GooDAnDReaDY/dsh-tts

README 摘要

dsh-tts Text-to-speech for the DeepSeek Harness Web GUI. When Speak agent replies is on, each finished assistant turn is synthesized on the host and played in the browser. API keys never reach the browser. A later messenger hub can call the same host chain ( POST /dsh-tts/speak or the speak text tool). This package does not talk to Telegram itself. Install Restart the Web UI, then hard-refresh the browser. Configure Settings - Speech : - Speak agent replies -- off by default. - Provider chain -- pick a provider, paste its API key in that row, then Save (or leave the key field). The value is written to the host credentials store immediately and is never sent back to the browser. The row shows Configured / Not set . Leave the field blank to keep an existing key. Local providers ( edge , piper , espeak ) have no key field. - A provider without a credential is skipped, not fatal. - Piper / eSpeak / edge-tts binaries and the Piper model path. Keys land in the same store the rest of DSH uses (credentials file, then the process environment). The plugin stores names, never values: Put additional keys in the same pool ( API KEY 2 , ...) if you use a key-rotation plugin. Providers Key Servic…

分类