dannydannydanny/dotfiles
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

phantom-ship/shipyard: add media-processing tools for feedback

Browse source 
Feedback now accepts photos, voice notes, video, documents etc. Phase
A captures + stores raw files (Pillow for EXIF strip); Phase B derives
OCR text, speech transcripts, poster frames, PDF text — all via
subprocess so each tool degrades gracefully if absent. Wire the
following into the shipyard service:

  - python3Packages.pillow → EXIF strip on captured photos
  - ffmpeg                 → poster frames + audio→16kHz WAV for whisper
  - tesseract (eng + rus)  → OCR (vyscul writes in Russian)
  - whisper-cpp            → speech-to-text for voice / audio / video
  - poppler_utils          → pdftotext for document attachments

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
DannyDannyDanny 2026-06-07 13:16:38 +02:00
parent 680c20483c
commit cc8cc05a08
5 changed files with 204 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

23
nixos/hosts/phantom-ship.nix
View file

@ -241,18 +241,37 @@ in
# Code deployed out-of-band via rsync to /home/danny/shipyard/
# (staying in-tree in ~/python-projects/26_shipyard/ until spun out to its own repo).
# Bot token (not in repo): ~danny/.secrets/telegram-bot-token-shipyard
# Data (feedback.jsonl, pointer cache): ~danny/.local/share/shipyard/
# Data (feedback.jsonl, feedback.db, pointer cache, feedback_media/):
# ~danny/.local/share/shipyard/
#
# Feedback now accepts photos / voice / video / docs / stickers etc.
# Phase A captures + stores raw files; Phase B derives OCR text
# (tesseract), speech transcripts (whisper-cpp), poster frames
# (ffmpeg) and PDF text (pdftotext) — all via subprocess, so each
# tool degrades gracefully if missing.
systemd.services.shipyard = let
pythonEnv = pkgs.python3.withPackages (ps: with ps; [
python-telegram-bot
httpx
pillow # EXIF strip on captured photos
]);
# tesseract with English + Russian tessdata — vyscul writes in
# Russian, screenshots can land in either language.
tesseractWithLangs = pkgs.tesseract.override {
enableLanguages = [ "eng" "rus" ];
};
in {
description = "Shipyard Telegram bot (mini-app launcher + feedback)";
after = [ "network-online.target" ];
wants = [ "network-online.target" ];
wantedBy = [ "multi-user.target" ];
path = [ pythonEnv ];
path = [
pythonEnv
pkgs.ffmpeg # video/animation posters, sticker decode
tesseractWithLangs # photo OCR
pkgs.whisper-cpp # voice/audio transcription
pkgs.poppler_utils # pdftotext (document handling)
];
environment = {
SHIPYARD_BOT_TOKEN_FILE = "/home/danny/.secrets/telegram-bot-token-shipyard";
# Owner-only commands (/admin, /grant, /revoke) — anyone else gets ignored.