sourcegraph
feedback: Rate-limit on pro plan is really annoying
Version
1.18.0
Areas for Improvement
- [ ] UI/UX
- [ ] Onboarding
- [ ] Docs
- [ ] Chat
- [ ] Commands
- [ ] Context
- [X] Response Quality
- [X] Other
What needs to be improved? Please describe how this affects the user experience and include a screenshot.
The rate limiting on autocompletion in the Pro plan is so frustrating. Literally, on your pricing page, dashboard, and almost everywhere else, you say "Unlimited," but it isn't. At the same time, your autocompletions are too short and not so good at the end. Why do I pay?
Describe the solution you'd like to see
Please provide the service that I pay for.
Describe any alternatives that could be considered
No response
Additional context
No response
Really curious what you're doing that you hit that limit... I'm pretty much exclusively using gpt4o right now all day and haven't hit the limit (yet?). Is that a monthly or a daily limit?
I only use the autocomplete feature, I never use the chat except for the "fix" feature, I program mostly in Rust and it seems like there is a daily limit, but here's the problem I pay for unlimited autocompletes, this kind of tool for me is really useful for Rust (when it works) to autocomplete where clauses and that kind of stuff that can be redundant and predictable but need an LLM to work.
I just realised that I configured phi3:3.8b via local ollama hence I probably don't run into the autocomplete limit .. it seems to perform way better than whatever is the default -- I also mostly code in Rust
I was actually shocked by how well it performs -- I tested it against falcon2:11b-q8_0, gemma:7b, codegemma:7b, deepseek-coder:33b, deepseek-coder:6.7b, hell even codellama:70b. It's extremely lightweight and produces by far the best results. That's for autocomplete though; for more complex tasks I still resort to GPT4.
Hey - if you're still running into rate limit issues, what's your sourcegraph username so I can look into this for you?
@chenkc805 I stopped my subscription so I can't confirm if it was fixed, but the last time I tested it (~2 weeks ago) it was still an issue