This repository was archived by the owner on Feb 8, 2024. It is now read-only.
Releases: npc-engine/npc-engine-unity
Releases · npc-engine/npc-engine-unity
Release v0.1.6
Release v0.1.5
- No processes lost/leaking when switching scenes or entering/leaving play mode
- NPC Engine process can be stopped
- Coroutine only interface
- num sampled parameter for chatbots that allows to improve results (more = better) in cost of performance
- Ability to set number of speaker ids in TTS config
- minor fixes
Please don't forget to redownload the models :)
Release v0.1.4
- Updated text generation model config to fix model prompt cropping. Please redownload default models or reexport any custom text generation models.
- Migrated importing custom models to separate pip package called
npc-engine-import-wizard. - NonPlayableCharacter no longer needs dialogue system to function (It will only generate replies)
- Fix to advanced demo setup. No Location and Character was set for PlayerCharacter
- Various small bug fixes
- Hotfix to the advanced demo
Release v0.1.3
- Option to access npc-engine through HTTP Server.
- Refactored models into multiprocessing services, with each service running in it's own process and communicating via ZMQ IPC sockets.
- Added NPC Engine Config component that shows available services, lets user to edit, remove them or download new services from the Huggingface Hub or import from the local filesystem.
- Added NPC Engine Manager component that allows users to start/stop NPC Engine server, services, shows services status.
- NPC Engine can now be used from the editor!
- Non Player Character has now custom editor that allows users to test all the related functionality in the editor (similarity, TTS, text generation)
- All the optional project parts that require dependencies are now moved into their own unitypackage files (Advanced demo and VIDE dialogues integration).
- Service APIs are now components that can be acquired via NPCEngineManager.GetAPI(), are non-blocking in regards to each-other and allow to specify service ID to use two different models of the same service type.
- Service resolution through method name, API name, service class or service id (folder holding the config).
- Device of the serivce can now be specified in the service config (via
providerskey)
e.g.
providers:
- 'gpu' or 'cpu'
- New CLI that can be installed via PIP (
pip install npc-engine[export]) and provide commands to manage and export models. - Versioned core docs.
- Renamed chatbot API to text generation API.
- Discord server!
- Patreon!