TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Show HN: SDK for AI Coding with Long Running Tasks

2 pointsby sean_30 days ago
It seems that using Cursor has become really repetitive. My workflow typically goes like this: - Get a task to accomplish on a codebase - Research the web and codebase to create a plan - Send the entire plan to cursor and sit back for it to finish<p>This flow works really well with a high level of accuracy for the most part and it&#x27;s becoming obvious that the margin between where it is today and &quot;perfect&quot; is quickly closing.<p>Because of this flow, I feel that I can replicate it through an SDK that uses a AI cli coder tool with low level control.<p>In this SDK, you can also use the best coder in the world per: <a href="https:&#x2F;&#x2F;aider.chat&#x2F;docs&#x2F;leaderboards&#x2F;" rel="nofollow">https:&#x2F;&#x2F;aider.chat&#x2F;docs&#x2F;leaderboards&#x2F;</a><p>As shown here:<p>```python from cloudcode import Local import os<p>def main(): # Use the current directory cwd = os.getcwd() example_file = &quot;example.py&quot;<p><pre><code> # Initialize the Aider SDK in architect mode agent = Local( working_dir=cwd, model=&quot;o3&quot;, # Main (planner) model editor_model=&quot;gpt-4.1&quot;, # Editor model for implementing changes architect_mode=True, use_git=False, api_key=os.getenv(&quot;CLOUDCODE_API_KEY&quot;) ) # Create or overwrite a simple Python file to modify using the SDK agent.create_file( example_file, &quot;&quot;&quot;def add(a, b): return a + b</code></pre> &quot;&quot;&quot; )<p><pre><code> # Run a coding task using the two-model workflow result = agent.code( prompt=&quot;make this function super cool. just make the math functions better&quot;, editable_files=[example_file] ) # Print the results print(&quot;\nTask completed!&quot;) print(f&quot;Success: {result[&#x27;success&#x27;]}&quot;) print(&quot;\nChanges made:&quot;) print(result[&quot;diff&quot;]) # Display cost information print(&quot;\nTask completed!&quot;) </code></pre> if __name__ == &quot;__main__&quot;: main()<p>```<p>Given that you can mimic your workflows from cursor and claude code into this sdk, you shouldn&#x27;t have to sit around on your computer waiting for the long running task to finish.<p>That&#x27;s why I built in a Sandbox class that mounts the ai coder into a remote sandbox directory instead of a local folder so you can deploy an ai coder to the cloud, enabling use cases such as coding from your phone and the browser.<p>I have already found so many low-handing fruit use cases for this sdk to automate prompting, information gathering, and auto-docs as shown in our github <a href="https:&#x2F;&#x2F;github.com&#x2F;LMSystems-ai&#x2F;cloud-coding">https:&#x2F;&#x2F;github.com&#x2F;LMSystems-ai&#x2F;cloud-coding</a><p>But I want to take it a step further.<p>You can get started for free today with this sdk and see our favorable pricing model here: <a href="https:&#x2F;&#x2F;docs.cloudcoding.ai&#x2F;pricing" rel="nofollow">https:&#x2F;&#x2F;docs.cloudcoding.ai&#x2F;pricing</a>

no comments

no comments