#tabbyml — Public Fediverse posts

Мы прожили с Copilot год, и вот что из этого вышло

Я год экспериментировал с on-premise Copilot — прямо над нашими разработчиками, — чтобы проверить: а правда ли эта штука разгоняет разработку на десятки процентов? Делюсь реальными метриками скорости и точности, разбираю, как оно работает на примере нашей инсталляции, и показываю результаты. По пути расскажу про все подводные камни: где ИИ стал турбоускорителем, а где подставил подножку и превратился в скрытую ловушку.

https://habr.com/ru/companies/k2tech/articles/1017022/

#copilot #ии #ииассистент #разработка #k2_cloud #к2_облако #искусственный_интеллект #continuedev #tabbyml #refactai

Мы прожили с Copilot год, и вот что из этого вышло

Я год экспериментировал с on-premise Copilot — прямо над нашими разработчиками, — чтобы проверить: а правда ли эта штука разгоняет разработку на десятки процентов? Делюсь реальными метриками скорости и точности, разбираю, как оно работает на примере нашей инсталляции, и показываю результаты. По пути расскажу про все подводные камни: где ИИ стал турбоускорителем, а где подставил подножку и превратился в скрытую ловушку.

https://habr.com/ru/companies/k2tech/articles/1017022/

#copilot #ии #ииассистент #разработка #k2_cloud #к2_облако #искусственный_интеллект #continuedev #tabbyml #refactai

Мы прожили с Copilot год, и вот что из этого вышло

Я год экспериментировал с on-premise Copilot — прямо над нашими разработчиками, — чтобы проверить: а правда ли эта штука разгоняет разработку на десятки процентов? Делюсь реальными метриками скорости и точности, разбираю, как оно работает на примере нашей инсталляции, и показываю результаты. По пути расскажу про все подводные камни: где ИИ стал турбоускорителем, а где подставил подножку и превратился в скрытую ловушку.

https://habr.com/ru/companies/k2tech/articles/1017022/

#copilot #ии #ииассистент #разработка #k2_cloud #к2_облако #искусственный_интеллект #continuedev #tabbyml #refactai

Мы прожили с Copilot год, и вот что из этого вышло

Я год экспериментировал с on-premise Copilot — прямо над нашими разработчиками, — чтобы проверить: а правда ли эта штука разгоняет разработку на десятки процентов? Делюсь реальными метриками скорости и точности, разбираю, как оно работает на примере нашей инсталляции, и показываю результаты. По пути расскажу про все подводные камни: где ИИ стал турбоускорителем, а где подставил подножку и превратился в скрытую ловушку.

https://habr.com/ru/companies/k2tech/articles/1017022/

#copilot #ии #ииассистент #разработка #k2_cloud #к2_облако #искусственный_интеллект #continuedev #tabbyml #refactai

I've been using Tabby, a self-hosted AI coding assistant, for the last week or so, and have found it quite useful at times. Particularly when working on a project with a lot of boilerplate, as it acts like a clever context aware copy and paste #tabbyml #genai #neovim https://www.tabbyml.com/

I've been using Tabby, a self-hosted AI coding assistant, for the last week or so, and have found it quite useful at times. Particularly when working on a project with a lot of boilerplate, as it acts like a clever context aware copy and paste #tabbyml #genai #neovim https://www.tabbyml.com/

I've been using Tabby, a self-hosted AI coding assistant, for the last week or so, and have found it quite useful at times. Particularly when working on a project with a lot of boilerplate, as it acts like a clever context aware copy and paste #tabbyml #genai #neovim https://www.tabbyml.com/

I've been using Tabby, a self-hosted AI coding assistant, for the last week or so, and have found it quite useful at times. Particularly when working on a project with a lot of boilerplate, as it acts like a clever context aware copy and paste #tabbyml #genai #neovim https://www.tabbyml.com/

I'm trying out the GH #Copilot alternative #TabbyML today - it was very easy to set up! For chat, I don't notice a huge difference with my trial run of Copilot a while ago. It seems as good, and as shitty, which is impressive.

FYI; I'm using StarCoder2-3B for code completion, and CodeGemma-7B-Instruct for chat. That comes in at about 14GB in total storage size. Memory usage increases with about ~5-11GB when asking a chat question.

I'm trying out the GH #Copilot alternative #TabbyML today - it was very easy to set up! For chat, I don't notice a huge difference with my trial run of Copilot a while ago. It seems as good, and as shitty, which is impressive.

FYI; I'm using StarCoder2-3B for code completion, and CodeGemma-7B-Instruct for chat. That comes in at about 14GB in total storage size. Memory usage increases with about ~5-11GB when asking a chat question.

I'm trying out the GH #Copilot alternative #TabbyML today - it was very easy to set up! For chat, I don't notice a huge difference with my trial run of Copilot a while ago. It seems as good, and as shitty, which is impressive.

FYI; I'm using StarCoder2-3B for code completion, and CodeGemma-7B-Instruct for chat. That comes in at about 14GB in total storage size. Memory usage increases with about ~5-11GB when asking a chat question.

Writing an article to explain what is #textual, a #Python package to build rich UI in terminal. And I realized that I could create a chatbot inside #neovim

Maybe #llm-ls and #TabbyML could be interested.

#ai #tui #textualize

FYI, the package is https://textual.textualize.io/

Writing an article to explain what is #textual, a #Python package to build rich UI in terminal. And I realized that I could create a chatbot inside #neovim

Maybe #llm-ls and #TabbyML could be interested.

#ai #tui #textualize

FYI, the package is https://textual.textualize.io/

Writing an article to explain what is #textual, a #Python package to build rich UI in terminal. And I realized that I could create a chatbot inside #neovim

Maybe #llm-ls and #TabbyML could be interested.

#ai #tui #textualize

FYI, the package is https://textual.textualize.io/

Writing an article to explain what is #textual, a #Python package to build rich UI in terminal. And I realized that I could create a chatbot inside #neovim

Maybe #llm-ls and #TabbyML could be interested.

#ai #tui #textualize

FYI, the package is https://textual.textualize.io/

Writing an article to explain what is #textual, a #Python package to build rich UI in terminal. And I realized that I could create a chatbot inside #neovim

Maybe #llm-ls and #TabbyML could be interested.

#ai #tui #textualize

FYI, the package is https://textual.textualize.io/

I've come across #tabbyml and have started using it as a replacement for GH copilot. It's open-source, comes with a VSCode extension and allows you to host your own LLM for chat and code complete (also locally). Currently running the StarCoder2-3B for code complete on my MacBook M2. First impression: model unsurprisingly less powerful than copilot but inference time is fast enough. Since it's running locally I don't even need WiFi. Very neat!

https://tabby.tabbyml.com

I've come across #tabbyml and have started using it as a replacement for GH copilot. It's open-source, comes with a VSCode extension and allows you to host your own LLM for chat and code complete (also locally). Currently running the StarCoder2-3B for code complete on my MacBook M2. First impression: model unsurprisingly less powerful than copilot but inference time is fast enough. Since it's running locally I don't even need WiFi. Very neat!

https://tabby.tabbyml.com

I've come across #tabbyml and have started using it as a replacement for GH copilot. It's open-source, comes with a VSCode extension and allows you to host your own LLM for chat and code complete (also locally). Currently running the StarCoder2-3B for code complete on my MacBook M2. First impression: model unsurprisingly less powerful than copilot but inference time is fast enough. Since it's running locally I don't even need WiFi. Very neat!

https://tabby.tabbyml.com

I've come across #tabbyml and have started using it as a replacement for GH copilot. It's open-source, comes with a VSCode extension and allows you to host your own LLM for chat and code complete (also locally). Currently running the StarCoder2-3B for code complete on my MacBook M2. First impression: model unsurprisingly less powerful than copilot but inference time is fast enough. Since it's running locally I don't even need WiFi. Very neat!

https://tabby.tabbyml.com

I've come across #tabbyml and have started using it as a replacement for GH copilot. It's open-source, comes with a VSCode extension and allows you to host your own LLM for chat and code complete (also locally). Currently running the StarCoder2-3B for code complete on my MacBook M2. First impression: model unsurprisingly less powerful than copilot but inference time is fast enough. Since it's running locally I don't even need WiFi. Very neat!

https://tabby.tabbyml.com

I wanted to jump up on the AI-powered-dev-assistant bandwagon, so I installed TabbyML into neovim and then ended up in a small rathole trying to get systemd to act like inetd. Details on my blog at https://place.org/~pj/blog/using-systemd-as-user-level-inetd.html

#TabbyML #neovim #systemd #inetd

I wanted to jump up on the AI-powered-dev-assistant bandwagon, so I installed TabbyML into neovim and then ended up in a small rathole trying to get systemd to act like inetd. Details on my blog at https://place.org/~pj/blog/using-systemd-as-user-level-inetd.html

#TabbyML #neovim #systemd #inetd