v1.0

I think this plugin has sufficient basic features needed in a LLM framework, so it is the time to bump it to v1.0

• Breaking change: replace Should Output Bos and Should Output Eos by Should Output Special
• A LLM_LEGAL.md to document information about legal issues on using LLM in games
• macOS build: support is on best effort basis since I don't have a mac myself
• store_text_finished signal for LlmDB node
• create function and LlmDBMetaDataType enum for LlmDB node
• Add Main GPU and Split Mode to control which gpu (if you have multiple) should be used in computation, this is also needed as a workaround if you have multiple drivers install for a single GPU (my fix is still under review on the upstream llama.cpp repo)
• Bug fixes

v0.4

• LlmDB node: manages a vector database for retrieval augmented generation
• Bug fix: dead lock
• Bug fix: non-English encoding
• Vulkan build is back, and provide an alternative cpu build

v0.3

• Breaking change: rename generate_text(prompt: String) -> String to generate_text_simple(prompt: String) -> String, you may also add 2 empty string argument (generate_text(prompt, "", "")) to achieve a similar result.
• Unfortunately, embedding does not work with vulkan backend at the moment, now change to cpu build and let's wait for the upstream fix
• Add GDEmbedding node for sentence embedding computation
• Add GDLlava node for multimodal model
• Introduce multiple run_* methods for GDLlama, GDEmbedding, and GDLlava to run computation in background,

v0.2.2, v0.2.3

• Disable LLAMA_NATIVE to ensure that the binary is compatible on most PC.

v0.2, v0.2.1

• Add intruct and interactive mode, also implemented the relevant api
• Add Json schema api
• Statically linked msvc runtime, hopefully it fixes #2
• Expose more llama.cpp parameters

v0.2.1

• Fix linux build

First release

v0.1

docs: fix grammar