AI Webcam to html

webcam2html.sh

#!/bin/bash

# Takes a photo and converts to a project text artifact.
# Just experimental for now.

set -euo pipefail
THISDIR="$(dirname "$0")"
#shellcheck disable=SC1091
source "$THISDIR/.env"

DEVICE=/dev/video1
SIZE=512x512

checks() {
  die() {
    echo "$*" >&2
    exit 1
  }
  requires() {
    command -v "$1" &>/dev/null || die "You need to install $1"
  }

  requires fswebcam
  requires base64
  requires jq
  requires jo
  requires curl
  [[ -n "$OPENAI_API_KEY" ]] || die "OPENAI_API_KEY not set"
  shellcheck -x "$0"

  # Android requirments
  # requires ffmpeg
  # ssh phone 'command -v termux-camera-photo' &>/dev/null || die "Cannot use phone"
}

main() {
  checks

  #TODO: prompt hardcoded for now.
  # prompt="Generate a Graphviz dot file based on the image."
  # prompt="Generate a PlantUML ERD file based on the image. And generate SQL DDL to build it."
  prompt="Generate an html form based on the image.  Style with Bulma."
  prompt="${prompt}\nOnly generate the raw text.  Do not produce commentary or surrounding markdown."

  photo="$(take_photo)"

  generate_data "$prompt" "$photo" | sendchat | sed '/```/d'
}

#usage: tmpfile --suffix=.jpg
tmpfile() {
  file="$(mktemp --dry-run "$@")"
  #shellcheck disable=SC2064
  trap "rm -f '$file'" EXIT
  echo -n "$file"
}

take_photo() {
  file="$(tmpfile --suffix=.jpg)"
  fswebcam -q -r "$SIZE" --jpeg 20 --no-banner -d "$DEVICE" "$file"
  local base64_image="$(base64 -w 0 "$file")"
  echo -n "data:image/jpeg;base64,${base64_image}"
}

#shellcheck disable=SC2034,SC2276
generate_data() {
  prompt="$1"; shift
  url="$1"; shift

  jo \
    model="gpt-4-vision-preview" \
    'messages[]'="$(jo \
      'role'="user" \
      'content[]'="$(jo type=text text="$prompt")" \
      'content[]'="$(jo type=image_url image_url="$(jo url="$url")")"
    )" \
    max_tokens=3000 \
    "$@"
}

sendchat() {
  # stdin/stdout will be request/response body.
  curl -sSf -X POST \
    "https://api.openai.com/v1/chat/completions" \
    -H "Authorization: Bearer $OPENAI_API_KEY" \
    -H "Content-Type: application/json" \
    -H "Accept: application/json" \
    --max-time 180 --retry 5 --retry-delay 3 \
    --json @- | \
    jq '[.choices[].message.content] | join("\n\n----\n\n")' -r
}

# Requires android, Termux, Termux::API, sshd, and "phone" host alias
take_android_photo() {
  file="ai.jpg"

  image="$(ssh phone 'echo $TMPDIR/ai.jpg')"
  #shellcheck disable=SC2029
  ssh phone "termux-camera-photo -c 0 '$image'"
  scp "phone:$image" "$file"

  ffmpeg -i ai.jpg -q:v 5 -vf scale=512:512 ai2.jpg
  rm ai.jpg
  mv ai2.jpg ai.jpg

  local base64_image="$(base64 -w 0 "$file")"
  echo -n "data:image/jpeg;base64,${base64_image}"

  #TODO: got lazy. need to use tmp files
}

main "$@"