[HN Gopher] Ask questions about Git repositories from a web UI
       ___________________________________________________________________
        
       Ask questions about Git repositories from a web UI
        
       Author : jcbhmr
       Score  : 18 points
       Date   : 2024-04-19 19:29 UTC (3 hours ago)
        
 (HTM) web link (greptile.com)
 (TXT) w3m dump (greptile.com)
        
       | vladdoster wrote:
       | Limiting by size in MiB is understandable for cost, but as a user
       | it put me off after trying multiple repositories (popular OSS and
       | personal) because I couldn't find one under that size and ended
       | up not trying it at all. Curious to know how much repo size
       | influenced the $20 pro cost.
        
       | CasperH2O wrote:
       | Seems nice, tried it on a really small repo of mine.
       | 
       | Painfully reminded me that my main project has a way to large
       | repo size because we store binaries in them.
       | 
       | Would be nice if they offered discounts for open source projects.
        
         | alkonaut wrote:
         | Can't it clone without fetching LFS files? Since the binaries
         | won't be analyzed anyway, they don't need to fetch them from
         | LFS.
         | 
         | (If you have large binaries that are not in LFS you should
         | consider converting them)
        
       | hoten wrote:
       | If the developers see this: any consideration for a free plan for
       | open source projects?
        
         | matrix_overload wrote:
         | Not the developer, but chances are, they are either using a
         | fine-tuned ChatGPT, or cloud-hosting something in the likes of
         | Llama. So they are getting quite a bill each month from one of
         | the 3 cloud landlords.
         | 
         | So until they find a good product/market fit, opening the
         | floodgates to free users would just drain whatever money they
         | raised faster than they would get any meaningful market
         | response.
        
       | nkozyra wrote:
       | I think this looks nice and compelling, but I wonder how you lock
       | down the seats here.
       | 
       | For an organization, they're going to be thinking "why not just
       | get one seat and share it?"
       | 
       | It's also a hard sell to developers, who can spin up a LLM+RAG a
       | lot easier than folks in spaces being similarly served by an
       | OpenAI-as-a-service proxy app. Pointing ollama at my codebase is
       | fairly trivial.
       | 
       | Not trying to rain on any parades; there are a lot of AI-powered
       | projects like this and the ones that are selling are those that
       | have proprietary data and serve less technical spaces.
        
         | replete wrote:
         | What method do you use for using oIlama with RAG like this? I
        
           | nkozyra wrote:
           | Vector storage of choice + Langchain
        
       | thaliaarchi wrote:
       | > Replace your docs
       | 
       | > Developers don't like writing docs. Now, they don't have to!
       | 
       | Well that's a terrible premise. AI can't get at design decisions
       | or anything that's not explicitly encoded in code, but still
       | connected. The "why" is extremely valuable.
        
       ___________________________________________________________________
       (page generated 2024-04-19 23:01 UTC)