs the Browser-Use agent with a recording hook:
 ```python
 #!/usr/bin/env python3
 #
 # Client to record and save Browser-Use activity.
 # Save this code to client.py and run with `python client.py`
 #
 import asyncio
 import requests
 from dotenv import load_dotenv
 from pyobjtojson import obj_to_json
 from langchain_openai import ChatOpenAI
 from browser_use import Agent
 # Load environment variables (for API keys)
 load_dotenv()
 def send_agent_history_step(data):
    """Send the agent step data to the recording API"""
    url = "http://127.0.0.1:9000/post_agent_history_step"
    response = requests.post(url, json=data)
    return response.json()
 async def record_activity(agent_obj):
    """Hook function that captures and records agent activity at each step"""
    website_html = None
    website_screenshot = None
    urls_json_last_elem = None
    model_thoughts_last_elem = None
    model_outputs_json_last_elem = None
    model_actions_json_last_elem = None
    extracted_content_json_last_elem = None
    print('--- ON_STEP_START HOOK ---')
    # Capture current page state
    website_html = await agent_obj.browser_session.get_page_html()
    website_screenshot = await agent_obj.browser_session.take_screenshot()
    # Make sure we have state history
    if hasattr(agent_obj, "state"):
        history = agent_obj.state.history
    else:
        history = None
        print("Warning: Agent has no state history")
        return
    # Process model thoughts
    model_thoughts = obj_to_json(
        obj=history.model_thoughts(),
        check_circular=False
    )
    if len(model_thoughts) > 0:
        model_thoughts_last_elem = model_thoughts[-1]
    # Process model outputs
    model_outputs = agent_obj.state.history.model_outputs()
    model_outputs_json = obj_to_json(
        obj=model_outputs,
        check_circular=False
    )
    if len(model_outputs_json) > 0:
        model_outputs_json_last_elem = model_outputs_json[-1]
    # Process model actions
    model_actions = agent_obj.state.history.model_actions()
    model_actions_json = obj_to_json(
        obj=model_actions,
        check_circular=False
    )
    if len(model_actions_json) > 0:
        model_actions_json_last_elem = model_actions_json[-1]
    # Process extracted content
    extracted_content = agent_obj.state.history.extracted_content()
    extracted_content_json = obj_to_json(
        obj=extracted_content,
        check_circular=False
    )
    if len(extracted_content_json) > 0:
        extracted_content_json_last_elem = extracted_content_json[-1]
    # Process URLs
    urls = agent_obj.state.history.urls()
    urls_json = obj_to_json(
        obj=urls,
        check_circular=False
    )
    if len(urls_json) > 0:
        urls_json_last_elem = urls_json[-1]
    # Create a summary of all data for this step
    model_step_summary = {
        "website_html": website_html,
        "website_screenshot": website_screenshot,
        "url": urls_json_last_elem,
        "model_thoughts": model_thoughts_last_elem,
        "model_outputs": model_outputs_json_last_elem,
        "model_actions": model_actions_json_last_elem,
        "extracted_content": extracted_content_json_last_elem
    }
    print("--- MODEL STEP SUMMARY ---")
    print(f"URL: {urls_json_last_elem}")
    # Send data to the API
    result = send_agent_history_step(data=model_step_summary)
    print(f"Recording API response: {result}")
 async def run_agent():
    """Run the Browser-Use agent with the recording hook"""
    agent = Agent(
        task="Compare the price of gpt-4o and DeepSeek-V3",
        llm=ChatOpenAI(model="gpt-4o"),
    )
    try:
        print("Starting Browser-Use agent with recording hook")
        await agent.run(
            on_step_start=record_activity,
            max_steps=30
        )
    except Exception as e:
        print(f"Error running agent: {e}")
 if __name__ == "__main__":
    # Check if API is running
    try:
        requests.get("http://127.0.0.1:9000")
        print("Recording API is available")
    except:
        print("Warning: Recording API may not be running. Start api.py first.")
    # Run the agent
    asyncio.run(run_agent())
 ```
 Contribution by Carlos A. Planchón.
 ### Working with the Recorded Data
 After running the agent, you'll find the recorded data in the `recordings` directory. Here's how you can use this data:
 1. **View recorded sessions**: Each JSON file contains a snapshot of agent activity for one step
 2. **Extract screenshots**: You can modify the API to save screenshots separately
 3. **Analyze agent behavior**: Use the recorded data to study how the agent navigates websites
 ### Extending the Example
 You can extend this recording system in several ways:
 1. **Save screenshots separately**: Uncomment the screenshot saving code in the API
 2. **Add a web dashboard**: Create a simple web interface to view recorded sessions
 3. **Add session IDs**: Modify the API to group steps by agent session
 4. **Add filtering**: Implement filters to record only specific types of actions
--- a/.github/instructions/output-format.instructions.md
+++ b/.github/instructions/output-format.instructions.md
@ -1,49 +0,0 @@
 ---
 description: "The default is text. But you can define a structured output format to make post-processing easier."
 applyTo: '**'
 ---
 ## Custom output format
 With [this example](https://github.com/browser-use/browser-use/blob/main/examples/features/custom_output.py) you can define what output format the agent should return to you.
 ```python
 from pydantic import BaseModel
 # Define the output format as a Pydantic model
 class Post(BaseModel):
 	post_title: str
 	post_url: str
 	num_comments: int
 	hours_since_post: int
 class Posts(BaseModel):
 	posts: List[Post]
 controller = Controller(output_model=Posts)
 async def main():
 	task = 'Go to hackernews show hn and give me the first  5 posts'
 	model = ChatOpenAI(model='gpt-4o')
 	agent = Agent(task=task, llm=model, controller=controller)
 	history = await agent.run()
 	result = history.final_result()
 	if result:
 		parsed: Posts = Posts.model_validate_json(result)
 		for post in parsed.posts:
 			print('\n--------------------------------')
 			print(f'Title:            {post.post_title}')
 			print(f'URL:              {post.post_url}')
 			print(f'Comments:         {post.num_comments}')
 			print(f'Hours since post: {post.hours_since_post}')
 	else:
 		print('No result')
 if __name__ == '__main__':
 	asyncio.run(main())
 ```
--- a/.github/instructions/real-browser.instructions.md
+++ b/.github/instructions/real-browser.instructions.md
@ -1,414 +0,0 @@
 ---
 description: "Connect to a remote browser or launch a new local browser."
 applyTo: '**'
 ---
 ## Overview
 Browser Use supports a wide variety of ways to launch or connect to a browser:
 - Launch a new local browser using playwright/patchright chromium (the default)
 - Connect to a remote browser using CDP or WSS
 - Use an existing playwright `Page`, `Browser`, or `BrowserContext` object
 - Connect to a local browser already running using `browser_pid`
 <Tip>
 Don't want to manage your own browser infrastructure? Try [☁️ Browser Use Cloud](https://browser-use.com) ➡️  
 We provide automatic CAPTCHA solving, proxies, human-in-the-loop automation, and more!
 </Tip>
 ## Connection Methods
 ### Method A: Launch a New Local Browser (Default)
 Launch a local browser using built-in default (playwright `chromium`) or a provided `executable_path`:
 ```python
 from browser_use import Agent, BrowserSession
 # If no executable_path provided, uses Playwright/Patchright's built-in Chromium
 browser_session = BrowserSession(
    # Path to a specific Chromium-based executable (optional)
    executable_path='/Applications/Google Chrome.app/Contents/MacOS/Google Chrome',  # macOS
    # For Windows: 'C:\\Program Files\\Google\\Chrome\\Application\\chrome.exe'
    # For Linux: '/usr/bin/google-chrome'
    # Use a specific data directory on disk (optional, set to None for incognito)
    user_data_dir='~/.config/browseruse/profiles/default',   # this is the default
    # ... any other BrowserProfile or playwright launch_persistnet_context config...
    # headless=False,
 )
 agent = Agent(
    task="Your task here",
    llm=llm,
    browser_session=browser_session,
 )
 ```
 We support most `chromium`-based browsers in `executable_path`, including [Brave](https://github.com/browser-use/browser-use/tree/main/examples/browser/stealth.py), [patchright chromium](https://github.com/Kaliiiiiiiiii-Vinyzu/patchright), [rebrowser](https://rebrowser.net/), Edge, and more. See [`examples/browser/stealth.py`](https://github.com/browser-use/browser-use/tree/main/examples/browser) for more. We do not support Firefox or Safari at the moment.
 <Warning>
    [As of Chrome v136](https://github.com/browser-use/browser-use/issues/1520), driving browsers with the default profile is [no longer supported](https://developer.chrome.com/blog/remote-debugging-port) for security reasons. Browser-Use has transitioned to creating a new dedicated profile for agents in: `~/.config/browseruse/profiles/default`. You can [open this profile](https://superuser.com/questions/377186/how-do-i-start-chrome-using-a-specified-user-profile) and log into everything you need your agent to have access to, and it will persist over time.
 </Warning>
 ### Method B: Connect Using Existing Playwright Objects
 Pass existing Playwright `Page`, `BrowserContext`, `Browser`, and/or `playwright` API object to `BrowserSession(...)`:
 ```python
 from browser_use import Agent, BrowserSession
 from playwright.async_api import async_playwright
 # from patchright.async_api import async_playwright   # stealth alternative
 async with async_playwright() as playwright:
    browser = await playwright.chromium.launch()
    context = await browser.new_context()
    page = await context.new_page()
    browser_session = BrowserSession(
        page=page,
        # browser_context=context,  # all these are supported
        # browser=browser,
        # playwright=playwright,
    )
    agent = Agent(
        task="Your task here",
        llm=llm,
        browser_session=browser_session,
    )
 ```
 You can also pass `page` directly to `Agent(...)` as a shortcut.
 ```python
 agent = Agent(
    task="Your task here",
    llm=llm,
    page=page,
 )
 ```
 ### Method C: Connect to Local Browser Using Browser PID
 Connect to a browser with open `--remote-debugging-port`:
 ```python
 from browser_use import Agent, BrowserSession
 # First, start Chrome with remote debugging:
 # /Applications/Google Chrome.app/Contents/MacOS/Google Chrome --remote-debugging-port=9242
 # Then connect using the process ID
 browser_session = BrowserSession(browser_pid=12345)  # Replace with actual Chrome PID
 agent = Agent(
    task="Your task here",
    llm=llm,
    browser_session=browser_session,
 )
 ```
 ### Method D: Connect to remote Playwright Node.js Browser Server via WSS URL
 Connect to Playwright Node.js server providers:
 ```python
 from browser_use import Agent, BrowserSession
 # Connect to a playwright server
 browser_session = BrowserSession(wss_url="wss://your-playwright-server.com/ws")
 agent = Agent(
    task="Your task here",
    llm=llm,
    browser_session=browser_session,
 )
 ```
 ### Method E: Connect to Remote Browser via CDP URL
 Connect to any remote Chromium-based browser:
 ```python
 from browser_use import Agent, BrowserSession
 # Connect to Chrome via CDP
 browser_session = BrowserSession(cdp_url="http://localhost:9222")
 agent = Agent(
    task="Your task here",
    llm=llm,
    browser_session=browser_session,
 )
 ```
 ## Security Considerations
 <Warning>
  When using any browser profile, the agent will have access to:
  - All its logged-in sessions and cookies
  - Saved passwords (if autofill is enabled)
  - Browser history and bookmarks
  - Extensions and their data
  Always review the task you're giving to the agent and ensure it aligns with your security requirements!
  Use `Agent(sensitive_data={'https://auth.example.com': {x_key: value}})` for any secrets, and restrict the browser with `BrowserSession(allowed_domains=['https://*.example.com'])`.
 </Warning>
 ## Best Practices
 1. **Use isolated profiles**: Create separate Chrome profiles for different agents to limit scope of risk:
   ```python
   browser_session = BrowserSession(
       user_data_dir='~/.config/browseruse/profiles/banking',
       # profile_directory='Default'
   )
   ```
 2. **Limit domain access**: Restrict which sites the agent can visit:
   ```python
   browser_session = BrowserSession(
       allowed_domains=['example.com', 'http*://*.github.com'],
   )
   ```
 3. **Enable `keep_alive=True`** If you want to use a single `BrowserSession` with more than one agent:
   ```python
   browser_session = BrowserSession(
       keep_alive=True,
       ...
   )
   await browser_session.start()  # start the session yourself before passing to Agent
   ...
   agent = Agent(..., browser_session=browser_session)
   await agent.run()
   ...
   await browser_session.kill()   # end the session yourself, shortcut for keep_alive=False + .stop()
   ```
 ## Re-Using a Browser
 A `BrowserSession` starts when the browser is launched/connected, and ends when the browser process exits/disconnects. A session internally manages a single live playwright browser context, and is normally auto-closed by the agent when its task is complete (*if* the agent started the session itself). If you pass an existing `BrowserSession` into an Agent, or if you set `BrowserSession(keep_alive=True)`, the session will not be closed and can be re-used between agents.
 Browser Use provides a number of ways to re-use profiles, sessions, and other configuration across multiple agents.
 - ✅ sequential agents can re-use a single `user_data_dir` in new `BrowserSession`s
 - ✅ sequential agents can re-use a single `BrowserSession` without closing it
 - ❌ parallel agents cannot run separate `BrowserSession`s using the same `user_data_dir`
 - ✅ parallel agents can run separate `BrowserSession`s using the same `storage_state`
 - ✅ parallel agents can share a single `BrowserSession`, working in different tabs
 - ⚠️ parallel agents can share a single `BrowserSession`, working in the same tab
 <Important>
 Multiple `BrowserSession`s (aka chrome processes) cannot share the same `user_data_dir` at the same time, but they can share a `storage_state` file or `BrowserProfile` config.
 </Important>
 ### Sequential Agents, Same Profile, Different Browser
 If you are only running one agent & browser at a time, they can re-use the same `user_data_dir` sequentially.
 ```python
 from browser_use import Agent, BrowserSession
 from langchain_openai import ChatOpenAI
 reused_profile = BrowserProfile(user_data_dir='~/.config/browseruse/profiles/default')
 agent1 = Agent(
    task="The first task...",
    llm=ChatOpenAI(model="gpt-4o-mini"),
    browser_profile=reused_profile,    # pass the profile in, it will auto-create a session
 )
 await agent1.run()
 agent2 = Agent(
    task="The second task...",
    llm=ChatOpenAI(model="gpt-4o-mini"),
    browser_profile=reused_profile,    # agent will auto-create its own new session
 )
 await agent2.run()
 ```
 > Make sure to never mix different browser versions or `executable_path`s with the same `user_data_dir`. Once run with a newer browser version, some migrations are applied to the dir and older browsers wont be able to read it.
 ### Sequential Agents, Same Profile, Same Browser
 If you are only running one agent at a time, they can re-use the same active `BrowserSession` and avoid having to relaunch chrome.
 Each agent will start off looking at the same tab the last agent ended off on.
 ```python
 from browser_use import Agent, BrowserSession
 from langchain_openai import ChatOpenAI
 reused_session = BrowserSession(
    user_data_dir='~/.config/browseruse/profiles/default',
    keep_alive=True,  # dont close browser after 1st agent.run() ends
 )
 await reused_session.start()   # when keep_alive=True, session must be started manually
 agent1 = Agent(
    task="The first task...",
    llm=ChatOpenAI(model="gpt-4o-mini"),
    browser_session=reused_session,
 )
 await agent1.run()
 agent2 = Agent(
    task="The second task...",
    llm=ChatOpenAI(model="gpt-4o-mini"),
    browser_session=reused_session,      # re-use the same session
 )
 await agent2.run()
 await reused_session.close()
 ```
 ### Parallel Agents, Same Browser, Multiple Tabs
 ```python
 from browser_use import Agent, BrowserSession
 from langchain_openai import ChatOpenAI
 shared_browser = BrowserSession(
    storage_state='/tmp/cookies.json',
    user_data_dir=None,
    keep_alive=True,
    headless=True,
 )
 await shared_browser.start()   # when keep_alive=True, you must start the session yourself
 agent1 = Agent(
    task="The first task...",
    llm=ChatOpenAI(model="gpt-4o-mini"),
    browser_session=shared_browser,              # pass the session in
 )
 agent2 = Agent(
    task="The second task...",
    llm=ChatOpenAI(model="gpt-4o-mini"),
    browser_session=shared_browser,              # re-use the same session
 )
 await asyncio.gather(agent1.run(), agent2.run()) # run in parallel
 await shared_browser.close()
 ```
 ### Parallel Agents, Same Browser, Same Tab
 <Warning>
 ⚠️ This mode is not recommended. Agents are not yet optimized to share the same tab in the same browser, they may interfere with each other or cause errors.
 </Warning>
 ```python
 from browser_use import Agent, BrowserSession
 from langchain_openai import ChatOpenAI
 from playwright.async_api import async_playwright
 playwright = await async_playwright().start()
 browser = await playwright.chromium.launch(headless=True)
 context = await browser.new_context()
 shared_page = await context.new_page()
 await shared_page.goto('https://example.com', wait_until='domcontentloaded')
 shared_session = BrowserSession(page=shared_page, keep_alive=True)
 await shared_session.start()
 agent1 = Agent(
    task="Fill out the form in section A...",
    llm=ChatOpenAI(model="gpt-4o-mini"),
    browser_session=shared_session
 )
 agent2 = Agent(
    task="Fill out the form in section B...",
    llm=ChatOpenAI(model="gpt-4o-mini"),
    browser_session=shared_session,
 )
 await asyncio.gather(agent1.run(), agent2.run()) # run in parallel
 await shared_session.kill()
 ```
 ### Parallel Agents, Same Profile, Different Browsers
 <Tip>
 This mode is the recommended default.
 </Tip>
 To share a single set of configuration or cookies, but still have agents working in their own browser sessions (potentially in parallel), use our provided `BrowserProfile` object.
 The recommended way to re-use cookies and localStorage state between separate parallel sessions is to use the [`storage_state`](https://docs.browser-use.com/customize/browser-settings#storage-state) option.
 ```bash
 # open a browser to log into sites you want the Agent to have access to
 playwright open https://example.com/ --save-storage=/tmp/auth.json
 playwright open https://example.com/ --load-storage=/tmp/auth.json
 ```
 ```python
 from browser_use.browser import BrowserProfile, BrowserSession
 shared_profile = BrowserProfile(
    headless=True,
    user_data_dir=None,               # use dedicated tmp user_data_dir per session
    storage_state='/tmp/auth.json',   # load/save cookies to/from json file
    keep_alive=True,                  # don't close the browser after the agent finishes
 )
 window1 = BrowserSession(browser_profile=profile_a)
 await window1.start()
 agent1 = Agent(browser_session=window1)
 window2 = BrowserSession(browser_profile=profile_a)
 await window2.start()
 agent2 = Agent(browser_session=window2)
 await asyncio.gather(agent1.run(), agent2.run())  # run in parallel
 await window1.save_storage_state()  # write storage state (cookies, localStorage, etc.) to auth.json
 await window2.save_storage_state()  # you must decide when to save manually
 # can also reload the cookies from the file into the active session if they change
 await window1.load_storage_state()
 await window1.close()
 await window2.close()
 ```
 ---
 ## Troubleshooting
 ### Chrome Won't Connect
 If you're having trouble connecting:
 1. **Close all Chrome instances** before trying to launch with a custom profile
 2. **Check if Chrome is running with debugging port**:
   ```bash
   ps aux | grep chrome | grep remote-debugging-port
   ```
 3. **Verify the executable path** is correct for your system
 4. **Check profile permissions** - ensure your user has read/write access
 ### Profile Lock Issues
 If you get a "profile is already in use" error:
 1. Close all Chrome instances
 2. The profile will automatically be unlocked when BrowserSession starts
 3. Alternatively, manually delete the `SingletonLock` file in the profile directory
 <Note>
  For more configuration options, see the [Browser Settings](/customize/browser-settings) documentation.
 </Note>
 ### Profile Version Issues
 The browser version you run must always be equal to or greater than the version used to create the `user_data_dir`.
 If you see errors like `Failed to parse Extensions` when launching, you're likely attempting to run an older browser with an incompatible `user_data_dir` that's already been migrated to a newer Chrome version.
 Playwright ships a version of chromium that's newer than the default stable Google Chrome release channel, so this can happen if you try to use
 a profile created by the default playwright chromium (e.g. `user_data_dir='~/.config/browseruse/profiles/default'`) with an older
 local browser like `executable_path='/Applications/Google Chrome.app/Contents/MacOS/Google Chrome'`.
--- a/.github/instructions/sensitive-data.instructions.md
+++ b/.github/instructions/sensitive-data.instructions.md
@ -1,198 +0,0 @@
 ---
 description: "Handle sensitive information securely and avoid sending PII & passwords to the LLM."
 applyTo: '**'
 ---
 ## Handling Sensitive Data
 When working with sensitive information like passwords or PII, you can use the `Agent(sensitive_data=...)` parameter to provide sensitive strings that the model can use in actions without ever seeing directly.
 ```python
 agent = Agent(
    task='Log into example.com as user x_username with password x_password',
    sensitive_data={
        'https://example.com': {
            'x_username': 'abc@example.com',
            'x_password': 'abc123456',  # 'x_placeholder': '<actual secret value>',
        },
    },
 )
 ```
 <Note>
 You should also configure [`BrowserSession(allowed_domains=...)`](https://docs.browser-use.com/customize/browser-settings#allowed-domains) to prevent the Agent from visiting URLs not needed for the task.
 </Note>
 ### Basic Usage
 Here's a basic example of how to use sensitive data:
 ```python
 from dotenv import load_dotenv
 load_dotenv()
 from langchain_openai import ChatOpenAI
 from browser_use import Agent, BrowserSession
 llm = ChatOpenAI(model='gpt-4o', temperature=0.0)
 # Define sensitive data
 # The LLM will only see placeholder names (x_member_number, x_passphrase), never the actual values
 sensitive_data = {
    'https://*.example.com': {
        'x_member_number': '123235325',
        'x_passphrase': 'abcwe234',
    },
 }
 # Use the placeholder names in your task description
 task = """
 1. go to https://travel.example.com
 2. sign in with your member number x_member_number and private access code x_passphrase
 3. extract today's list of travel deals as JSON
 """
 # Recommended: Limit the domains available for the entire browser so the Agent can't be tricked into visiting untrusted URLs
 browser_session = BrowserSession(allowed_domains=['https://*.example.com'])
 agent = Agent(
    task=task,
    llm=llm,
    sensitive_data=sensitive_data,   # Pass the sensitive data to the agent
    browser_session=browser_session, # Pass the restricted browser_session to limit URLs Agent can visit
    use_vision=False,                # Disable vision or else the LLM might see entered values in screenshots
 )
 async def main():
    await agent.run()
 if __name__ == '__main__':
    asyncio.run(main())
 ```
 In this example:
 1. The LLM only ever sees the `x_member_number` and `x_passphrase` placeholders in prompts
 2. When the model wants to use your password it outputs x_passphrase - and we replace it with the actual value in the DOM
 3. When sensitive data appear in the content of the current page, we replace it in the page summary fed to the LLM - so that the model never has it in its state.
 4. The browser will be entirely prevented from going to any site not under `https://*.example.com`
 This approach ensures that sensitive information remains secure while still allowing the agent to perform tasks that require authentication.
 ---
 ### Best Practices
 - Always restrict your sensitive data to only the exact domains that need it, `https://travel.example.com` is better than `*.example.com`
 - Always restrict [`BrowserSession(allowed_domains=[...])`](https://docs.browser-use.com/customize/browser-settings#allowed-domains) to only the domains the agent needs to visit to accomplish its task. This helps guard against prompt injection attacks, jailbreaks, and LLM mistakes.
 - Only use `sensitive_data` for strings that can be inputted verbatim as text. The LLM never sees the actual values, so it can't "understand" them, adapt them, or split them up for multiple input fields. For example, you can't ask the Agent to click through a datepicker UI to input the sensitive value `1990-12-31`. For these situations you can implement a [custom function](/customize/custom-functions) the LLM can call that updates the DOM using Python / JS.
 - Don't use `sensitive_data` for login credentials, it's better to use [`storage_state`](docs.browser-use.com/customize/browser-settings#storage-state) or a [`user_data_dir`](/customize/browser-settings#user-data-dir) to log into the sites the agent needs in advance & reuse the cookies:  
 ```bash
 # open a browser to log into the sites you need & save the cookies
 $ playwright open https://accounts.google.com --save-storage auth.json
 ```
 Then use those cookies when the agent runs:
 ```python
 agent = Agent(..., browser_session=BrowserSession(storage_state='./auth.json'))
 ```
 <Warning>
 Warning: Vision models still see the screenshot of the page by default - where the sensitive data might be visible.
 It's recommended to set `Agent(use_vision=False)` when working with `sensitive_data`.
 </Warning>
 <a name="allowed_domains"></a>
 <a name="domain-pattern-format"></a>
 ### Allowed Domains
 Domain patterns in `sensitive_data` follow the same format as [`allowed_domains`](https://docs.browser-use.com/customize/browser-settings#allowed-domains):
 - `example.com` - Matches only `https://example.com/*`
 - `*.example.com` - Matches `https://example.com/*` and any subdomain `https://*.example.com/*`
 - `http*://example.com` - Matches both `http://` and `https://` protocols for `example.com/*`
 - `chrome-extension://*` - Matches any Chrome extension URL e.g. `chrome-extension://anyextensionid/options.html`
 > **Security Warning**: For security reasons, certain patterns are explicitly rejected:
 > 
 > - Wildcards in TLD part (e.g., `example.*`) are **not allowed** (`google.*` would match `google.ninja`, `google.pizza`, etc. which is a bad idea)
 > - Embedded wildcards (e.g., `g*e.com`) are rejected to prevent overly broad matches
 > - Multiple wildcards like `*.*.domain` are not supported currently, open an issue if you need this feature
 The default protocol when no scheme is specified is now `https://` for enhanced security.
 For convenience the system will validate that all domain patterns used in `Agent(sensitive_data)` are also included in `BrowserSession(allowed_domains)`.
 ### Missing or Empty Values
 When working with sensitive data, keep these details in mind:
 - If a key referenced by the model (`<secret>key_name</secret>`) is missing from your `sensitive_data` dictionary, a warning will be logged but the substitution tag will be preserved.
 - If you provide an empty value for a key in the `sensitive_data` dictionary, it will be treated the same as a missing key.
 - The system will always attempt to process all valid substitutions, even if some keys are missing or empty.
 ---
 ### Full Example
 Here's a more complex example demonstrating multiple domains and sensitive data values.
 ```python
 from dotenv import load_dotenv
 load_dotenv()
 from langchain_openai import ChatOpenAI
 from browser_use import Agent, BrowserSession
 llm = ChatOpenAI(model='gpt-4o', temperature=0.0)
 # Domain-specific sensitive data
 sensitive_data = {
    'https://*.google.com': {'x_email': '...', 'x_pass': '...'},
    'chrome-extension://abcd1243': {'x_api_key': '...'},
    'http*://example.com': {'x_authcode': '123123'}
 }
 # Set browser session with allowed domains that match all domain patterns in sensitive_data
 browser_session = BrowserSession(
    allowed_domains=[
        'https://*.google.com',
        'chrome-extension://abcd',
        'http://example.com',   # Explicitly include http:// if needed
        'https://example.com'   # By default, only https:// is matched
    ]
 )
 # Pass the sensitive data to the agent
 agent = Agent(
    task="Log into Google, then check my account information",
    llm=llm,
    sensitive_data=sensitive_data,
    browser_session=browser_session,
    use_vision=False,
 )
 async def main():
    await agent.run()
 if __name__ == '__main__':
    asyncio.run(main())
 ```
 With this approach:
 1. The Google credentials (`x_email` and `x_pass`) will only be used on Google domains (any subdomain, https only)
 2. The API key (`x_api_key`) will only be used on pages served by the specific Chrome extension `abcd1243`
 3. The auth code (`x_authcode`) will only be used on `http://example.com/*` or `https://example.com/*`
--- a/.github/instructions/supported-models.instructions.md
+++ b/.github/instructions/supported-models.instructions.md
@ -1,294 +0,0 @@
 ---
 description: "Guide to using different LangChain chat models with Browser Use"
 applyTo: '**'
 ---
 ## Overview
 Browser Use supports various LangChain chat models. Here's how to configure and use the most popular ones. The full list is available in the [LangChain documentation](https://python.langchain.com/docs/integrations/chat/).
 ## Model Recommendations
 We have yet to test performance across all models. Currently, we achieve the best results using GPT-4o with an 89% accuracy on the [WebVoyager Dataset](https://browser-use.com/posts/sota-technical-report). DeepSeek-V3 is 30 times cheaper than GPT-4o. Gemini-2.0-exp is also gaining popularity in the community because it is currently free.
 We also support local models, like Qwen 2.5, but be aware that small models often return the wrong output structure-which lead to parsing errors. We believe that local models will improve significantly this year.
 <Note>
  All models require their respective API keys. Make sure to set them in your
  environment variables before running the agent.
 </Note>
 ## Supported Models
 All LangChain chat models, which support tool-calling are available. We will document the most popular ones here.
 ### OpenAI
 OpenAI's GPT-4o models are recommended for best performance.
 ```python
 from langchain_openai import ChatOpenAI
 from browser_use import Agent
 # Initialize the model
 llm = ChatOpenAI(
    model="gpt-4o",
    temperature=0.0,
 )
 # Create agent with the model
 agent = Agent(
    task="Your task here",
    llm=llm
 )
 ```
 Required environment variables:
 ```bash .env
 OPENAI_API_KEY=
 ```
 ### Anthropic
 ```python
 from langchain_anthropic import ChatAnthropic
 from browser_use import Agent
 # Initialize the model
 llm = ChatAnthropic(
    model_name="claude-3-5-sonnet-20240620",
    temperature=0.0,
    timeout=100, # Increase for complex tasks
 )
 # Create agent with the model
 agent = Agent(
    task="Your task here",
    llm=llm
 )
 ```
 And add the variable:
 ```bash .env
 ANTHROPIC_API_KEY=
 ```
 ### Azure OpenAI
 ```python
 from langchain_openai import AzureChatOpenAI
 from browser_use import Agent
 from pydantic import SecretStr
 import os
 # Initialize the model
 llm = AzureChatOpenAI(
    model="gpt-4o",
    api_version='2024-10-21',
    azure_endpoint=os.getenv('AZURE_OPENAI_ENDPOINT', ''),
    api_key=SecretStr(os.getenv('AZURE_OPENAI_KEY', '')),
 )
 # Create agent with the model
 agent = Agent(
    task="Your task here",
    llm=llm
 )
 ```
 Required environment variables:
 ```bash .env
 AZURE_OPENAI_ENDPOINT=https://your-endpoint.openai.azure.com/
 AZURE_OPENAI_KEY=
 ```
 ### Gemini
 > [!IMPORTANT]
 > `GEMINI_API_KEY` was the old environment var name, it should be called `GOOGLE_API_KEY` as of 2025-05.
 ```python
 from langchain_google_genai import ChatGoogleGenerativeAI
 from browser_use import Agent
 from dotenv import load_dotenv
 # Read GOOGLE_API_KEY into env
 load_dotenv()
 # Initialize the model
 llm = ChatGoogleGenerativeAI(model='gemini-2.0-flash-exp')
 # Create agent with the model
 agent = Agent(
    task="Your task here",
    llm=llm
 )
 ```
 Required environment variables:
 ```bash .env
 GOOGLE_API_KEY=
 ```
 ### DeepSeek-V3
 The community likes DeepSeek-V3 for its low price, no rate limits, open-source nature, and good performance.
 The example is available [here](https://github.com/browser-use/browser-use/blob/main/examples/models/deepseek.py).
 ```python
 from langchain_deepseek import ChatDeepSeek
 from browser_use import Agent
 from pydantic import SecretStr
 from dotenv import load_dotenv
 import os
 load_dotenv()
 api_key = os.getenv("DEEPSEEK_API_KEY")
 # Initialize the model
 llm=ChatDeepSeek(base_url='https://api.deepseek.com/v1', model='deepseek-chat', api_key=SecretStr(api_key))
 # Create agent with the model
 agent = Agent(
    task="Your task here",
    llm=llm,
    use_vision=False
 )
 ```
 Required environment variables:
 ```bash .env
 DEEPSEEK_API_KEY=
 ```
 ### DeepSeek-R1
 We support DeepSeek-R1. Its not fully tested yet, more and more functionality will be added, like e.g. the output of it'sreasoning content.
 The example is available [here](https://github.com/browser-use/browser-use/blob/main/examples/models/deepseek-r1.py).
 It does not support vision. The model is open-source so you could also use it with Ollama, but we have not tested it.
 ```python
 from langchain_deepseek import ChatDeepSeek
 from browser_use import Agent
 from pydantic import SecretStr
 from dotenv import load_dotenv
 import os
 load_dotenv()
 api_key = os.getenv("DEEPSEEK_API_KEY")
 # Initialize the model
 llm=ChatDeepSeek(base_url='https://api.deepseek.com/v1', model='deepseek-reasoner', api_key=SecretStr(api_key))
 # Create agent with the model
 agent = Agent(
    task="Your task here",
    llm=llm,
    use_vision=False
 )
 ```
 Required environment variables:
 ```bash .env
 DEEPSEEK_API_KEY=
 ```
 ### Ollama
 Many users asked for local models. Here they are.
 1. Download Ollama from [here](https://ollama.ai/download)
 2. Run `ollama pull model_name`. Pick a model which supports tool-calling from [here](https://ollama.com/search?c=tools)
 3. Run `ollama start`
 ```python
 from langchain_ollama import ChatOllama
 from browser_use import Agent
 from pydantic import SecretStr
 # Initialize the model
 llm=ChatOllama(model="qwen2.5", num_ctx=32000)
 # Create agent with the model
 agent = Agent(
    task="Your task here",
    llm=llm
 )
 ```
 Required environment variables: None!
 ### Novita AI
 [Novita AI](https://novita.ai) is an LLM API provider that offers a wide range of models. Note: choose a model that supports function calling.
 ```python
 from langchain_openai import ChatOpenAI
 from browser_use import Agent
 from pydantic import SecretStr
 from dotenv import load_dotenv
 import os
 load_dotenv()
 api_key = os.getenv("NOVITA_API_KEY")
 # Initialize the model
 llm = ChatOpenAI(base_url='https://api.novita.ai/v3/openai', model='deepseek/deepseek-v3-0324', api_key=SecretStr(api_key))
 # Create agent with the model
 agent = Agent(
    task="Your task here",
    llm=llm,
    use_vision=False
 )
 ```
 Required environment variables:
 ```bash .env
 NOVITA_API_KEY=
 ```
 ### X AI
 [X AI](https://x.ai) is an LLM API provider that offers a wide range of models. Note: choose a model that supports function calling.
 ```python
 from langchain_openai import ChatOpenAI
 from browser_use import Agent
 from pydantic import SecretStr
 from dotenv import load_dotenv
 import os
 load_dotenv()
 api_key = os.getenv("GROK_API_KEY")
 # Initialize the model
 llm = ChatOpenAI(
    base_url='https://api.x.ai/v1',
    model='grok-3-beta',
    api_key=SecretStr(api_key)
 )
 # Create agent with the model
 agent = Agent(
    task="Your task here",
    llm=llm,
    use_vision=False
 )
 ```
 Required environment variables:
 ```bash .env
 GROK_API_KEY=
 ```
 ## Coming soon
 (We are working on it)
 - Groq
 - Github
 - Fine-tuned models
--- a/.github/instructions/system-prompt.instructions.md
+++ b/.github/instructions/system-prompt.instructions.md
@ -1,76 +0,0 @@
 ---
 description: "Customize the system prompt to control agent behavior and capabilities"
 applyTo: '**'
 ---
 ## Overview
 You can customize the system prompt in two ways:
 1. Extend the default system prompt with additional instructions
 2. Override the default system prompt entirely
 <Note>
  Custom system prompts allow you to modify the agent's behavior at a
  fundamental level. Use this feature carefully as it can significantly impact
  the agent's performance and reliability.
 </Note>
 ### Extend System Prompt (recommended)
 To add additional instructions to the default system prompt:
 ```python
 extend_system_message = """
 REMEMBER the most important RULE:
 ALWAYS open first a new tab and go first to url wikipedia.com no matter the task!!!
 """
 ```
 ### Override System Prompt
 <Warning>
  Not recommended! If you must override the [default system
  prompt](https://github.com/browser-use/browser-use/blob/main/browser_use/agent/system_prompt.md),
  make sure to test the agent yourself.
 </Warning>
 Anyway, to override the default system prompt:
 ```python
 # Define your complete custom prompt
 override_system_message = """
 You are an AI agent that helps users with web browsing tasks.
 [Your complete custom instructions here...]
 """
 # Create agent with custom system prompt
 agent = Agent(
    task="Your task here",
    llm=ChatOpenAI(model='gpt-4'),
    override_system_message=override_system_message
 )
 ```
 ### Extend Planner System Prompt
 You can customize the behavior of the planning agent by extending its system prompt:
 ```python
 extend_planner_system_message = """
 PRIORITIZE gathering information before taking any action.
 Always suggest exploring multiple options before making a decision.
 """
 # Create agent with extended planner system prompt
 llm = ChatOpenAI(model='gpt-4o')
 planner_llm = ChatOpenAI(model='gpt-4o-mini')
 agent = Agent(
 	task="Your task here",
 	llm=llm,
 	planner_llm=planner_llm,
 	extend_planner_system_message=extend_planner_system_message
 )
 ```
--- a/.gitignore
+++ b/.gitignore
@ -12,7 +12,6 @@ oauth_providers.csv
 .venv
 .env
 #.sensitive.json
 log_*.log
 domains.txt
--- a/.sensitive.json
+++ b/.sensitive.json
@ -1,22 +0,0 @@
 {
    "google.com": {
        "x_username": "whs.imnya.ng@gmail.com",
        "x_password": "Vb1Mz9pgjY8JVs"
    },
    "accounts.google.com": {
        "x_username": "whs.imnya.ng@gmail.com",
        "x_password": "Vb1Mz9pgjY8JVs"
    },
    "naver.com": {
        "x_username": "oauth-j93es",
        "x_password": "whs31234"
    },
    "nid.naver.com": {
        "x_username": "oauth-j93es",
        "x_password": "whs31234"
    },
    "github.com": {
        "x_username": "imnyang-bot",
        "x_password": "6PuVXCH9tpQLNm"
    }
 }
--- a/.vscode/settings.json
+++ b/.vscode/settings.json
@ -1,3 +0,0 @@
 {
    "rust-analyzer.initializeStopped": true
 }
--- a/README.md
+++ b/README.md
@ -1,50 +1,30 @@
 # 참고하면 좋을만한 것
 - [ ] 일부 웹사이트는 사용자의 언어에 따라 OAuth 옵션을 바꾸기도 합니다.
 - [ ] https://docs.browser-use.com/customize/custom-functions
 # 환경 설정
-요구 사항
+이 프로젝트는 [uv](https://docs.astral.sh/uv/getting-started/installation/)라는 Python 패키지 관리자를 사용하여 설정해야합니다.
- [uv](https://docs.astral.sh/uv/getting-started/installation/) - Python Package Manager Written by Rust
+또한 [oauth-backend](https://github.com/j93es/oauth-backend)가 설정된 상태여야만 합니다.
 - [oauth-backend](https://github.com/j93es/oauth-backend)
 - [Google Chrome](https://www.google.com/intl/ko_kr/chrome/)
 ---
 > [oauth-backend](https://github.com/j93es/oauth-backend) 프록시를 사용한다면 이 가이드에 따라 인증서 또한 설정되어야만 합니다.
 >
 > 그렇지 않으면 실행되지 않습니다.
 >
 > 윈도우 환경에서는 `sudo certutil -addstore root mitmproxy-ca-cert.cer`로 인증합니다.
 > 
 > Sudo가 활성화되어있지 않은 환경에서는 관리자로 상향된 쉘에서 실행합니다.
 >
 > MacOS 환경에서는 `sudo security add-trusted-cert -d -p ssl -p basic -k /Library/Keychains/System.keychain ~/.mitmproxy/mitmproxy-ca-cert.pem`으로 인증합니다.
 >
 > 다른 플렛폼은 수동으로 설정되어야만 합니다.
 > https://docs.mitmproxy.org/stable/concepts/certificates/
 ---
 uv 설치 후 다음과 같은 명령어를 입력합니다.
-```sh
+```
 uv sync
 ```
 venv와 패키지가 설치가 됩니다.
---
+browser_use가 Playwright에 대한 의존성이 있어 브라우저 설치가 필요합니다
 ~~browser_use가 Playwright에 대한 의존성이 있어 브라우저 설치가 필요합니다~~
 스텔스 기능 때문에 Google Chrome이 필요합니다.
 만약 설치가 되어 있지 않다면 
 ```
-playwright install chrome
+playwright install chromium --with-deps --no-shell
 ```
 ---
 다음과 같은 명령어로 실행합니다.
-```sh
+```
 uv run main.py
 ```
@ -52,41 +32,14 @@ Environment는 .env.example에 따라 설정되어야합니다.
 .env.example을 .env로 복사하여서 사용해주세요.
-# 로그인 방안
+로그인을 수행하지 않을 OAuth Provider는 prompt에서 제거합니다.
 ## 쿠키와 로컬 스토리지 설정 방법 (추천)
 ![1](./docs/image.png)
 ```sh
 uv run playwright open https://google.com/ --save-storage=./data/storage_state.json
 ```
 위 명령어를 실행하면 playwright Browser가 하나 열리는데 여기서 원하는 프로바이더를 모두 로그인 한 후에 브라우저를 정상적으로 닫으면 ./data/storage_state.json 경로에 쿠키, 로컬스토리지를 저장한 파일이 생성됩니다.
 ## Browser Use에게 직접 로그인 요청 (선택)
 <details>
 위에 쿠키와 로컬스토리지 설정 방법과 혼용해서 사용가능합니다.
 `.sensitive.example.json`을 `.sensitive.json`으로 복사해서
 안에 있는 예시 내용을 참고해서 작성해주시면 됩니다.
 더 자세한 내용은 
 [Sensitive Data - Browser Use](https://docs.browser-use.com/customize/sensitive-data)를 참고하시면 좋을 것 같습니다.
 [Sensitive Data - Browser Use](https://docs.browser-use.com/customize/sensitive-data)에서도 권장하지 않는 방법인만큼 애매하긴 하지만 쿠키와 로컬 스토리지를 저장하기 어려운 경우나 일부 flow에서 접근이 어려운 경우 사용해주세요.
 </details>
 # 실행
 domains.txt는 실행시 자동으로 다운로드 됩니다.
 ```sh
 # domains.txt 받기
 curl "https://f.imnya.ng/.whs/tp-domains/data/domains/latest.txt" -o domains.txt
 ```
 ```sh
 # ./run.sh {domains.txt 시작 줄} {domains.txt 끝 줄} {HTML 검사 Skip}
 ./run.sh 12540 13000 False
 ```
@ -96,8 +49,3 @@ curl "https://f.imnya.ng/.whs/tp-domains/data/domains/latest.txt" -o domains.txt
 # ./run.ps1 {domains.txt 시작 줄} {domains.txt 끝 줄} {HTML 검사 Skip}
 ./run.ps1 12540 13000 False
 ```
 # 참고하면 좋을만한 것
 - [ ] 일부 웹사이트는 사용자의 언어에 따라 OAuth 옵션을 바꾸기도 합니다.
 - [ ] https://docs.browser-use.com/customize/custom-functions
--- a/docs/image.png
+++ b/docs/image.png
--- a/lib/agents/init.py
+++ b/lib/agents/init.py
@ -0,0 +1,9 @@
 from lib.agents.get_sso_list import get_sso_list
 # 업데이트될 버전 import 아직 개발 중
 from lib.agents.get_sso_list_v2 import get_sso_list as get_sso_list_v2
 from lib.agents.login_google import login_google
 __all__ = [
    "get_sso_list",
    "login_google",
 ]
--- a/lib/agents/get_sso_list/init.py
+++ b/lib/agents/get_sso_list/init.py
@ -0,0 +1,3 @@
 from lib.agents.get_sso_list.get_sso_list import get_sso_list
 __all__ = ["get_sso_list"]
--- a/lib/agents/get_sso_list/get_sso_list.py
+++ b/lib/agents/get_sso_list/get_sso_list.py
@ -0,0 +1,22 @@
 from lib.agents.get_sso_list.prompt import get_sso_list_task, FindLoginPageResponse
 from lib.browser_use_utils.run_task import run_task
 NOT_FOUND_LOGIN_PAGE = 0
 FOUND_LOGIN_PAGE = 1
 async def get_sso_list(target_url) -> tuple[bool, str | FindLoginPageResponse | None]:
    task = get_sso_list_task
    ReturnModel = FindLoginPageResponse
    success, response = await run_task(target_url, ReturnModel, task)
    if not success:
        return False, response
    if isinstance(response, str):
        return False, response
    return True, response
--- a/lib/agents/get_sso_list/prompt.py
+++ b/lib/agents/get_sso_list/prompt.py
@ -0,0 +1,68 @@
 from pydantic import BaseModel
 class FindLoginPageResponse(BaseModel):
    msg: str | None = None
    url: str | None = None
    sso_list: list[str] = []  # List of SSO providers found on the login page
 get_sso_list_task = """
 You are an expert in finding login pages.
 Your task is to navigate to the login page of the given URL. Follow the steps below strictly and return results only in the specified format.
 ※ You are NOT allowed to navigate to URLs that are not directly discoverable within the initial domain. Do NOT use search engines or guess external login URLs.
 0. INITIAL BLOCK CHECK
 - If the browser is blocked when trying to access the page — due to firewall, CAPTCHA, regional restrictions, or other access denials — immediately terminate the process and return the following JSON:
    ```json
    {
        "msg": "Blocked",
        "url": "",
        "sso_list": []
    }
    ```
 - Do NOT proceed to further steps in this case.
 1. LOGIN PAGE NAVIGATION
 - Navigate only to a **client-side (non-enterprise)** login page within the provided domain.
 - Do NOT rely on external tools, search engines, or links not directly found on the site.
 - If a consent popup (e.g. for privacy/cookies) appears, you MUST dismiss or close it before proceeding.
 - Since step 0 confirmed access, assume the page now loads properly.
 2. SSO BUTTON IDENTIFICATION
 - On the login page, look for the following social login (SSO) buttons:
  - Google, GitHub, Facebook, LinkedIn, Microsoft, Naver, Slack, Etc.
 - ✅ Proceed only if it is clearly an **actual SSO button**.
 - ❌ Exclude the following:
  - Passkey-related buttons
  - Username/password fields
  - Email-based login
  - Non-OAuth methods such as certificate or phone verification
 3. RETURN FORMAT
 - If the login page is successfully found, return:
    ```json
    {
        "msg": "Login page found",
        "url": "https://example.com/login",
        "sso_list": ["Google", "GitHub"]
    }
    ```
 - If the login page cannot be found, return:
    ```json
    {
        "msg": "Login page not found",
        "url": "",
        "sso_list": []
    }
    ```
 - If blocked (as in step 0), return:
    ```json
    {
        "msg": "Blocked",
        "url": "",
        "sso_list": []
    }
    ```
 - Return ONLY the JSON object. Do NOT include any explanation, logging, or extra output.
 """
--- a/lib/agents/get_sso_list_v2/init.py
+++ b/lib/agents/get_sso_list_v2/init.py
@ -0,0 +1,3 @@
 from lib.agents.get_sso_list_v2 import get_sso_list
 __all__ = ["get_sso_list"]
--- a/lib/agents/get_sso_list_v2/get_sso_list.py
+++ b/lib/agents/get_sso_list_v2/get_sso_list.py
@ -0,0 +1,20 @@
 from lib.agents.get_sso_list_v2.prompt import get_sso_list_task, FindLoginPageResponse
 from lib.browser_use_utils.run_task import run_task
 # TODO - Split find login page agent and get SSO list agent
 async def get_sso_list(target_url) -> tuple[bool, str | FindLoginPageResponse | None]:
    task = get_sso_list_task
    ReturnModel = FindLoginPageResponse
    success, response = await run_task(target_url, ReturnModel, task)
    if not success:
        return False, response
    if isinstance(response, str):
        return False, response
    return True, response
--- a/lib/agents/get_sso_list_v2/prompt.py
+++ b/lib/agents/get_sso_list_v2/prompt.py
@ -0,0 +1,68 @@
 from pydantic import BaseModel
 class FindLoginPageResponse(BaseModel):
    msg: str | None = None
    url: str | None = None
    sso_list: list[str] = []  # List of SSO providers found on the login page
 get_sso_list_task = """
 You are an expert in finding login pages.
 Your task is to navigate to the login page of the given URL. Follow the steps below strictly and return results only in the specified format.
 ※ You are NOT allowed to navigate to URLs that are not directly discoverable within the initial domain. Do NOT use search engines or guess external login URLs.
 0. INITIAL BLOCK CHECK
 - If the browser is blocked when trying to access the page — due to firewall, CAPTCHA, regional restrictions, or other access denials — immediately terminate the process and return the following JSON:
    ```json
    {
        "msg": "Blocked",
        "url": "",
        "sso_list": []
    }
    ```
 - Do NOT proceed to further steps in this case.
 1. LOGIN PAGE NAVIGATION
 - Navigate only to a **client-side (non-enterprise)** login page within the provided domain.
 - Do NOT rely on external tools, search engines, or links not directly found on the site.
 - If a consent popup (e.g. for privacy/cookies) appears, you MUST dismiss or close it before proceeding.
 - Since step 0 confirmed access, assume the page now loads properly.
 2. SSO BUTTON IDENTIFICATION
 - On the login page, look for the following social login (SSO) buttons:
  - Google, GitHub, Facebook, LinkedIn, Microsoft, Naver, Slack, Etc.
 - ✅ Proceed only if it is clearly an **actual SSO button**.
 - ❌ Exclude the following:
  - Passkey-related buttons
  - Username/password fields
  - Email-based login
  - Non-OAuth methods such as certificate or phone verification
 3. RETURN FORMAT
 - If the login page is successfully found, return:
    ```json
    {
        "msg": "Login page found",
        "url": "https://example.com/login",
        "sso_list": ["Google", "GitHub"]
    }
    ```
 - If the login page cannot be found, return:
    ```json
    {
        "msg": "Login page not found",
        "url": "",
        "sso_list": []
    }
    ```
 - If blocked (as in step 0), return:
    ```json
    {
        "msg": "Blocked",
        "url": "",
        "sso_list": []
    }
    ```
 - Return ONLY the JSON object. Do NOT include any explanation, logging, or extra output.
 """
--- a/lib/agents/login_google/init.py
+++ b/lib/agents/login_google/init.py
@ -0,0 +1,3 @@
 from lib.agents.login_google.login_google import login_google
 __all__ = ["login_google"]
--- a/lib/agents/login_google/login_google.py
+++ b/lib/agents/login_google/login_google.py
@ -0,0 +1,11 @@
 from lib.agents.login_google.prompt import login_google_task, LoginGoogleResponse
 from lib.browser_use_utils.run_task import run_task
 async def login_google(target_url) -> tuple[bool, str | LoginGoogleResponse | None]:
    task = login_google_task
    ReturnModel = LoginGoogleResponse
    success, response = await run_task(target_url, ReturnModel, task)
    if not success:
        return False, None
    return True, response
--- a/lib/agents/login_google/prompt.py
+++ b/lib/agents/login_google/prompt.py
@ -0,0 +1,63 @@
 from pydantic import BaseModel
 from lib.config import GOOGLE_ID, GOOGLE_PASSWORD
 class LoginGoogleResponse(BaseModel):
    msg: str | None = None
    status: str | None = None  # "success", "mfa_required", "google_blocked", "sso_not_found", "login_page_not_found", "invalid_credentials"
    final_url: str | None = None
 login_google_task = f"""
 You are a web automation agent.
 Your task is to visit the given domain and perform a full login via the **Google SSO button**, following all steps strictly as described below.
 ▶ Target: Find a login page inside this domain that allows "Sign in with Google", and use it to complete login via Google.
 Instructions:
 1. If any cookie or privacy popups appear, dismiss or accept them.
 2. Navigate through the site's UI to find the **login or sign-in page** (e.g., via buttons like "Log In", "Sign In", "Get Started").
    - Only follow links within the same domain.
 3. On the login page, look for a clearly labeled **Google SSO button** — typically labeled as:
    - "Continue with Google"
    - "Sign in with Google"
    - or a button with the Google 'G' icon
 4. Click the **Google login button**.
    - ⚠️ The Google login flow MUST open in a **new browser tab** (not a new window or popup).
    - ❌ If the login opens in a new **window** or **popup**, do NOT continue. Immediately stop and return the appropriate status.
 5. Check if the user is **already logged in to Google and immediately redirected back to the original site** without showing a Google login screen.
    - ✅ If so, treat the login as successful and return immediately.
 6. If redirected to the Google login page:
    - If a **CAPTCHA**, **MFA prompt**, or a request for **ID/password entry** appears, do NOT proceed.
    - Immediately stop and return the appropriate status.
 7. If login proceeds without interruptions, wait for redirection back to the original site and record the final URL.
 Credentials to use for Google login:
 - Email: {GOOGLE_ID}
 - Password: {GOOGLE_PASSWORD}
 Constraints:
 - Do NOT use search engines or guess URLs.
 - Do NOT use autofill, saved sessions, or cookies.
 - Do NOT proceed with login if:
    - The login opens in a new window (only tabs are allowed)
    - CAPTCHA or MFA appears
    - ID/password input is required
 - If the user is already logged in to Google and redirected back automatically, stop there and report success.
 - If the login page cannot be found, return "login_page_not_found".
 - If the Google login button is not found, return "sso_not_found".
 - If a page such as a sign-up page appears, consider it a successful login and terminate immediately.
 Final Output:
 Return the result in the following format only:
 ```json
 {{
    "msg": "Google login completed",
    "status": "success" | "already_logged_in" | "mfa_required" | "captcha_triggered" | "window_blocked" | "idpw_required" | "google_blocked" | "sso_not_found" | "login_page_not_found",
    "final_url": "<url_after_login_redirect or empty string>"
 }}
 ```
 - Return ONLY the JSON object. Do NOT include any explanation, logging, or extra output.
 """
--- a/lib/browser_use_utils/init.py
+++ b/lib/browser_use_utils/init.py
@ -0,0 +1,15 @@
 from lib.browser_use_utils.clean_resources import clean_resources, clean_agent_resources, clean_session_resources
 from lib.browser_use_utils.create_google_ai import create_google_ai
 from lib.browser_use_utils.get_profile import get_profile
 from lib.browser_use_utils.run_agent import run_agent
 from lib.browser_use_utils.run_task import run_task
 __all__ = [
    "clean_resources",
    "clean_agent_resources",
    "clean_session_resources",
    "create_google_ai",
    "get_profile",
    "run_agent",
    "run_task",
 ]
--- a/lib/browser_use_utils/clean_resources.py
+++ b/lib/browser_use_utils/clean_resources.py
@ -0,0 +1,21 @@
 async def clean_agent_resources(agent=None):
    """에이전트 리소스를 정리하는 함수"""
    if agent:
        try:
            await agent.close()
        except Exception as e:
            print(f"⚠️ 에이전트 리소스 정리 실패: {e}")
 async def clean_session_resources(session=None):
    """브라우저 리소스를 정리하는 함수"""
    if session:
        try:
            await session.close()
        except Exception as e:
            print(f"⚠️ 브라우저 리소스 정리 실패: {e}")
 async def clean_resources(agent=None, session=None):
    """리소스를 정리하는 함수"""
    await clean_agent_resources(agent)
    await clean_session_resources(session)
--- a/lib/browser_use_utils/create_google_ai.py
+++ b/lib/browser_use_utils/create_google_ai.py
@ -7,7 +7,7 @@ class QuotaExhaustedHandler(BaseCallbackHandler):
            print("⚠️ API 쿼터가 소진되었습니다. 재시도 로직에 위임합니다...")
            # backoff handled in scan_one_url
-def CreateChatGoogleGenerativeAI(model: str):
+def create_google_ai(model: str):
    """재시도 로직이 포함된 LLM 생성"""
    if model == "fallback":
        print("⚠️ Fallback 모델을 사용합니다. Envorinment 변수를 확인하세요.")
--- a/lib/browser_use_utils/get_profile.py
+++ b/lib/browser_use_utils/get_profile.py
@ -1,31 +1,12 @@
 import os
 from pathlib import Path
 from dotenv import load_dotenv
 from browser_use import BrowserProfile
-# Load environment variables
+async def get_storage_state():
 load_dotenv(override=True)
 def setup_proxy():
    """Configure proxy settings from environment variables."""
    proxy_host = os.getenv("PROXY_HOST")
    proxy_port = os.getenv("PROXY_PORT")
    if proxy_host and proxy_port:
        proxy_url = f"http://{proxy_host}:{proxy_port}"
        print(f"🔗 Using proxy: {proxy_host}:{proxy_port}")
        return proxy_url
    else:
        print("🔗 No proxy configured, using direct connection.")
        return None
 async def setup_storage_state():
    """Setup browser storage state for session persistence."""
    # Get the script directory to ensure correct path resolution
-    script_dir = Path(__file__).parent.parent.parent.parent
+    storage_state_path = Path("data/storage_state.json")
-    storage_state_path = script_dir / "data" / "storage_state.json"
+    storage_state_temp_path = Path("data/storage_state_temp.json")
    storage_state_temp_path = script_dir / "data" / "storage_state_temp.json"
    print(f"📂 Storage state path: {storage_state_path}")
    print(f"📂 Temp storage state path: {storage_state_temp_path}")
@ -44,6 +25,20 @@ async def setup_storage_state():
    return None
 def get_proxy_url():
    """Configure proxy settings from environment variables."""
    proxy_host = os.getenv("PROXY_HOST")
    proxy_port = os.getenv("PROXY_PORT")
    if proxy_host and proxy_port:
        proxy_url = f"http://{proxy_host}:{proxy_port}"
        print(f"🔗 Using proxy: {proxy_host}:{proxy_port}")
        return proxy_url
    else:
        print("🔗 No proxy configured, using direct connection.")
        return None
 def get_browser_args():
    """Get browser arguments for enhanced compatibility and security."""
    return [
@ -73,3 +68,30 @@ def get_browser_args():
        # Language
        f"--lang={os.getenv('LANG', 'en_US')}",
    ]
 async def get_profile():
    proxy_url = get_proxy_url()
    storage_state_path = await get_storage_state()
    profile = BrowserProfile(
        # Security settings
        disable_security=True,
        stealth=True,
        # Display settings
        headless=False,
        device_scale_factor=1,
        window_size={"width": 1600, "height": 900},
        viewport={"width": 1600, "height": 900},
        # Data persistence
        user_data_dir=None,
        storage_state=storage_state_path,
        # Network settings
        proxy={"server": proxy_url} if proxy_url else None,
        # Additional arguments
        args=get_browser_args(),
    )
    return profile
--- a/lib/browser_use_utils/run_agent.py
+++ b/lib/browser_use_utils/run_agent.py
@ -0,0 +1,40 @@
 from typing import Any
 from pydantic import BaseModel
 from lib.browser_use_utils.clean_resources import clean_agent_resources
 from lib.config import GOOGLE_MODEL
 from browser_use import (
    Agent,
    Controller,
 )
 from lib.browser_use_utils.create_google_ai import create_google_ai
 async def run_agent(session, initial_actions, ReturnModel: type[BaseModel], task: str) -> tuple[bool, str, Any | None]:
    controller = Controller(output_model=ReturnModel, exclude_actions=['search_google'])
    agent = Agent(
        browser_session=session,
        initial_actions=initial_actions,
        task=task,
        llm=create_google_ai(GOOGLE_MODEL),
        controller=controller,
    )
    try:
        response = await agent.run()
        final_result = response.final_result()
        if final_result is None:
            return False, "LLM이 반환한 최종 결과가 없습니다.", None
    except Exception as e:
        # API 쿼터 문제인지 확인
        if "ResourceExhausted" in str(e) or "429" in str(e):
            return False, "API 쿼터 에러로 인한 실패", None
        # 일반 에러 처리
        else:
            return False, "일반 에러로 인한 실패", None
    finally:
        await clean_agent_resources(agent)
    return True, "ok", final_result
--- a/lib/browser_use_utils/run_task.py
+++ b/lib/browser_use_utils/run_task.py
@ -0,0 +1,40 @@
 import json
 from typing import Any
 from pydantic import BaseModel
 from browser_use import (
    BrowserSession
 )
 from patchright.async_api import async_playwright as async_patchright
 from lib.utils.logger import logger
 from lib.browser_use_utils import get_profile, clean_session_resources, run_agent
 async def run_task(target_url: str, ReturnModel: type[BaseModel], task: str) -> tuple[bool, type[BaseModel] | None]:
    session = BrowserSession(
        playwright=(await async_patchright().start()),
        browser_profile=await get_profile(),
    )
    initial_actions = [{"open_tab": {"url": target_url}}]
    seccess, msg, final_result = await run_agent(session=session,
                                           initial_actions=initial_actions,
                                           ReturnModel=ReturnModel,
                                           task=task)
    if not seccess:
        logger(f"⚠️ LLM 실행 실패: {target_url} | {msg}")
        print(f"⚠️ LLM 실행 실패: {target_url} | {msg}")
        await clean_session_resources(session)
        return False, None
    try:
        data = json.loads(final_result)
        resp = ReturnModel(**data)
        return True, resp
    except Exception as e:
        logger(f"⚠️ LLM 응답 결과 파싱 실패: {target_url} | {e}\n원본 결과: {data.msg}")
        print(f"⚠️ LLM 응답 결과 파싱 실패: {target_url} | {e}\n원본 결과: {data.msg}")
        return False, None
    finally:
        await clean_session_resources(session)
--- a/lib/utils/config.py
+++ b/lib/utils/config.py
@ -4,5 +4,7 @@ load_dotenv(verbose=True, override=True)
 BACKEND_URL = os.getenv("BACKEND_URL", "http://localhost:11081")
 GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
-GOOGLE_MODEL = os.getenv("GOOGLE_MODEL", "gemini-2.5-flash-preview-05-20")
+GOOGLE_MODEL = os.getenv("GOOGLE_MODEL", "gemini-2.5-flash")
-GOOGLE_PLANNER_MODEL = os.getenv("GOOGLE_PLANNER_MODEL", "gemini-2.5-pro-preview-06-05")
+
 GOOGLE_ID = os.getenv("GOOGLE_ID", "google")
 GOOGLE_PASSWORD = os.getenv("GOOGLE_PASSWORD", "google")
--- a/lib/llm/prompt/init.py
+++ b/lib/llm/prompt/init.py
@ -1,142 +0,0 @@
 import os
 from dotenv import load_dotenv
 load_dotenv(override=True)
 # Extended planner prompt
 extend_planner_system_message = f"""
 🎯 목적: 웹 자동화를 위한 **SSO 로그인 리디렉션 URL 수집**
 📌 주의사항 (전제 조건)
 - ❌ **검색 엔진(Google, Bing 등) 사용 금지**
 - ✅ **초기 제공된 URL 내에서만 탐색**
 - ❌ 직접 이동하거나 추측한 링크 클릭 금지
 - ⛔ 추측한 URL은 대답하거나 클릭하지 마세요
 - OAuth가 아닌 일반 로그인은 무시
 - OAuth가 없다면 **즉시 중단**하고 빈 배열 반환
 ---
 ## 🧩 Step 0: 페이지 차단(Block) 여부 확인
 초기 URL의 로그인 페이지에 접근하여 다음 사항을 점검합니다:
 - 🚫 페이지 차단됨 (Firewall, Access Denied 등) → 즉시 중단
 - 🔒 CAPTCHA는 통과 가능 (해결하고 계속 진행)
 - ❗ 로그인 UI가 정상적으로 로드되지 않으면 중단
 📤 차단 시 즉시 반환:
 ```json
 [
  {{
    "provider": "Blocked",
    "oauth_uri": "-"
  }}
 ]
 ````
 ---
 ## 🔍 Step 1: 로그인 페이지 탐색
 * 초기 URL에 접속하여 **클라이언트용 로그인 페이지**로 진입합니다.
 * 쿠키 동의, 개인정보 안내 등 팝업은 무시하거나 닫고 계속 진행하세요.
 * 페이지가 정상 로드되었다고 가정합니다.
 ---
 ## 👀 Step 2: SSO 로그인 버튼 식별
 아래 **OAuth SSO 버튼들만** 유효합니다:
 * ✅ Google, GitHub, Facebook, LinkedIn, Microsoft, Naver
 **유효한 버튼 기준**:
 * OAuth 인증 흐름을 실제로 트리거
 * `window.location` 또는 `<a href=...>` 또는 JS로 redirect가 발생
 **제외 버튼들 (클릭 금지)**:
 * ❌ 일반 로그인, 패스키, 이메일/전화번호, 인증서 기반, 비밀번호 입력
 ---
 ## ✅ Step 3: 모든 SSO 버튼 클릭 및 로그인 시도
 > 각 SSO 로그인 버튼을 클릭한 뒤 반드시 아래 절차를 **완전히 수행**해야 합니다.
 각 SSO 버튼에 대해 다음을 수행:
 1. 버튼 클릭
 2. 🌐 페이지가 이동되면, **현재 주소창(URL)을 확인하여 리디렉션된 OAuth URL**을 `oauth_uri`로 저장  
   → 예: `https://accounts.google.com/o/oauth2/auth?...`
 3. ✅ 로그인 진행:
   - 로그인 페이지에서 OAuth 인증을 완료합니다.
   - sign in with your username(email) x_username and password is x_password
   - 버튼같은게 안눌리면 새로고침을 해봐
   - **로그인 완료 후 authorize 등 버튼이 있으면 클릭**
   - GitHub같은 경우 Authorize 버튼이 뜨는데 오래걸릴 수 있음, 기다려야 할 수도 있음
   - 만약 버튼을 눌러도 반응이 없을 경우 새로고침을 한번 해주세요.
   - 로그인 실패 시에는 다음 SSO 버튼을 클릭합니다.
 4. 로그인이 성공하면 모두 쿠키를 삭제하고 다음 SSO 버튼을 클릭합니다.
 5. 다음 SSO 버튼으로 반복 진행
 쿠키 삭제 방법:
 chrome://settings/clearBrowserData에 들어가서 삭제해주세요.
 🛑 절대 아래와 같이 해석하지 말 것:
 - ❌ 버튼 클릭 후 페이지 로딩만 기다리고 돌아가기
 - ❌ URL 저장 없이 go_back() 호출
 📤 각 로그인 후 다음 형식으로 결과 저장:
 ```json
 [
  {{
    "provider": "Google",
    "oauth_uri": "https://example.com/auth/google?client_id=..."
  }}
 ]
 ````
 ````
 ---
 ### ✨ 추가 안전 장치: "뒤로가기(go_back) 호출 조건" 제한
 ```text
 🛑 뒤로가기(go_back)은 다음 조건이 모두 충족될 때만 사용 => 다만 로그인 실패 시, 뒤로가기 수행:
 - ✅ 로그인 흐름이 완료됨 (예: redirect back to app, or callback URL)
 - ✅ 현재 리디렉션 URL이 수집됨
 - ✅ 결과에 저장 후 다음 버튼 탐색을 위해 복귀 필요할 때
 ```
 ---
 ## 🚫 Step 4: 버튼 없음 또는 예외 발생 시
 * 유효한 SSO 버튼이 **전혀 없을 경우**
 * 예외, 오류 등 발생 시
 📤 즉시 중단 후 다음 형식으로 반환:
 ```json
 []
 ```
 ---
 ## 📎 중요 규칙 요약
 * ✅ **모든 SSO 로그인은 반드시 실행** (가능한 버튼은 모두 클릭)
 * 🔁 단계는 반드시 순서대로 진행
 * 🔐 로그인은 쿠키/세션으로 유지된 상태에서 수행
 * 🚫 직접 ID/PW 입력하지 않음
 * ⛔ 추측 URL 클릭 금지
 * ❗ 예외 발생 시 반드시 규정된 JSON 포맷만 반환
 ---
 """
--- a/lib/utils/init.py
+++ b/lib/utils/init.py
@ -1,40 +1,20 @@
-from lib.utils.config import (
+from lib.utils.env_checker import check_env_variables
-    BACKEND_URL,
+from lib.utils.is_html import is_html_url
-    GOOGLE_API_KEY,
+from lib.utils.logger import logger
-    GOOGLE_MODEL,
+from lib.utils.notify_backend import notify_backend
-    GOOGLE_PLANNER_MODEL,
+from lib.utils.progress_checker import save_progress, load_progress
-)
+# v2 import => 아직 개발 중
 from lib.utils.progress_checker_v2 import ProgressChecker
 from lib.utils.read_txt import read_lines_between
 from lib.utils.save_oauth_providers import save_oauth_providers
-
+__all__ = [
-def show_info():
+    "check_env_variables",
-    print("🔧 환경 설정:")
+    "is_html_url",
-    print(browser_use_version())
+    "logger",
-    print(f"🔗 Backend URL: {BACKEND_URL}")
+    "notify_backend",
-    print(
+    "read_lines_between",
-        f"🔑 Google API Key: {'*' * (len(GOOGLE_API_KEY) - 4) + GOOGLE_API_KEY[-4:] if GOOGLE_API_KEY else None}"
+    "save_progress",
-    )
+    "load_progress",
-    print(f"🌐 Google Model: {GOOGLE_MODEL}")
+    "save_oauth_providers",
-    print(f"🌐 Google Planner Model: {GOOGLE_PLANNER_MODEL}")
+]
 def browser_use_version():
    try:
        # run uv pip show browser-use
        import subprocess
        result = subprocess.run(
            ["uv", "pip", "show", "browser-use"],
            capture_output=True,
            text=True,
            check=True,
        )
        print("📦 Browser Use 패키지 정보:")
        return result.stdout.strip()
    except ImportError:
        return None
 def env_cheker():
    if GOOGLE_API_KEY is None:
        raise ValueError("GOOGLE_API_KEY 환경변수가 설정되지 않았습니다.")
--- a/lib/utils/browser_use/init.py
+++ b/lib/utils/browser_use/init.py
@ -1,31 +0,0 @@
 from lib.utils.browser_use.func import *
 # Initialize configuration
 proxy_url = setup_proxy()
 # Create browser profile
 async def GetProfile():
    storage_state_path = await setup_storage_state()
    profile = BrowserProfile(
        # Security settings
        disable_security=True,
        stealth=True,
        # Display settings
        headless=False,
        device_scale_factor=1,
        window_size={"width": 1600, "height": 900},
        viewport={"width": 1600, "height": 900},
        # Data persistence
        user_data_dir=None,
        storage_state=storage_state_path,
        # Network settings
        proxy={"server": proxy_url} if proxy_url else None,
        # Additional arguments
        args=get_browser_args(),
    )
    return profile
--- a/lib/utils/browser_use/clean_resources.py
+++ b/lib/utils/browser_use/clean_resources.py
@ -1,25 +0,0 @@
 from pathlib import Path
 async def clean_resources(agent=None, session=None):
    """리소스를 정리하는 함수"""
    storage_state_temp_path = Path("./data/storage_state_temp.json").resolve()
    if storage_state_temp_path.exists():
        try:
            # remove file
            print(f"🗑️ 임시 스토리지 상태 파일 삭제 중: {storage_state_temp_path}")
            # unlink removes the file
            storage_state_temp_path.unlink()
            print("🗑️ 임시 스토리지 상태 파일 삭제 완료.")
        except Exception as e:
            print(f"⚠️ 임시 스토리지 상태 파일 삭제 실패: {e}")
    if agent:
        try:
            await agent.close()
        except Exception as e:
            print(f"⚠️ 에이전트 리소스 정리 실패: {e}")
    if session:
        try:
            await session.close()
        except Exception as e:
            print(f"⚠️ 세션 리소스 정리 실패: {e}")
--- a/lib/utils/browser_use/model.py
+++ b/lib/utils/browser_use/model.py
@ -1,11 +0,0 @@
 from typing import List
 from pydantic import BaseModel
 # 출력 모델
 class OAuth(BaseModel):
    provider: str
    oauth_uri: str
 class OAuthList(BaseModel):
    oauth_providers: List[OAuth]
--- a/lib/utils/browser_use/sensitive_data.py
+++ b/lib/utils/browser_use/sensitive_data.py
@ -1,21 +0,0 @@
 # read json file .sensitive.json
 import json
 import os
 def GetSensitiveData():
    """
    Reads sensitive data from a .sensitive.json file in the current directory.
    Returns:
        dict: A dictionary containing the sensitive data.
    """
    file_path = os.path.join(os.getcwd(), '.sensitive.json')
    if not os.path.exists(file_path):
        return None
    with open(file_path, 'r') as file:
        sensitive_data = json.load(file)
    return sensitive_data
--- a/lib/utils/env_checker.py
+++ b/lib/utils/env_checker.py
@ -0,0 +1,15 @@
 import os
 from dotenv import load_dotenv
 load_dotenv()
 def check_env_variables():
    """환경변수 체크 함수"""
    required_vars = [
        "BACKEND_URL",
        "GOOGLE_API_KEY",
        "GOOGLE_MODEL",
    ]
    for var in required_vars:
        if os.getenv(var) is None:
            raise ValueError(f"{var} 환경변수가 설정되지 않았습니다.")
--- a/lib/utils/notify_backend.py
+++ b/lib/utils/notify_backend.py
@ -1,6 +1,6 @@
 import requests
-from lib.utils.config import BACKEND_URL
+from lib.config import BACKEND_URL
 def notify_backend(target_url):
    # Backend에 스캔 시작을 알림
--- a/lib/utils/progress_checker.py
+++ b/lib/utils/progress_checker.py
@ -0,0 +1,21 @@
 import json
 import os
 from pathlib import Path
 progress_file = Path("data/scan_progress.json")
 def save_progress(current_progress):
    """현재 진행 상황을 파일에 저장"""
    with open(progress_file, 'w', encoding='utf-8') as f:
        json.dump(current_progress, f, ensure_ascii=False, indent=2)
 def load_progress():
    """이전 진행 상황을 파일에서 불러오기"""
    if os.path.exists(progress_file):
        try:
            with open(progress_file, 'r', encoding='utf-8') as f:
                return json.load(f)
        except:
            return None
    return None
--- a/lib/utils/progress_checker_v2.py
+++ b/lib/utils/progress_checker_v2.py
@ -0,0 +1,25 @@
 import json
 import os
 from pathlib import Path
 progress_file = Path("data/scan_progress.json")
 class ProgressChecker:
    def __init__(self, filepath):
        self.filepath = filepath
        self.progress = self.load_progress()
    def save(self):
        """현재 진행 상황을 파일에 저장"""
        with open(self.filepath, 'w', encoding='utf-8') as f:
            json.dump(self.progress, f, ensure_ascii=False, indent=2)
    def load(self):
        """이전 진행 상황을 파일에서 불러오기"""
        if os.path.exists(self.filepath):
            try:
                with open(self.filepath, 'r', encoding='utf-8') as f:
                    return json.load(f)
            except:
                return None
        return None
--- a/lib/utils/save_oauth_providers.py
+++ b/lib/utils/save_oauth_providers.py
@ -0,0 +1,13 @@
 import csv
 import os
 def save_oauth_providers(url, oauth_entries):
    csv_file = "./oauth_providers.csv"
    file_exists = os.path.isfile(csv_file)
    with open(csv_file, "a", newline="", encoding="utf-8") as f:
        writer = csv.writer(f)
        if not file_exists:
            writer.writerow(["issuer", "provider", "oauth_uri"])
        for entry in oauth_entries:
            writer.writerow([url, entry.provider or None, entry.oauth_uri or None])
    print(f"✅ OAuth providers saved to {csv_file}\n")
--- a/main.py
+++ b/main.py
@ -1,69 +1,59 @@
 import asyncio
 import json
 import os
 import csv
 import argparse
 from pathlib import Path
 import signal
 from dotenv import load_dotenv
 from lib.config import BACKEND_URL
 from lib.utils import notify_backend, is_html_url, read_lines_between, save_progress, load_progress, check_env_variables 
 from lib.agents import get_sso_list, login_google
 load_dotenv()
-from browser_use import (
+check_env_variables()
-    Agent,
+
-    BrowserSession,
+backend_url = BACKEND_URL
-    Controller,
+
-    ActionResult,
+login_agents = {
-)
+    "google": login_google
-from patchright.async_api import async_playwright as async_patchright, Page
+}
-from pydantic import BaseModel
+
 # ── URL별로 Browser를 새로 띄우는 함수 ──
 async def scan_one_url(url: str, skip_html_check: bool = False):
    target_url = url if url.startswith("http") else f"https://{url}"
    print(f"🚀 Starting scan for: {target_url}")
    # 1) URL이 HTML 페이지인지 확인
    if not is_html_url(target_url) and not skip_html_check:
        print(f"❌ {target_url} 은(는) HTML이 아닙니다. 스킵합니다.")
        return
    # Backend에 스캔 시작 알림
    notify_backend(target_url)
    success, response = await get_sso_list(target_url)
    if not success:
        return
    if len(response.sso_list) == 0:
        return
    for sso in response.sso_list:
        target_login_agent = login_agents.get(sso.lower())
        if target_login_agent:
            print(f"🔍 {target_url} 에서 SSO 발견: {sso}, 로그인 시도 중...")
            success, login_response = await target_login_agent(target_url)
            if not success:
                print(f"⚠️ {target_url} 에서 {sso} 로그인 실패")
                continue
            print(f"✅ {target_url} 에서 {sso} 로그인 성공: {login_response.final_url}")
        else:
            print(f"✅ {target_url} 에서 SSO 발견: {sso} | TODO")
    # Backend에 스캔 완료 알림
    # 오탐 검증
 from lib.utils import env_cheker
 from lib.utils.backend_client import notify_backend
 from lib.utils.browser_use import model
 from lib.utils.browser_use.clean_resources import clean_resources
 from lib.utils.browser_use.func import setup_storage_state
 from lib.utils.browser_use.sensitive_data import GetSensitiveData
 from lib.utils.config import BACKEND_URL, GOOGLE_MODEL, GOOGLE_PLANNER_MODEL
 from lib.utils.is_html import is_html_url
 from lib.utils.read_txt import read_lines_between
 from lib.llm.prompt import extend_planner_system_message
 from lib.utils.logger import logger
 import lib.utils.browser_use as browser_use
 from lib.llm import CreateChatGoogleGenerativeAI
 load_dotenv(verbose=True, override=True)
 # Exponential backoff settings
 INITIAL_BACKOFF = int(os.getenv("INITIAL_BACKOFF", "60"))  # seconds
 MAX_BACKOFF = int(os.getenv("MAX_BACKOFF", "600"))  # seconds
 # 진행 상황 추적을 위한 전역 변수
 current_progress = {"current_index": 0, "total": 0, "current_url": "", "start_line": 0}
 progress_file = Path("data/scan_progress.json")
 env_cheker()
 if os.getenv("LMNR_PROJECT_API_KEY"):
    from lmnr import Laminar
    Laminar.initialize(project_api_key=os.getenv("LMNR_PROJECT_API_KEY"))
 def save_progress():
    """현재 진행 상황을 파일에 저장"""
    with open(progress_file, 'w', encoding='utf-8') as f:
        json.dump(current_progress, f, ensure_ascii=False, indent=2)
 def load_progress():
    """이전 진행 상황을 파일에서 불러오기"""
    if os.path.exists(progress_file):
        try:
            with open(progress_file, 'r', encoding='utf-8') as f:
                return json.load(f)
        except:
            return None
    return None
 def signal_handler(signum, frame):
    """Ctrl+C 시그널 핸들러"""
@ -73,127 +63,14 @@ def signal_handler(signum, frame):
    print(f"   - 전체: {current_progress['total']}개 URL")
    print(f"   - 완료: {current_progress['current_index']}개 URL")
    print(f"   - 현재 처리 중: {current_progress['current_url']}")
-    print(f"   - domains.txt의 {current_progress['start_line'] + current_progress['current_index']}번째 줄")
+    print(f"   - domains.txt의 {current_progress['current_index']}번째 줄")
    print(f"   - 진행률: {current_progress['current_index']}/{current_progress['total']} ({current_progress['current_index']/current_progress['total']*100:.1f}%)")
    print("="*60)
-    save_progress()
+    save_progress(current_progress)
    print(f"💾 진행 상황이 {progress_file}에 저장되었습니다.")
    exit(0)
 # 시그널 핸들러 등록
 signal.signal(signal.SIGINT, signal_handler)
 # ── URL별로 Browser를 새로 띄우는 함수 ──
 async def scan_one_url(url: str, skip_html_check: bool = False):
    await setup_storage_state()
    target_url = url if url.startswith("http") else f"https://{url}"
    print(f"🚀 Starting scan for: {target_url}")
    # 1) URL이 HTML 페이지인지 확인
    if not is_html_url(target_url) and not skip_html_check:
        print(f"❌ {target_url} 은(는) HTML이 아닙니다. 스킵합니다.")
        return
    # Backend에 스캔 시작을 알림
    notify_backend(target_url)
    agent = None
    session = None
    try_cnt = 0
    while True:
        # BrowserSession에 profile 전달
        session = BrowserSession(
            playwright=(await async_patchright().start()),
            browser_profile=await browser_use.GetProfile(),
        )
        # Agent 생성 및 실행 (단일 try-except with 백오프)
        initial_actions = [{"open_tab": {"url": target_url}}]
        controller = Controller(output_model=model.BaseModel, exclude_actions=['search_google'])
        print("🤖 LLM 모델 초기화 및 스캔 시작...")
        print("Available actions:", list(controller.registry.registry.actions.keys()))
        try:
            agent = Agent(
                browser_session=session,
                initial_actions=initial_actions,
                sensitive_data=GetSensitiveData(),
                task=(
                    "Navigate to the login page, identify all OAuth provider buttons (excluding Passkey), "
                    "and for each one: click the button, follow the full OAuth login flow as far as possible "
                    "with a real user account (without using a fake or non-existent account), and capture the "
                    "final redirect URL after login. Do not stop at just collecting the initial authorization URL—"
                    "actually perform the login step like a real user would. "
                    "If the OAuth buttons do not appear immediately, wait briefly to allow the page to load completely before proceeding. "
                    "Always log out before starting the login process, and make sure to attempt the login again from a clean state."
                ),
                llm=CreateChatGoogleGenerativeAI(GOOGLE_MODEL),
                planner_llm=CreateChatGoogleGenerativeAI(GOOGLE_PLANNER_MODEL),
                controller=controller,
                extend_planner_system_message=extend_planner_system_message,
            )
            response = await agent.run()
            final_result = response.final_result()
            if final_result is None:
                raise ValueError("final_result()가 None을 반환했습니다.")
        except Exception as e:
            await clean_resources(agent, session)
            # API 쿼터 문제인지 확인
            if "ResourceExhausted" in str(e) or "429" in str(e):
                wait = min(INITIAL_BACKOFF * (2**try_cnt), MAX_BACKOFF)
                print(f"⚠️ API 쿼터 에러: {e}. {wait}초 대기 후 재시도합니다...")
                await asyncio.sleep(wait)
                try_cnt += 1
                if try_cnt >= 3:
                    print(f"❌ {url} 스캔 실패: API 쿼터 문제가 지속됩니다.")
                    logger(f"❌ {url} 스캔 실패: API 쿼터 문제: {e}")
                    return
                continue
            # 일반 에러 처리
            try_cnt += 1
            if try_cnt >= 3:
                print(f"❌ {url} 스캔 실패: 에러: {e}")
                logger(f"❌ {url} 스캔 실패: 에러: {e}")
                return
            print(f"⚠️ 에러 발생: {e}. {try_cnt}번째 재시도 중...")
            await asyncio.sleep(30)
            continue
        # 스캔 결과 처리
        data = json.loads(final_result)
        try:
            oauth_entries = [model.OAuth(**entry) for entry in data["oauth_providers"]]
        except Exception as e:
            raise ValueError(f"결과 파싱 실패: {e}\n원본 결과: {final_result}")
        print("-" * 50)
        print(f"🔗 Scanned URL: {url}\n")
        print("🔐 Detected OAuth Providers and URLs:")
        for entry in oauth_entries:
            if "<" in entry.oauth_uri or "..." in entry.oauth_uri:
                print(
                    f"⚠️ WARNING: {entry.provider} URL may be masked or incomplete:\n{entry.oauth_uri}\n"
                )
            else:
                print(f"- {entry.provider}: {entry.oauth_uri}")
        print("-" * 50)
        # CSV에 저장 (append)
        csv_file = "./data/oauth_providers.csv"
        file_exists = os.path.isfile(csv_file)
        with open(csv_file, "a", newline="", encoding="utf-8") as f:
            writer = csv.writer(f)
            if not file_exists:
                writer.writerow(["issuer", "provider", "oauth_uri"])
            for entry in oauth_entries:
                writer.writerow([url, entry.provider, entry.oauth_uri])
        await clean_resources(agent, session)
        break
 async def loop(
    filepath: str, start_line: int, end_line: int, skip_html_check: bool = False
 ):
@ -214,8 +91,8 @@ async def loop(
        print(f"   - 이전 완료: {prev_progress['current_index']}/{prev_progress['total']}")
        print(f"   - 마지막 처리: {prev_progress.get('current_url', 'N/A')}")
-        resume = input("이어서 진행하시겠습니까? (y/n): ").lower().strip()
+        resume = input("이어서 진행하시겠습니까? (Y/n): ").lower().strip()
-        if resume == 'y':
+        if resume != 'n':
            current_progress["current_index"] = prev_progress["current_index"]
            target_list = target_list[current_progress["current_index"]:]
            print(f"✅ {current_progress['current_index']}번째부터 재개합니다.")
@ -229,64 +106,61 @@ async def loop(
        current_progress["current_index"] = actual_index
        print(f"\n🔄 Processing {actual_index + 1}/{current_progress['total']}: {url}")
-        print(f"📍 domains.txt의 {start_line + actual_index}번째 줄")
+        print(f"📍 domains.txt의 {actual_index}번째 줄")
        # URL들 사이에 API 쿼터 회복을 위한 대기 시간 추가
        if actual_index > 0:
            print("⏳ API 쿼터 보호를 위해 30초 대기 중...")
            await asyncio.sleep(30)
        await scan_one_url(url, skip_html_check=skip_html_check)
        # 진행 상황 저장
        current_progress["current_index"] = actual_index + 1
-        save_progress()
+        save_progress(current_progress)
        print("⏳ API 쿼터 보호를 위해 10초 대기 중...")
        await asyncio.sleep(10)
    print(f"\n🎉 모든 스캔이 완료되었습니다! ({current_progress['total']}개 URL)")
    # 완료 후 진행 상황 파일 삭제
    if os.path.exists(progress_file):
        os.remove(progress_file)
 def main():
    parser = argparse.ArgumentParser(
        prog="domain_scanner",
-        description="도메인 목록 파일에서 지정한 줄 범위를 읽어 SSO 스캔을 수행합니다.",
+        description="도메인 목록 파일에서 지정한 줄 범위를 읽어 SSO 스캔을 수행합니다."
    )
    # 커맨드라인 인자로 받을 옵션들 정의
    parser.add_argument(
-        "-f",
+        "-f", "--file",
        "--file",
        type=str,
        required=True,
-        help="도메인 목록이 들어 있는 텍스트 파일 경로 (예: ./domains.txt)",
+        help="도메인 목록이 들어 있는 텍스트 파일 경로 (예: ./domains.txt)"
    )
    parser.add_argument(
-        "-s", "--start", type=int, required=True, help="읽기 시작 줄 번호 (1-based)"
+        "-s", "--start",
        type=int,
        required=True,
        help="읽기 시작 줄 번호 (1-based)"
    )
    parser.add_argument(
-        "-e", "--end", type=int, required=True, help="읽기 종료 줄 번호 (1-based)"
+        "-e", "--end",
        type=int,
        required=True,
        help="읽기 종료 줄 번호 (1-based)"
    )
    parser.add_argument(
-        "-skh",
+        "-skh", "--skip-html-check",
        "--skip-html-check",
        type=bool,
        default=False,
-        help="HTML 페이지 체크를 건너뛰고 모든 URL을 스캔합니다. (기본값: False)",
+        help="HTML 페이지 체크를 건너뛰고 모든 URL을 스캔합니다. (기본값: False)"
    )
    args = parser.parse_args()
    # 인자값을 비동기 함수에 전달
-    asyncio.run(
+    asyncio.run(loop(
-        loop(
+        filepath=args.file,
-            filepath=args.file,
+        start_line=args.start,
-            start_line=args.start,
+        end_line=args.end,
-            end_line=args.end,
+        skip_html_check=args.skip_html_check
-            skip_html_check=args.skip_html_check,
+    ))
        )
    )
 if __name__ == "__main__":
--- a/pyproject.toml
+++ b/pyproject.toml
@ -5,7 +5,6 @@ description = "Add your description here"
 readme = "README.md"
 requires-python = ">=3.13"
 dependencies = [
-    "browser-use[memory]>=0.2.7",
+    "browser-use[memory]==0.2.7",
-    "lmnr[all]>=0.6.10",
+    "patchright==1.52.5",
    "patchright>=1.52.5",
 ]
--- a/run.ps1
+++ b/run.ps1
@ -3,7 +3,10 @@
 $PYTHON_SCRIPT = "main.py"
 # 도메인 목록 파일 경로 (Python 스크립트 실행 시 -f 옵션에 전달)
-$DOMAIN_FILE = "./data/domains.txt"
+$DOMAIN_FILE = "./domains.txt"
 # 몇 줄씩(chunk) 나눠서 실행할지
 $CHUNK_SIZE = 10
 # ─────────────
 # https://f.imnya.ng/.whs/tp-domains/data/domains/latest.txt
@ -23,14 +26,27 @@ $START_LINE = [int]$args[0]
 $END_LINE = [int]$args[1]
 $SKIP_HEADER = if ($args.Count -eq 3) { $args[2] } else { "False" }
-$timestamp = Get-Date -Format "yyyy-MM-dd HH:mm:ss"
+# START_LINE부터 END_LINE까지 CHUNK_SIZE 만큼씩 반복
-Write-Host "[$timestamp] Processing lines $START_LINE to $END_LINE..."
+$current = $START_LINE
 while ($current -le $END_LINE) {
    # 각 청크 구간의 마지막 줄 계산
    $chunk_end = $current + $CHUNK_SIZE - 1
    if ($chunk_end -gt $END_LINE) {
        $chunk_end = $END_LINE
    }
-# Python 스크립트 실행
+    $timestamp = Get-Date -Format "yyyy-MM-dd HH:mm:ss"
-# -f DOMAIN_FILE: 도메인 목록 파일 경로
+    Write-Host "[$timestamp] Processing lines $current to $chunk_end..."
 # -s START_LINE : 읽기 시작 줄
 # -e END_LINE   : 읽기 끝 줄
 # -skh SKIP_HEADER: 헤더 스킵 여부
 uv run $PYTHON_SCRIPT -f $DOMAIN_FILE -s $START_LINE -e $END_LINE -skh $SKIP_HEADER
-Write-Host "처리 완료."
+    # Python 스크립트 실행
    # -f DOMAIN_FILE: 도메인 목록 파일 경로
    # -s current  : 읽기 시작 줄
    # -e chunk_end: 읽기 끝 줄
    # -skh SKIP_HEADER: 헤더 스킵 여부
    uv run $PYTHON_SCRIPT -f $DOMAIN_FILE -s $current -e $chunk_end -skh $SKIP_HEADER
    # 다음 청크의 시작 값 설정
    $current = $chunk_end + 1
 }
 Write-Host "모든 청크 처리 완료."
--- a/run.sh
+++ b/run.sh
@ -2,10 +2,11 @@
 # ── 설정 부분 ──
 PYTHON_SCRIPT="main.py"
-DOMAIN_FILE="./data/domains.txt"
+DOMAIN_FILE="./domains.txt"
 CHUNK_SIZE=10
 # ─────────────
-curl "https://f.imnya.ng/.whs/tp-domains/data/domains/latest.txt" -o $DOMAIN_FILE
+# curl "https://f.imnya.ng/.whs/tp-domains/data/domains/latest.txt" -o $DOMAIN_FILE
 # 인자 개수 확인
 if [ $# -lt 2 ]; then
@ -22,7 +23,18 @@ if [ -z "$SKH_OPTION" ]; then
  SKH_OPTION="False"
 fi
-echo "[$(date '+%Y-%m-%d %H:%M:%S')] Processing lines ${START_LINE} to ${END_LINE}..."
+current=$START_LINE
-uv run "$PYTHON_SCRIPT" -f "$DOMAIN_FILE" -s "$START_LINE" -e "$END_LINE" -skh $SKH_OPTION
+while [ "$current" -le "$END_LINE" ]; do
  chunk_end=$(( current + CHUNK_SIZE - 1 ))
  if [ "$chunk_end" -gt "$END_LINE" ]; then
    chunk_end=$END_LINE
  fi
-echo "처리 완료."
+  echo "[$(date '+%Y-%m-%d %H:%M:%S')] Processing lines ${current} to ${chunk_end}..."
  uv run "$PYTHON_SCRIPT" -f "$DOMAIN_FILE" -s "$current" -e "$chunk_end" -skh $SKH_OPTION
  current=$(( chunk_end + 1 ))
  sleep 1  # 1초 대기
 done
 echo "모든 청크 처리 완료."
--- a/uv.lock
+++ b/uv.lock
Author	SHA1	Message	Date
James	8c22a52774	Merge pull request #21 from j93es/feat/j93es Feat/j93es	2025-06-27 10:26:42 +09:00
tv0924@icloud.com	e93e64756c	[Refactor] 리팩터링	2025-06-27 10:18:24 +09:00
tv0924@icloud.com	65150b3514	[Prompt] 로그인 후, 가입페이지가 나왔을때, 무한 반복되는 오류 해결	2025-06-27 09:50:12 +09:00
tv0924@icloud.com	bbd2d6d636	[Refactor] 리팩터링	2025-06-27 09:45:50 +09:00
tv0924@icloud.com	26b40e0e65	[Refactor] 파일 구조 변경	2025-06-26 23:35:58 +09:00
tv0924@icloud.com	b7e6afb227	[Update] agent 호출 구조 변경	2025-06-26 23:33:59 +09:00
tv0924@icloud.com	495b3a52da	[Update]	2025-06-22 22:19:30 +09:00
tv0924@icloud.com	92967ed353	[Update] new logic	2025-06-22 20:35:56 +09:00
		`@ -0,0 +1,3 @@`
							`from lib.agents.get_sso_list.get_sso_list import get_sso_list`

							`__all__ = ["get_sso_list"]`
		`@ -0,0 +1,3 @@`
							`from lib.agents.get_sso_list_v2 import get_sso_list`

							`__all__ = ["get_sso_list"]`
		`@ -0,0 +1,3 @@`
							`from lib.agents.login_google.login_google import login_google`

							`__all__ = ["login_google"]`