Ling0925
/
suna
mirror of https://github.com/kortix-ai/suna.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'PRODUCTION' into sync/production

This commit is contained in:
sharath 2025-06-05 07:31:54 +00:00
parent 11e8b4f7bd a1a71a05de
commit 4ef4eeceb1
No known key found for this signature in database
9 changed files with 61 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
backend/agent/run.py
View File

@ -290,7 +290,9 @@ async def run_agent(
latest_user_message = await client.table('messages').select('*').eq('thread_id', thread_id).eq('type', 'user').order('created_at', desc=True).limit(1).execute()
if latest_user_message.data and len(latest_user_message.data) > 0:
data = json.loads(latest_user_message.data[0]['content'])
data = latest_user_message.data[0]['content']
if isinstance(data, str):
data = json.loads(data)
trace.update(input=data['content'])
while continue_execution and iteration_count < max_iterations:
@ -327,14 +329,16 @@ async def run_agent(
latest_browser_state_msg = await client.table('messages').select('*').eq('thread_id', thread_id).eq('type', 'browser_state').order('created_at', desc=True).limit(1).execute()
if latest_browser_state_msg.data and len(latest_browser_state_msg.data) > 0:
try:
browser_content = json.loads(latest_browser_state_msg.data[0]["content"])
browser_content = latest_browser_state_msg.data[0]["content"]
if isinstance(browser_content, str):
browser_content = json.loads(browser_content)
screenshot_base64 = browser_content.get("screenshot_base64")
screenshot_url = browser_content.get("screenshot_url")
screenshot_url = browser_content.get("image_url")
# Create a copy of the browser state without screenshot data
browser_state_text = browser_content.copy()
browser_state_text.pop('screenshot_base64', None)
browser_state_text.pop('screenshot_url', None)
browser_state_text.pop('image_url', None)
if browser_state_text:
temp_message_content_list.append({
@ -348,6 +352,7 @@ async def run_agent(
"type": "image_url",
"image_url": {
"url": screenshot_url,
"format": "image/jpeg"
}
})
elif screenshot_base64:
@ -369,7 +374,7 @@ async def run_agent(
latest_image_context_msg = await client.table('messages').select('*').eq('thread_id', thread_id).eq('type', 'image_context').order('created_at', desc=True).limit(1).execute()
if latest_image_context_msg.data and len(latest_image_context_msg.data) > 0:
try:
image_context_content = json.loads(latest_image_context_msg.data[0]["content"])
image_context_content = latest_image_context_msg.data[0]["content"] if isinstance(latest_image_context_msg.data[0]["content"], dict) else json.loads(latest_image_context_msg.data[0]["content"])
base64_image = image_context_content.get("base64")
mime_type = image_context_content.get("mime_type")
file_path = image_context_content.get("file_path", "unknown file")

2
backend/sandbox/README.md
View File

@ -20,7 +20,7 @@ You can modify the sandbox environment for development or to add new capabilitie
```
cd backend/sandbox/docker
docker compose build
docker push kortix/suna:0.1.2
docker push kortix/suna:0.1.3
```
3. Test your changes locally using docker-compose

10
backend/sandbox/docker/Dockerfile
View File

@ -96,11 +96,6 @@ WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
# Copy server script
COPY . /app
COPY server.py /app/server.py
COPY browser_api.py /app/browser_api.py
# Install Playwright and browsers with system dependencies
ENV PLAYWRIGHT_BROWSERS_PATH=/ms-playwright
# Install Playwright package first
@ -111,6 +106,11 @@ RUN playwright install chromium
# Verify installation
RUN python -c "from playwright.sync_api import sync_playwright; print('Playwright installation verified')"
# Copy server script
COPY . /app
COPY server.py /app/server.py
COPY browser_api.py /app/browser_api.py
# Set environment variables
ENV PYTHONUNBUFFERED=1
ENV CHROME_PATH=/ms-playwright/chromium-*/chrome-linux/chrome

31
backend/sandbox/docker/browser_api.py
View File

@ -1,5 +1,5 @@
from fastapi import FastAPI, APIRouter, HTTPException, Body
from playwright.async_api import async_playwright, Browser, Page
from playwright.async_api import async_playwright, Browser, BrowserContext, Page
from pydantic import BaseModel
from typing import Optional, List, Dict, Any
import asyncio
@ -282,6 +282,7 @@ class BrowserAutomation:
def __init__(self):
self.router = APIRouter()
self.browser: Browser = None
self.browser_context: BrowserContext = None
self.pages: List[Page] = []
self.current_page_index: int = 0
self.logger = logging.getLogger("browser_automation")
@ -341,6 +342,7 @@ class BrowserAutomation:
try:
self.browser = await playwright.chromium.launch(**launch_options)
self.browser_context = await self.browser.new_context(viewport={'width': 1024, 'height': 768})
print("Browser launched successfully")
except Exception as browser_error:
print(f"Failed to launch browser: {browser_error}")
@ -348,6 +350,7 @@ class BrowserAutomation:
print("Retrying with minimal options...")
launch_options = {"timeout": 90000}
self.browser = await playwright.chromium.launch(**launch_options)
self.browser_context = await self.browser.new_context(viewport={'width': 1024, 'height': 768})
print("Browser launched with minimal options")
try:
@ -356,13 +359,20 @@ class BrowserAutomation:
self.current_page_index = 0
except Exception as page_error:
print(f"Error finding existing page, creating new one. ( {page_error})")
page = await self.browser.new_page(viewport={'width': 1024, 'height': 768})
page = await self.browser_context.new_page()
print("New page created successfully")
self.pages.append(page)
self.current_page_index = 0
# Navigate directly to google.com instead of about:blank
await page.goto("https://www.google.com", wait_until="domcontentloaded", timeout=30000)
print("Navigated to google.com")
try:
self.browser_context.on("page", self.handle_page_created)
except Exception as e:
print(f"Error setting up page event handler: {e}")
traceback.print_exc()
print("Browser initialization completed successfully")
except Exception as e:
@ -372,8 +382,17 @@ class BrowserAutomation:
async def shutdown(self):
"""Clean up browser instance on shutdown"""
if self.browser_context:
await self.browser_context.close()
if self.browser:
await self.browser.close()
async def handle_page_created(self, page: Page):
"""Handle new page creation"""
await asyncio.sleep(0.5)
self.pages.append(page)
self.current_page_index = len(self.pages) - 1
print(f"Page created: {page.url}; current page index: {self.current_page_index}")
async def get_current_page(self) -> Page:
"""Get the current active page"""
@ -958,6 +977,7 @@ class BrowserAutomation:
# Give time for any navigation or DOM updates to occur
await page.wait_for_load_state("networkidle", timeout=5000)
await asyncio.sleep(1)
# Get updated state after action
dom_state, screenshot, elements, metadata = await self.get_updated_browser_state(f"click_coordinates({action.x}, {action.y})")
@ -977,6 +997,7 @@ class BrowserAutomation:
# Try to get state even after error
try:
await asyncio.sleep(1)
dom_state, screenshot, elements, metadata = await self.get_updated_browser_state("click_coordinates_error_recovery")
return self.build_action_result(
False,
@ -1076,7 +1097,7 @@ class BrowserAutomation:
await page.wait_for_load_state("networkidle", timeout=5000)
except Exception as wait_error:
print(f"Timeout or error waiting for network idle after click: {wait_error}")
await asyncio.sleep(1) # Fallback wait
await asyncio.sleep(1)
# Get updated state after action
dom_state, screenshot, elements, metadata = await self.get_updated_browser_state(f"click_element({action.index})")
@ -1161,6 +1182,7 @@ class BrowserAutomation:
# Fallback to xpath
await page.fill(f"//{element.tag_name}[{action.index}]", action.text)
await asyncio.sleep(1)
# Get updated state after action
dom_state, screenshot, elements, metadata = await self.get_updated_browser_state(f"input_text({action.index}, '{action.text}')")
@ -1192,6 +1214,7 @@ class BrowserAutomation:
page = await self.get_current_page()
await page.keyboard.press(action.keys)
await asyncio.sleep(1)
# Get updated state after action
dom_state, screenshot, elements, metadata = await self.get_updated_browser_state(f"send_keys({action.keys})")
@ -1267,7 +1290,7 @@ class BrowserAutomation:
try:
print(f"Attempting to open new tab with URL: {action.url}")
# Create new page in same browser instance
new_page = await self.browser.new_page()
new_page = await self.browser_context.new_page()
print(f"New page created successfully")
# Navigate to the URL

4
backend/sandbox/docker/docker-compose.yml
View File

@ -6,12 +6,12 @@ services:
dockerfile: ${DOCKERFILE:-Dockerfile}
args:
TARGETPLATFORM: ${TARGETPLATFORM:-linux/amd64}
image: kortix/suna:0.1.2.8
image: kortix/suna:0.1.3
ports:
- "6080:6080" # noVNC web interface
- "5901:5901" # VNC port
- "9222:9222" # Chrome remote debugging port
- "8000:8000" # API server port
- "8003:8003" # API server port
- "8080:8080" # HTTP server port
environment:
- ANONYMIZED_TELEMETRY=${ANONYMIZED_TELEMETRY:-false}

2
backend/utils/config.py
View File

@ -159,7 +159,7 @@ class Configuration:
STRIPE_PRODUCT_ID_STAGING: str = 'prod_SCgIj3G7yPOAWY'
# Sandbox configuration
SANDBOX_IMAGE_NAME = "kortix/suna:0.1.2.8"
SANDBOX_IMAGE_NAME = "kortix/suna:0.1.3"
SANDBOX_ENTRYPOINT = "/usr/bin/supervisord -n -c /etc/supervisor/conf.d/supervisord.conf"
# LangFuse configuration

2
docs/SELF-HOSTING.md
View File

@ -115,7 +115,7 @@ As part of the setup, you'll need to:
1. Create a Daytona account
2. Generate an API key
3. Create a Docker image:
- Image name: `kortix/suna:0.1.2.8`
- Image name: `kortix/suna:0.1.3`
- Entrypoint: `/usr/bin/supervisord -n -c /etc/supervisor/conf.d/supervisord.conf`
## Manual Configuration

15
frontend/src/components/thread/tool-views/BrowserToolView.tsx
View File

@ -65,7 +65,7 @@ export function BrowserToolView({
const [imageError, setImageError] = React.useState(false);
try {
const topLevelParsed = safeJsonParse<{ content?: string }>(toolContent, {});
const topLevelParsed = safeJsonParse<{ content?: any }>(toolContent, {});
const innerContentString = topLevelParsed?.content || toolContent;
if (innerContentString && typeof innerContentString === 'string') {
const toolResultMatch = innerContentString.match(/ToolResult\([^)]*output='([\s\S]*?)'(?:\s*,|\s*\))/);
@ -116,7 +116,18 @@ export function BrowserToolView({
screenshotUrl = finalParsedOutput?.image_url || null;
}
}
}
} else if (innerContentString && typeof innerContentString === "object") {
screenshotUrl = (() => {
if (!innerContentString) return null;
if (!("tool_execution" in innerContentString)) return null;
if (!("result" in innerContentString.tool_execution)) return null;
if (!("output" in innerContentString.tool_execution.result)) return null;
if (!("image_url" in innerContentString.tool_execution.result.output)) return null;
if (typeof innerContentString.tool_execution.result.output.image_url !== "string") return null;
return innerContentString.tool_execution.result.output.image_url;
})()
}
} catch (error) {
}

2
setup.py
View File

@ -237,7 +237,7 @@ def collect_daytona_info():
print_info("Then, generate an API key from 'Keys' menu")
print_info("After that, go to Images (https://app.daytona.io/dashboard/images)")
print_info("Click '+ Create Image'")
print_info(f"Enter 'kortix/suna:0.1.2.8' as the image name")
print_info(f"Enter 'kortix/suna:0.1.3' as the image name")
print_info(f"Set '/usr/bin/supervisord -n -c /etc/supervisor/conf.d/supervisord.conf' as the Entrypoint")
input("Press Enter to continue once you've completed these steps...")