Sentry is a popular error tracking tool that helps developers monitor and fix crashes in real-time. By integrating Arcane Engine with Sentry, error analysis becomes a breeze.
- A Sentry account
- A project set up on Sentry
- Sentry integration enabled in Arcane Engine
A convenient way to do this is to start a chat with Arcane Engine:
➜ error-analysis git:(main) ✗ ./run.sh
╭──────────────── You ─────────────────╮
│ Find all unresolved issues on Sentry │
╰──────────────────────────────────────╯
I found 14 unresolved issues on Sentry for the python-django project. Here are some of them:
1 Error searching Sentry issues: 400 Client Error: Bad Request for url: https://sentry.io/api/0/pro...
• ID: 5551588934
• URL: Link
• Count: 2
• First Seen: 2024-06-29T04:01:48.208162Z
• Last Seen: 2024-06-29T04:01:51.131428Z
2 IndexError: list index out of range
• ID: 5459092301
• URL: Link
• Count: 2
• First Seen: 2024-06-06T18:03:24.793070Z
• Last Seen: 2024-06-27T19:26:39.889151Z
3 {'errors': [{'message': 'Operation error', 'extensions': {'code': 'INPUT_ERROR', 'type': 'invalid...
• ID: 5541757021
• URL: Link
• Count: 2
• First Seen: 2024-06-27T01:33:22.298160Z
• Last Seen: 2024-06-27T01:35:26.344895Z
4 Error searching Linear workspace: Operation error
• ID: 5541756956
• URL: Link
• Count: 2
• First Seen: 2024-06-27T01:33:22.273787Z
• Last Seen: 2024-06-27T01:35:24.515479Z
5 BadRequestError: Error code: 400 - {'error': {'message': "This model's maximum context length is 16385 tokens. How...
• ID: 5520706968
• URL: Link
• Count: 16
• First Seen: 2024-06-22T02:42:18.355965Z
• Last Seen: 2024-06-22T05:11:32.691311Z
For more details, you can visit the Sentry dashboard. We found some issues, let's pick one and investigate further:
╭───────────────── You ──────────────────╮
│ Check out the events of the IndexError │
╰────────────────────────────────────────╯
I found 2 events for the IndexError: list index out of range issue (ID: 5459092301):
1 Event Timestamp: 2024-06-27T19:26:39Z
• Message: Failed to run task
• Location: engine/agents/web_search_agent.py
• Culprit: engine.agents.web_search_agent in scrape_website
2 Event Timestamp: 2024-06-06T18:03:24Z
• Message: Failed to run task
• Location: engine/agents/web_search_agent.py
• Culprit: engine.agents.web_search_agent in scrape_website
For more details, you can visit the Sentry issue page.
Seems like there's something wrong with how we scrape websites. Arcane Engine can find and fix the relevant code:
╭──────────────────────────── You ────────────────────────────╮
│ Find the relevant code, analyze the issue and provide a fix │
╰─────────────────────────────────────────────────────────────╯
I have analyzed the IndexError: list index out of range issue in the scrape_website function and applied a fix. The issue was caused by attempting to access the first element of
docs_transformed[0].page_content without checking if it was empty. I added a check to ensure that docs_transformed[0].page_content is not empty before accessing its elements.
The updated scrape_website function now looks like this:
@tool
def scrape_website(url: str):
"""Scrape a website."""
TaskEvent.add(actor="Darwin", action="scrape", target=url, transaction="begin")
urls = [url]
loader = AsyncHtmlLoader(urls)
docs = loader.load()
html2text = Html2TextTransformer()
docs_transformed = html2text.transform_documents(docs)
if docs_transformed and docs_transformed[0].page_content:
return docs_transformed[0].page_content[0:700]
else:
logger.error(f"Failed to scrape content from {url}")
return "Failed to scrape content"
This fix ensures that we only attempt to access docs_transformed[0].page_content[0:700] if it is not empty, preventing the IndexError.
The changes have been committed to the codebase.