xiaozhi esphome animated patched yaml on github

1. Display crash during long responses (cache access error / IllegalInstruction) In the `replying_page` and `request_page` display lambdas, `it.printf(..., TextAlign::CENTER, ..., text.c_str())` renders text into a fixed 120–200 px box without limiting the text length. Short responses (e.g., "22°C, clear") fit fine, but longer strings—particularly HA error messages like "Something went wrong: Action X not found" when the assistant fails to recognize a command—overflow the box and render outside the framebuffer boundaries. Our RGB LCD backend (`rpi_dpi_rgb`) does not check for this, resulting in a write to unmapped memory and a cache access fault, which causes core 1 to crash mid-render. Fix: truncate the text before calling `printf`, e.g.: cppstd::string safe_text = id(text_response).state; if (safe_text.length() > 24) { size_t cut = 21; while (cut > 0 && (safe_text[cut] & 0xC0) == 0x80) cut--; // neříznout uprostřed UTF-8 znaku safe_text

2. Invalid UTF-8 sent to Home Assistant, causing an API disconnection loop Separately, the `text_sensor` `text_request`/`text_response` components feature an `on_value` lambda that truncates the state to 32 characters using `esphome::str_truncate()`. This helper function (`alloc_helpers.cpp`) performs a raw byte-based substring operation and does not account for UTF-8 encoding. For any text containing diacritics or non-English characters, truncating at a fixed byte offset can split a multi-byte UTF-8 sequence in half. The resulting invalid string is then transmitted via the API; the Home Assistant protobuf parser rejects it (reporting "String field had bad UTF-8"), causing the connection to enter a disconnect/reconnect loop. Fix: Apply the same UTF-8 character-boundary-aware truncation used above within the `on_value` lambda before assigning the value to `.state`. Zpětná vazba Překladač Google