Add endpoint to stream HAR file for tasks

[ngalongc]

Add endpoint to stream HAR file for tasks

Overview

This PR introduces a new endpoint that allows users to retrieve the HAR (HTTP Archive) file associated with a specific task. This addition is particularly valuable for AI security agents and developers who need access to raw network traffic data.

Key Features

• New GET endpoint: /tasks/{task_id}/har
• Streams HAR file content for efficient handling of large files
• Uses aiofiles for non-blocking file operations

Motivation

• Offers valuable insights into network traffic for debugging and analysis purposes
• Enhances the API's utility for users who require detailed HTTP transaction information

Implementation Details

• Utilizes StreamingResponse to handle large HAR files without memory issues
• Implements error handling for file not found and other potential exceptions
• Integrates with existing artifact management system

Testing

• Tested with various HAR file sizes to ensure proper streaming behavior
• Verified error handling for non-existent tasks and missing HAR files

Security Considerations

• Endpoint respects existing organization-based access controls
• No sensitive information is exposed beyond what's already in the HAR file

Future Improvements

• Consider adding compression options for very large HAR files
• Explore the possibility of filtering HAR content based on user-specified criteria

Feedback Request

I'd appreciate a review on:

1. The streaming implementation
2. Error handling approach - have I covered all necessary scenarios?
3. Any potential security concerns I might have overlooked

Looking forward to your feedback!

---

:rocket:	This description was created by Ellipsis for commit 5da44261909c202e38e6e016be45612ec7714d51

Summary:

Adds a new endpoint in skyvern/forge/sdk/routes/agent_protocol.py to stream HAR files for tasks using non-blocking operations and error handling.

Key points:

• Adds GET /tasks/{task_id}/har endpoint in skyvern/forge/sdk/routes/agent_protocol.py.
• Streams HAR file content using StreamingResponse and aiofiles.
• Implements get_task_har function to retrieve HAR files for tasks.
• Introduces stream_file function for non-blocking file reading.
• Handles errors for missing files and other exceptions.
• Integrates with existing artifact management and access control systems.

---

Generated with :heart: by ellipsis.dev

[ngalongc]

I will try to fix this asap, max 1 week.

[suchintan]

I will try to fix this asap, max 1 week.

Sounds good. Happy to merge it in afterwards!