26 June 2023
- Add this changelog :)
- Add sha256 hashes on release so you can verify the binaries
- All binaries are automatically generated with Github actions
- Add signal handling for SIGHUP (macOS, Linux) and CTRL_CLOSE_EVENT (Windows) to fix issue
#16
- This allows you to run chat as a subprocess. The chat subprocess now quits properly if parent app is closed.
- Version information
- Fix segfault on
/help
22 June 2023
- Pull request from @154pinkchairs merged. Thanks. :)
- The pull request
#18
has the two fixes below: - Properly handle file paths including tildes
18e9f36
- Handle buffer allocation errors
6800dfb
- Better debug mode compilation. May fix issue
#9
16 June 2023
- Adds
--save_dir
option so you can change save directory location - Default location is
./saves
on the same directory as the chat binary - See issue
#13
for more details
15 June 2023
- Fixes for old macOS.
- Use
-DOLD_MACOS=ON
option when compiling with CMake. - Tested to compile on High Sierra and Xcode 10
14 June 2023
- You can name saves with
./save NAME
and./load NAME
- You can toggle saving and loading off with
--no-saves
flag
13 June 2023
- Save/load state with
./save
and./load
- Reset context with
./reset
, help with./help
- Makes a
./saves
folder - Note that a single save can take up to 2Gb
- You can wrap the AI response with tokens using
--b_token
and--e_token
- See issue
#12
for more details
5 June 2023
- Fix when using json to specify names for logfiles. Fixes issue
#11
4 June 2023
- Fix said ability to reset context... :)
3 June 2023
- Ability to reset context
30 May 2023
- Save and load chat logs
- Use
--save_log
and--load_log
- AVX512 option for compilation
-DAVX512=ON
17 May 2023
- Update gpt4all backend to v0.1.1
61a963a
- Full Windows Visual Studio compatibility. Finally fixes issue
#1
- Builds from source on aarch64 Linux. Fixes issue
#3
- Full MPT support. Fixes issue
#4
16 May 2023
- Code cleaning and reordering
llmodel_create_model()
function
13 May 2023
- Add support for MPT models
- Uses gpt4all-backend
12 May 2023
- First pull request from @itz-coffee merged. Thanks. :)
- The pull request
#2
adds the feature below: - Add --no-animation flag
fdc2ac3
- Support for old macOS
4 May 2023
- Parse parameters from json files
- Use
-j FNAME
or--load_json FNAME
3 May 2023
- MinGW compilation on Windows
1 May 2023
- v0.1.4 had no tags
- It was part of
cmake-release.yml
rewrite to enable MinGWe7e1ebf
1 May 2023
- Add loading of prompt template files
- Use
--load_template
for loading - See
prompt_template_sample.txt
for a sample
30 April 2023
- Automatic memory handling for the model
29 April 2023
- Windows compilation fixes
29 April 2023
- Before this, progress was in GPTJ-chat and Llama-chat
- First version