🌿 Fern Regeneration -- March 6, 2025 (#31)

fern-api[bot] · justinsunyt · web-flow · commit 781d24a1c63f · 2025-03-05T21:51:24.000-05:00
* SDK regeneration

* feat: update prompts + move to model folder

* feat: update prompts in test

* feat: update deprecated docstrings

---------

Co-authored-by: fern-api &lt;115122769+fern-api[bot]@users.noreply.github.com&gt;
Co-authored-by: Justin Sun &lt;justinsunyt@gmail.com&gt;
diff --git a/pyproject.toml b/pyproject.toml
@@ -1,6 +1,9 @@
+[project]
+name = "scrapybara"
+
 [tool.poetry]
 name = "scrapybara"
-version = "2.3.0"
+version = "2.3.1"
 description = ""
 readme = "README.md"
 authors = []
diff --git a/src/scrapybara/anthropic/__init__.py b/src/scrapybara/anthropic/__init__.py
@@ -3,6 +3,7 @@
 from pydantic import Field
 
 from ..types.act import Model
+from datetime import datetime
 
 
 class Anthropic(Model):
@@ -29,3 +30,93 @@ def __init__(
         api_key: Optional[str] = None,
     ) -> None:
         super().__init__(provider="anthropic", name=name, api_key=api_key)
+
+
+UBUNTU_SYSTEM_PROMPT = f"""<SYSTEM_CAPABILITY>
+* You have access to an Ubuntu VM with internet connectivity
+* You can install Ubuntu applications using the bash tool (use curl over wget)
+* To run GUI applications with the bash tool, use a subshell, e.g. "(DISPLAY=:1 xterm &)", make sure to include the parantheses
+* GUI apps will appear but may take time to load - confirm with an extra screenshot
+* Chromium is the default browser
+* Start Chromium via the bash tool "(DISPLAY=:1 chromium &)", but interact with it visually via the computer tool
+* If you need to read a HTML file:
+  - Open with the address bar in Chromium
+* For commands with large text output:
+  - Redirect to a temp file
+  - Use str_replace_editor or grep with context (-B and -A flags) to view output
+* When viewing pages:
+  - Zoom out to see full content, or
+  - Scroll to ensure you see everything
+* When interacting with a field, always clear the field first using "ctrl+A" and "delete"
+  - Take an extra screenshot after clicking "enter" to confirm the field is properly submitted and move the mouse to the next field
+* Computer function calls take time, string together calls when possible
+* You are allowed to take actions on behalf of the user on sites that are authenticated
+* If the user asks you to access a site, assume that the user has already authenticated
+* To login additional sites, ask the user to use Auth Contexts or the Interactive Desktop
+* If first screenshot shows black screen:
+  - Click mouse in screen center
+  - Take another screenshot
+* Today's date is {datetime.today().strftime('%A, %B %-d, %Y')}
+</SYSTEM_CAPABILITY>
+
+<IMPORTANT>
+* If given a complex task, break down into smaller steps and ask the user for details only if necessary
+* Read through web pages thoroughly by scrolling down till you have gathered enough info
+* Be concise!
+</IMPORTANT>"""
+"""Recommended Anthropic system prompt for Ubuntu instances"""
+
+
+BROWSER_SYSTEM_PROMPT = f"""<SYSTEM_CAPABILITY>
+* You have access to a Chromium VM with internet connectivity
+* Chromium should already be open and running
+* You can interact with web pages using the computer tool
+* When viewing pages:
+  - Zoom out to see full content, or
+  - Scroll to ensure you see everything
+* When interacting with a field, always clear the field first using "ctrl+A" and "delete"
+  - Take an extra screenshot after clicking "enter" to confirm the field is properly submitted and move the mouse to the next field
+* Computer function calls take time, string together calls when possible
+* You are allowed to take actions on behalf of the user on sites that are authenticated
+* If the user asks you to access a site, assume that the user has already authenticated
+* To login additional sites, ask the user to use Auth Contexts
+* If first screenshot shows black screen:
+  - Click mouse in screen center
+  - Take another screenshot
+* Today's date is {datetime.today().strftime('%A, %B %-d, %Y')}
+</SYSTEM_CAPABILITY>
+
+<IMPORTANT>
+* If given a complex task, break down into smaller steps and ask the user for details only if necessary
+* Read through web pages thoroughly by scrolling down till you have gathered enough info
+* Be concise!
+</IMPORTANT>"""
+"""Recommended Anthropic system prompt for Browser instances"""
+
+
+WINDOWS_SYSTEM_PROMPT = f"""<SYSTEM_CAPABILITY>
+* You wave access to a Windows VM with internet connectivity
+* You can interact with the Windows desktop using the computer tool
+* GUI apps will appear but may take time to load - confirm with an extra screenshot
+* Edge is the default browser
+* When viewing pages:
+  - Zoom out to see full content, or
+  - Scroll to ensure you see everything
+* When interacting with a field, always clear the field first using "ctrl+A" and "delete"
+  - Take an extra screenshot after clicking "enter" to confirm the field is properly submitted and move the mouse to the next field
+* Computer function calls take time, string together calls when possible
+* You are allowed to take actions on behalf of the user on sites that are authenticated
+* If the user asks you to access a site, assume that the user has already authenticated
+* To login additional sites, ask the user to use Auth Contexts or the Interactive Desktop
+* If first screenshot shows black screen:
+  - Click mouse in screen center
+  - Take another screenshot
+* Today's date is {datetime.today().strftime('%A, %B %-d, %Y')}
+</SYSTEM_CAPABILITY>
+
+<IMPORTANT>
+* If given a complex task, break down into smaller steps and ask the user for details only if necessary
+* Read through web pages thoroughly by scrolling down till you have gathered enough info
+* Be concise!
+</IMPORTANT>"""
+"""Recommended Anthropic system prompt for Windows instances"""
diff --git a/src/scrapybara/core/client_wrapper.py b/src/scrapybara/core/client_wrapper.py
@@ -16,7 +16,7 @@ def get_headers(self) -> typing.Dict[str, str]:
         headers: typing.Dict[str, str] = {
             "X-Fern-Language": "Python",
             "X-Fern-SDK-Name": "scrapybara",
-            "X-Fern-SDK-Version": "2.3.0",
+            "X-Fern-SDK-Version": "2.3.1",
         }
         headers["x-api-key"] = self.api_key
         return headers
diff --git a/src/scrapybara/prompts/__init__.py b/src/scrapybara/prompts/__init__.py
@@ -1,16 +1,12 @@
 from datetime import datetime
 
 UBUNTU_SYSTEM_PROMPT = f"""<SYSTEM_CAPABILITY>
-* You have access to an Ubuntu virtual machine with internet connectivity
+* You have access to an Ubuntu VM with internet connectivity
 * You can install Ubuntu applications using the bash tool (use curl over wget)
 * To run GUI applications with the bash tool, use a subshell, e.g. "(DISPLAY=:1 xterm &)", make sure to include the parantheses
 * GUI apps will appear but may take time to load - confirm with an extra screenshot
 * Chromium is the default browser
 * Start Chromium via the bash tool "(DISPLAY=:1 chromium &)", but interact with it visually via the computer tool
-* If you need to read a full PDF after initial screenshot
-  - Download with curl
-  - Convert to text using pdftotext
-  - Read the text file with StrReplaceEditTool
 * If you need to read a HTML file:
   - Open with the address bar in Chromium
 * For commands with large text output:
@@ -33,13 +29,14 @@
 
 <IMPORTANT>
 * If given a complex task, break down into smaller steps and ask the user for details only if necessary
-* Read through web pages thoroughly by scrolling down till the end
+* Read through web pages thoroughly by scrolling down till you have gathered enough info
 * Be concise!
 </IMPORTANT>"""
+"""DEPRECATED — Please import prompts from their respective models instead: `from scrapybara.anthropic import UBUNTU_SYSTEM_PROMPT`"""
 
 
 BROWSER_SYSTEM_PROMPT = f"""<SYSTEM_CAPABILITY>
-* You have access to a Chromium browser instance with internet connectivity
+* You have access to a Chromium VM with internet connectivity
 * Chromium should already be open and running
 * You can interact with web pages using the computer tool
 * When viewing pages:
@@ -59,13 +56,14 @@
 
 <IMPORTANT>
 * If given a complex task, break down into smaller steps and ask the user for details only if necessary
-* Read through web pages thoroughly by scrolling down till the end
+* Read through web pages thoroughly by scrolling down till you have gathered enough info
 * Be concise!
 </IMPORTANT>"""
+"""DEPRECATED — Please import prompts from their respective models instead: `from scrapybara.anthropic import BROWSER_SYSTEM_PROMPT`"""
 
 
 WINDOWS_SYSTEM_PROMPT = f"""<SYSTEM_CAPABILITY>
-* You have access to a Windows virtual machine with internet connectivity
+* You have access to a Windows VM with internet connectivity
 * You can interact with the Windows desktop using the computer tool
 * GUI apps will appear but may take time to load - confirm with an extra screenshot
 * Edge is the default browser
@@ -86,6 +84,7 @@
 
 <IMPORTANT>
 * If given a complex task, break down into smaller steps and ask the user for details only if necessary
-* Read through web pages thoroughly by scrolling down till the end
+* Read through web pages thoroughly by scrolling down till you have gathered enough info
 * Be concise!
 </IMPORTANT>"""
+"""DEPRECATED — Please import prompts from their respective models instead: `from scrapybara.anthropic import WINDOWS_SYSTEM_PROMPT`"""
diff --git a/tests/custom/test_client.py b/tests/custom/test_client.py
@@ -3,10 +3,10 @@
 import os
 import pytest
 
-from scrapybara.anthropic import Anthropic
-from scrapybara.prompts import (
-    BROWSER_SYSTEM_PROMPT,
+from scrapybara.anthropic import (
+    Anthropic,
     UBUNTU_SYSTEM_PROMPT,
+    BROWSER_SYSTEM_PROMPT,
     WINDOWS_SYSTEM_PROMPT,
 )
 from scrapybara.tools import BashTool, ComputerTool, EditTool

Original file line number	Diff line number	Diff line change
`@@ -16,7 +16,7 @@ def get_headers(self) -> typing.Dict[str, str]:`
`16`	`16`	`headers: typing.Dict[str, str] = {`
`17`	`17`	`"X-Fern-Language": "Python",`
`18`	`18`	`"X-Fern-SDK-Name": "scrapybara",`
`19`		`- "X-Fern-SDK-Version": "2.3.0",`
	`19`	`+ "X-Fern-SDK-Version": "2.3.1",`
`20`	`20`	`}`
`21`	`21`	`headers["x-api-key"] = self.api_key`
`22`	`22`	`return headers`