v0.6

New feature: Added llama.rn backend to run models natively.

• Simply download a model to your phone and Import it. (Note that it creates a duplicate of the model file within the app, so it may be wise to remove the downloaded model to save on space.)
• Load the model to allocate it into memory and start chatting.
• Prompt processing is very slow, expect it to take a few minutes on a new chat.
• llama.rn does not have GPU support as of yet, so prompt processing will be very slow.

v0.6.1

• Initial implementation of Open Router and Chat Completions, currently does not respect Instruct formatting.
• Added System TTS

v0.6.2

• Implemented alternate greetings.
• Fixed issues with macro replacement.

v0.6.2a

• Fixed macro replacement still not working.

v0.6.3

• Optimized context building.
• Added elipses when waiting for response.
• Improved startup routine.

v0.6.3a

• Fixed issues with local inferencing.

v0.6.4

• Added dynatemp for KAI endpoint.
• Allow newlines in Instruct presets.

v0.6.4a

• Fixed a crash when entering CharInfo menu.

v0.6.5

• Bumped llama.rn version to 0.3.0-rc.14
• Added Seed value to local inferencing.