AI Code Assistant Suggestions
Hey guys,
I've been meaning to equip my team with different code assistants for their daily workflow. While I do use Gemini via GitHub Copilot for my personal projects I cannot do that here as the work we do is extremely confidential, and cannot be exposed in any way out of our servers. Due to this limitation using closed source models (openAI, Gemini, etc) is out of the question.
While my team is open to experiment with Windsurf and Cursor using open source models to work on our codebase(s), them being closed source is a negative for us, and thus we're probably gonna stick with VSC, and the assistants we can get from VSC extensions.
Features that we definitely will use:
- switching between different models
- use of agentic modes and chat modes
- MCP servers
- complete codebase traversal
Things that are good to have but not strictly necessary:
- Ability to change the Sampling Temperature
- Browse Open Source Codebases (or the internet in general) for some fixes
I've been meaning to try different code assistants and am considering to check out Cline and Continue.dev, with ollama based models or pulling deepseek (or other) models from Hugging Face.
How have your experiences been with these 2?
Given my use cases and constraints is it wise to check them out or should I steer clear of any of them?
Are there any other ones that I should be on the lookout for?
I've been meaning to equip my team with different code assistants for their daily workflow. While I do use Gemini via GitHub Copilot for my personal projects I cannot do that here as the work we do is extremely confidential, and cannot be exposed in any way out of our servers. Due to this limitation using closed source models (openAI, Gemini, etc) is out of the question.
While my team is open to experiment with Windsurf and Cursor using open source models to work on our codebase(s), them being closed source is a negative for us, and thus we're probably gonna stick with VSC, and the assistants we can get from VSC extensions.
Features that we definitely will use:
- switching between different models
- use of agentic modes and chat modes
- MCP servers
- complete codebase traversal
Things that are good to have but not strictly necessary:
- Ability to change the Sampling Temperature
- Browse Open Source Codebases (or the internet in general) for some fixes
I've been meaning to try different code assistants and am considering to check out Cline and Continue.dev, with ollama based models or pulling deepseek (or other) models from Hugging Face.
How have your experiences been with these 2?
Given my use cases and constraints is it wise to check them out or should I steer clear of any of them?
Are there any other ones that I should be on the lookout for?
