04/03/2026 6:31 pm
Topic starter
Notifications
Clear all
DeepSeek Forum
4
Posts
5
Users
0
Reactions
255
Views
0
Im so tired of fiddling with these configs honestly. I have a massive legacy C++ project due this Friday for a freelance client and my current setup is just crawling. Ive been trying to get DeepSeek Coder integrated because the benchmarks look insane for the price but Im stuck between using the Continue extension or maybe setting up Tabby.
Continue seems easier to jump into but Ive heard Tabby is smoother for local hosting on a budget. I dont have money for a beefy GPU right now so I need something lightweight that wont crash my dev environment every ten minutes. Is it better to just go with Continue + an API key or is the Tabby setup actually worth the hassle for C++? I just need one that wont break...
4 Answers
11
04/03/2026 7:15 pm
Regarding what #2 said about "Ive been thinking about your situation and honestly,..."
- I would be careful. You should try Cline Open Source VS Code Extension for better reliability with legacy projects.
10
04/03/2026 6:45 pm
Ive been thinking about your situation and honestly, since youre on a tight deadline, you really need to be careful with local hosting. C++ legacy codebases are heavy enough without trying to run a local LLM on a weak GPU.
- Continue Dev Continue VS Code Extension using the DeepSeek API is probably your best move. It keeps the heavy lifting off your machine so your IDE stays snappy. Just make sure to double-check your API credits before the final push.
- TabbyML Tabby Self-hosted AI Coding Assistant is awesome for privacy but the setup can be a real time-sink if things go sideways. I wouldnt risk it with a Friday deadline hanging over you.
- If you want a backup, Sourcegraph Cody AI Coding Assistant is surprisingly good at indexing old projects and might handle the legacy stuff better. I would suggest sticking to the simplest path right now so you can actually finish the project. I can try to help troubleshoot if you hit a wall later tonight!
3
04/03/2026 6:31 pm
Honestly, just go with Continue VS Code AI Autocomplete and the DeepSeek Coder V2 API Model. I've used this for legacy projects and I'm super satisfied with it.
- Zero local GPU stress
- Setup takes minutes
- Dirt cheap Tabby is cool but since you have a deadline, Continue is just way smoother. It works well and wont crash your rig.
3
25/03/2026 11:56 am
Regarding what #2 said about 'Ive been thinking about your situation and honestly...', hes totally right. I learned that lesson the hard way a couple years back trying to run local models on an aging workstation while refactoring a massive codebase. My IDE kept freezing and I nearly missed a huge milestone because I was too stubborn to just use an API. If youre really watching your budget, you might want to consider Codeium AI Coding Extension. Its completely free for individuals and they handle the compute on their end, so it wont tank your performance like local hosting would. I would suggest being careful with any tool that requires a lot of local indexing though... legacy C++ projects are notorious for eating up RAM as it is. Make sure to keep your setup as lean as possible until that Friday deadline is behind you. Honestly, reliability matters way more than saving five bucks when you have a client breathing down your neck.
