Step-by-step physical assembly guide for transplanting an M5Stack CoreS3 into the official StackChan Body. Detaching DIN Base, 4-pin wiring, and initial boot test.
Tested with 16 memories and 10 questions: a date bonus on cosine got the 9/9 memory and a 0.441 café-au-lait memory over the 0.45 cutoff, with 0.016 headroom left on hit-free questions.
Ryzen 7 5800HS test before StackChan: subject-less memories got claimed by the character, a 0.45 cosine threshold flipped on a comma, and ModelScope's embedding API vanished mid-test.
Qwen3-Embedding-0.6B on CPU vs ModelScope's API on a Ryzen 7 5800HS voice server. Local embedding adds up to +2.6s per voice turn by competing for CPU; the API adds none.
Tested Qwen3-Embedding-0.6B on CPU and Qdrant's local mode, then checked the 1.5-1.7GB RAM footprint against a Ryzen 7 5800HS voice server with only 4GB VRAM.
A pre-implementation design for adding Qwen3-Embedding and Qdrant memory to an RTX 3050 Ti and CoreS3 voice-chat stack without depending on StackChanWorld's API. It keeps raw logs, searchable memory, persona, and body separate.
CoreS3 voice chat rebuilt on action=async: filler on job_id, 0.7s polls, gapless sentence chunks, 8.3s to first reply, plus a Key unit (U144) talk button with port auto-detect.
Tested on a CoreS3: talk into the mic, get a spoken reply. 44-byte RIFF header on the PCM buffer, HTTPClient multipart POST built in PSRAM, 12.4–13.8s sync round trips.
Tap-to-record 5s at 48kHz into PSRAM, instant playback on CoreS3. The mic/speaker swap on the shared I2S bus worked with 200ms waits; the freeze was the serial monitor's DTR/RTS reset.
3 face PNGs embedded in flash decode to sprites in 87ms. CoreS3 touch stops at y=239 so Core2-style BtnA/B/C never fire; tap, hold and flick gestures instead.
ESP32 can't join Tailscale, so a VPS PHP relay feeds the CoreS3. Chunked WAV via writeToStream, SD mounts broken by any pre-mount LCD draw, NVS volume.
CoreS3 voice chat spec check before firmware: playWav needs linear PCM 8-16bit up to 2ch (any rate), mic/speaker share one I2S bus, 128x128 sprite lip sync fits in RAM.