Integration with Copilot is very slow

[lispercat]

Maybe there is a quick fix for it, but on the same machine when I talk to Copilot (for company account) via the Web interface it takes at most 2 sec, even when it has a lot of context and searches company files.
When CopilotChat it takes ~10 sec even for very simple questions in which I don't ask to analyze any code.
Is there a way to accelerate the process?

[deathbeam]

Is it only first question or further questions as well? Also depends on model, gpt-4o is a lot faster than claude for example. But on first question the plugin needs to fetch models, agents and check policies, question after will cache this data so should be fine.

[lispercat]

it's all questions, first and subsequent. The default gpt-4o is selected.
If I open nvim and just open CopilotChat it's pretty quick but as soon as I open a file from a project it gets really sluggish and seems like to get slower as I keep asking questions.

[deathbeam]

it's all questions, first and subsequent. The default gpt-4o is selected. If I open nvim and just open CopilotChat it's pretty quick but as soon as I open a file from a project it gets really sluggish and seems like to get slower as I keep asking questions.

well are you opening big file? as the content of the file needs to be sent every time when asking the question (default behaviour is sent whole buffer). and the history also needs to be sent every time, thats how every LLM works mostly. 10 seconds is still a lot tho, maybe it could be also curl related?

Can you output what :checkhealth CopilotChat says?

[deathbeam]

I merged some optimizations but im still curious about your curl version + the size of file

[lispercat]

Here is the checkhealth output:
CopilotChat: require("CopilotChat.health").check()

CopilotChat.nvim ~

• If you are facing any issues, also see :CopilotChatDebugInfo for more information.

CopilotChat.nvim [core] ~

• OK nvim: NVIM v0.10.1
  Build type: RelWithDebInfo
  LuaJIT 2.1.1713484068
  Run ":verbose version" for more info

CopilotChat.nvim [commands] ~

• OK curl: curl 7.81.0 (x86_64-pc-linux-gnu) libcurl/7.81.0 OpenSSL/3.0.2 zlib/1.2.11 brotli/1.0.9 zstd/1.4.8 libidn2/2.3.2 libpsl/0.21.0 (+libidn2/2.3.2) libssh/0.9.6/openssl/zlib nghttp2/1.43.0 librtmp/2.3 OpenLDAP/2.5.17
  Release-Date: 2022-01-05
  Protocols: dict file ftp ftps gopher gophers http https imap imaps ldap ldaps mqtt pop3 pop3s rtmp rtsp scp sftp smb smbs smtp smtps telnet tftp
  Features: alt-svc AsynchDNS brotli GSS-API HSTS HTTP2 HTTPS-proxy IDN IPv6 Kerberos Largefile libz NTLM NTLM_WB PSL SPNEGO SSL TLS-SRP UnixSockets zstd
• OK git: git version 2.34.1

CopilotChat.nvim [dependencies] ~

• OK plenary: installed
• OK copilot: copilot.lua
• WARNING tiktoken_core: missing, optional for accurate token counting. See README for installation instructions.
• OK treesitter[markdown]: installed
• WARNING treesitter[diff]: missing, optional for better diff highlighting. Install "nvim-treesitter/nvim-treesitter" plugin and run ":TSInstall diff".

[deathbeam]

Can you try on latest canary? Added some status reporting for embedding files as well.

Also how big is the file again? char count/line numbers will do.

You could also try upgrading curl, 7.81 is very old and see if it helps.

[lispercat]

Looks like upgrading to latest curl (latest from github) doesn't change much.
Also what I noticed is that even for a small project with small file sizes, as the time goes by it gets slower and slower to the point that nvim takes 100% of CPU and I need to restart it. Maybe it's just happening to me.