closes#2931 (slightly changed goal based on advice from the discord,
generate full messages, not 'x' tokens. Full messages are more useful
data)
- [x] Create draft selection UI
- [x] Draft inference
- [x] Option to regenerate drafts and serve 3 new ones
- [x] Remember last viewed sibling message
- [x] Store selected draft training data for RLHF
- ~~[ ] Disable drafts when queue is too long / server is under load~~
(Suggested to leave to next PR in the discord)
- [x] Draft markdown rendering
- [x] 'Used plugin' UI for drafts
- [x] Resolve merge conflicts
---------
Co-authored-by: notmd <33456881+notmd@users.noreply.github.com>
Co-authored-by: notmd <notmd1811@gmail.com>
Added dependabot file to implement weekly dependency version checks for
Node JS
changes in this PR:
adds a new file under `.github/` called `depndabot.yml`.
---------
Co-authored-by: notmd <33456881+notmd@users.noreply.github.com>
Fixes#2285
Changes in the PR:
* revert auth callbacks to refer back to the inference server (same as
it was before), the callback urls in the corresponding apps must be
updated.
* Added trusted clients which can communicate with API keys, these keys
need to be added to the github env / secrets, I don't have the access
rights to do that
* Drop all auth code from website
Changes here conflict with #2221, although not significantly, only the
urls, maybe this could be addressed as part of #2286
fixes#2081
Test Evidence:
Since this is a bit tricky to fix and test on this repo
I tested by building postgres images on my private repo actions.
Co-authored-by: James Melvin <melvin@gameface.ai>
To allow convenient imports/processing of OA data by 3rd party projects
the oasst-jsonl data file format schemas are moved into a separate new
python module called oasst_data. Code related to reading oasst-jsonl
files will be added later.
- add lang column to message_tree_state, create multi-column index on
`state`, `lang`
- stop generating initial prompt tasks for languages with more than
`max_prompt_lottery_waiting ` prompts in `prompt_lottery_waiting` state
- add `MAX_PROMPT_LOTTERY_WAITING` workflow variable
* Read the signin env variables from server props so that they always use the runtime variables
* Rename the env variables to be private
* Updating docker and ansible changes to reflect env variable name change
* Disabling captcha during CI due to failures and cleaning up a build step
* Add our current architecture to the blog
* Update docs/blog/2023-02-11-architecture.md
Co-authored-by: Andrew Maguire <andrewm4894@gmail.com>
---------
Co-authored-by: Andrew Maguire <andrewm4894@gmail.com>
