Releases: serge-chat/serge
Releases · serge-chat/serge
0.1.1
Changelog
Fixed a bug with the release where you couldn't build the image locally
Fixed the wrong python version in the docker image
Fixed PR workflow that wasn't releasing tagged images properly
What's Changed
- CI Fixes and Updates by @gaby in #224
- Add missing fields to cleanup job by @gaby in #225
- Remove PR tagged builds by @gaby in #226
Full Changelog: 0.1.0...0.1.1
Contributors
gaby
Assets 2
0.1.0 - Moar Models
Changelog
Quite a few things, but in no particular order:
- We got rid of the custom wrapper around llama.cpp and we now use the awesome python bindings provided by
llama-cpp-python. - We also got rid of MongoDB, now Redis is used to handle the chat history using a wrapper provided by langchain. Getting rid of MongoDB means Serge should now be fully compatible with older CPUs that don't support AVX instructions 🎉 Please file a bug report if you still have compatibility issues with your specific hardware.
- We cleaned up the UI a bit and made the colors more uniform all around.
- We added the ability to delete an already downloaded model.
- Speaking of models, we added support for the open assistant models as well as Vicuna. They should be available from the download page in Serge.
- As a result of those changes, plus some work by @gaby the final image should now be a lot lighter.
- We fixed a bug related to passing the temperature parameter in the UI. Thanks @agronholm and welcome!
And we're finally bumping the version to 0.1, since we got rid of MongoDB it seemed like the right time.
What's Changed
- Bump daisyui from 2.51.4 to 2.51.5 in /web by @dependabot in #185
- Bump huggingface-hub from 0.13.3 to 0.13.4 in /api by @dependabot in #191
- Bump websockets from 11.0.1 to 11.0.2 in /api by @dependabot in #192
- Bump motor from 3.1.1 to 3.1.2 in /api by @dependabot in #187
- Bump @typescript-eslint/eslint-plugin from 5.58.0 to 5.59.0 in /web by @dependabot in #188
- Update llama.cpp to master-dcdd65e by @gaby in #193
- Bump @typescript-eslint/parser from 5.58.0 to 5.59.0 in /web by @dependabot in #186
- Bump typescript from 4.9.5 to 5.0.4 in /web by @dependabot in #189
- Bump eslint from 8.36.0 to 8.38.0 in /web by @dependabot in #190
- Update Dockerfile to fix CI/CD by @gaby in #194
- Bump filelock from 3.11.0 to 3.12.0 in /api by @dependabot in #198
- Bump @sveltejs/kit from 1.15.5 to 1.15.7 in /web by @dependabot in #195
- Bump @iconify/svelte from 3.1.1 to 3.1.2 in /web by @dependabot in #197
- Bump vite from 4.2.1 to 4.2.2 in /web by @dependabot in #196
- Bump sse-starlette from 1.3.3 to 1.3.4 in /api by @dependabot in #200
- Update llama.cpp tag to master-f7d0509 by @gaby in #201
- Bump @iconify/svelte from 3.1.2 to 3.1.3 in /web by @dependabot in #205
- Bump postcss from 8.4.22 to 8.4.23 in /web by @dependabot in #203
- Bump vite from 4.2.2 to 4.3.0 in /web by @dependabot in #206
- Bump email-validator from 2.0.0.post1 to 2.0.0.post2 in /api by @dependabot in #204
- Fixed name of the temperature parameter by @agronholm in #207
- Bump vite from 4.3.0 to 4.3.1 in /web by @dependabot in #208
- Bug fixes and improvements by @gaby in #202
- Use python bindings, integrate with LangChain and get rid of MongoDB by @nsarrazin in #148
- Bump loguru from 0.6.0 to 0.7.0 in /api by @dependabot in #216
- Bump eslint from 8.38.0 to 8.39.0 in /web by @dependabot in #215
- Use Debian Slim for base images, Fixes to CI actions by @gaby in #221
New Contributors
- @agronholm made their first contribution in #207
Full Changelog: 0.0.5...0.1.0
Contributors
agronholm, gaby, and 2 other contributors
Assets 2
Serge 0.0.5 - Mostly Maintenance
Updates
Mostly a housekeeping update, here are the things of note:
- @gaby introduced dependabot, and a new github action allowing us to create docker images for incoming PRs! This will let us test without having to pull everything locally, which should be great.
- @snxraven added a new button to start a new chat with the same parameters as the current chat, which will save you the trouble of reinputting the same parameters every time.
- @johncadengo fixed a bug where the context window length wasn't being passed properly, this should allow for longer chats if you increase the context window in the UI.
- We're up to date with the latest
llama.cppchanges, which should bring improved performance and hardware compatibility.
What's Changed
- Add link to llama.cpp git repo by @gaby in #126
- Performance improvements for AVX2 via llama.cpp by @gaby in #129
- Correctly pass the context window length by @johncadengo in #138
- Update llama.cpp to master-eeaa7b0 by @gaby in #144
- Adding a new session button which uses current session settings by @snxraven in #136
- Update llama.cpp to master-180b693 by @gaby in #147
- Update llama.cpp to master-be87b6e by @gaby in #153
- Fix LLaMA spelling mistakes by @gaby in #160
- Expand dependabot coverage by @gaby in #161
- Update llama.cpp tag to master-489537e by @gaby in #172
- Bump @sveltejs/adapter-static from 2.0.1 to 2.0.2 in /web by @dependabot in #167
- Bump fastapi from 0.95.0 to 0.95.1 in /api by @dependabot in #170
- Bump @iconify/svelte from 3.1.0 to 3.1.1 in /web by @dependabot in #166
- Bump vite from 4.2.0 to 4.2.1 in /web by @dependabot in #162
- Bump svelte-check from 3.1.4 to 3.2.0 in /web by @dependabot in #164
- Bump beanie from 1.17.0 to 1.18.0 in /api by @dependabot in #168
- Bump @typescript-eslint/parser from 5.56.0 to 5.58.0 in /web by @dependabot in #163
- Bump filelock from 3.10.7 to 3.11.0 in /api by @dependabot in #165
- Build and publish PR images by @gaby in #173
- Bump email-validator from 1.3.1 to 2.0.0.post1 in /api by @dependabot in #169
- Update docker-image.yml by @gaby in #183
- Bump packaging from 23.0 to 23.1 in /api by @dependabot in #182
- Bump chipkent/action-cleanup-package from 1.0.1 to 1.0.3 by @dependabot in #174
- Bump sentencepiece from 0.1.97 to 0.1.98 in /api by @dependabot in #175
- Bump svelte from 3.57.0 to 3.58.0 in /web by @dependabot in #180
- Bump orjson from 3.8.9 to 3.8.10 in /api by @dependabot in #179
- Bump postcss from 8.4.21 to 8.4.22 in /web by @dependabot in #178
- Bump @typescript-eslint/eslint-plugin from 5.56.0 to 5.58.0 in /web by @dependabot in #176
- Bump tailwindcss from 3.2.7 to 3.3.1 in /web by @dependabot in #181
- Bump websockets from 10.4 to 11.0.1 in /api by @dependabot in #171
- Bump @sveltejs/kit from 1.12.0 to 1.15.5 in /web by @dependabot in #177
New Contributors
- @johncadengo made their first contribution in #138
- @dependabot made their first contribution in #167
Full Changelog: 0.0.4...0.0.5
Contributors
johncadengo, gaby, and 2 other contributors
Assets 2
Serge 0.0.4 - GPT4All Support
Cool things added:
- GPT4All support has been added! You can just grab it from the download manager.
- Improved support for CPUs with AVX but no AVX2 support
- Hitting ctrl+enter lets you send messages
- Files added to the /usr/src/app/weights folder will now be automatically picked up
What's Changed
- docs: correct the name 7B-native on Readme by @pabl-o-ce in #120
- Update to latest llama.cpp by @nsarrazin in #118
- Update llama.cpp to master-3525899 by @gaby in #122
- When a user hits CTRL+ENTER send input by @snxraven in #121
- main.py now recursively looks for .bin files. by @axolotlite in #46
- Update llama.cpp to master-e986f94 by @gaby in #124
New Contributors
- @pabl-o-ce made their first contribution in #120
- @snxraven made their first contribution in #121
- @axolotlite made their first contribution in #46
Full Changelog: 0.0.3...0.0.4
Contributors
gaby, snxraven, and 3 other contributors
Assets 2
Serge 0.0.3! Now with a download manager.
Get started
docker run -d -v weights:/usr/src/app/weights -v datadb:/data/db/ -p 8008:8008 ghcr.io/nsarrazin/serge:latestAnd that's literally it. Just open http://localhost:8008/ to see the available models.
What's Changed
- Improvements to github actions by @gaby in #79
- Add fix for login-action into ghcr.io by @gaby in #81
- Proper dev & deployment images. Converts front-end to SPA. by @nsarrazin in #84
- Create CODE_OF_CONDUCT.md by @nsarrazin in #88
- Move the llama.cpp compilation step to after the image is built by @nsarrazin in #87
- Set default "n_threads" value to half of the available threads by @ryan-mca in #86
- Update README.md by @paraskevasleivadaros in #94
- API Refactor & Model Manager by @nsarrazin in #101
- Update llama.cpp to master-5a5f8b1 by @gaby in #103
New Contributors
- @gaby made their first contribution in #79
- @ryan-mca made their first contribution in #86
- @paraskevasleivadaros made their first contribution in #94
Full Changelog: 0.0.2...0.0.3
Contributors
gaby, paraskevasleivadaros, and 2 other contributors
Assets 2
0.0.2
What's Changed
- Initial support for Kubernetes by @FenarkSEC in #71
- run the image deployment process when a release is published by @nsarrazin in #74
New Contributors
- @FenarkSEC made their first contribution in #71
Full Changelog: 0.0.1...0.0.2
Contributors
nsarrazin and FenarkSEC
Assets 2
Serge 0.0.1
What's Changed
First release! Let's see if the github actions work.
New Contributors
- @nsarrazin made their first contribution in #14
- @thomasleveil made their first contribution in #22
- @oganexon made their first contribution in #51
- @mavaa made their first contribution in #39
- @louisoutin made their first contribution in #59
- @jsonsmth made their first contribution in #62
- @robotdjman made their first contribution in #26
Full Changelog: https://github.com/nsarrazin/serge/commits/0.0.1
Contributors
thomasleveil, mavaa, and 5 other contributors