All notable changes to this project are documented in this file.
- Text-to-speech playback using Piper TTS, synthesized per sentence as responses stream.
- Audio output saved to
src/data/audioand played automatically through system speakers. - Persistent conversation history stored in
src/data/history.json, including per-message metadata (chunk count, token count, elapsed time, temperature, max tokens). .gitignoreentries for generated audio and history files.
StreamBuffernow flushes on sentence boundaries instead of per-token, improving TTS chunk quality./historycommand displays full message content instead of a truncated preview.
- Command handler in
main.pyno longer exits the chat loop after every command; only/exitterminates the session. /historycrash caused byNonemessage content./historycrash caused by_print_separatorbeing used in a string concatenation instead of being called directly.- Verbose logging now respects a global toggle instead of always printing request and chunk details.
- Basic streaming chat loop using Replicate.
- JSON event logging for chat requests, successes, and errors.