Replace X clipboard with LLM-ready cleaned up context

webclip.sh

#!/bin/bash

# Exit unless xclip and html2markdown are installed, using the return code of 'which'
if ! which xclip >/dev/null 2>&1; then
    echo "xclip is not installed" >&2
    exit 2
fi
if ! which html2markdown >/dev/null 2>&1; then
    echo "html2markdown is not installed" >&2
    exit 2
fi

function warn() {
    if [[ -t 1 ]]; then
        echo "$@" >&2
    else
        gxmessage -buttons OK -default OK "$@"
    fi
}

# Get next sequential number from counter file
COUNTER_FILE="/tmp/webclip-counter"
LOCK_FILE="${COUNTER_FILE}.lock"
touch "$LOCK_FILE"

exec 9>"$LOCK_FILE"
if ! flock --timeout 2 9; then
    warn "Could not acquire lock within 2 seconds"
    exec 9>&-
    exit 3
fi
if [[ ! -f "$COUNTER_FILE" ]]; then
    echo "1" > "$COUNTER_FILE"
fi
SERIAL=$(cat "$COUNTER_FILE")
echo $((SERIAL + 1)) > "$COUNTER_FILE"
flock -u 9
exec 9>&-
TEMP="/tmp/webclip-$(printf "%03d" "$SERIAL")"

# Get rich text clipboard and URL
declare -a HTML_TARGETS=("text/html" "text/html;*" "html" "HTML")
declare -a TEXT_TARGETS=("text/plain" "text/plain;*" "UTF8_STRING" "STRING" "TEXT")
declare -a URL_TARGETS=("chromium/x-source-url" "text/x-moz-url")

# Find best available HTML target
HTML_TARGET=""
TEXT_TARGET=""
while IFS= read -r target; do
    for supported in "${HTML_TARGETS[@]}"; do
        if [[ "$target" == "$supported" ]]; then
            HTML_TARGET="$target"
            break 2
        fi
    done
done < <(xclip -selection clipboard -o -t TARGETS)
if [[ -z "$HTML_TARGET" ]]; then
    while IFS= read -r target; do
        for supported in "${TEXT_TARGETS[@]}"; do
            if [[ "$target" == "$supported" ]]; then
                TEXT_TARGET="$target"
                break 2
            fi
        done
    done < <(xclip -selection clipboard -o -t TARGETS)
    if [[ -z "$TEXT_TARGET" ]]; then
        warn "Could not find a supported clipboard format"
        exit 1
    fi
fi

# Find best available URL target
URL_TARGET=""
while IFS= read -r target; do
    for supported in "${URL_TARGETS[@]}"; do
        if [[ "$target" == "$supported" ]]; then
            URL_TARGET="$target"
            break 2
        fi
    done
done < <(xclip -selection clipboard -o -t TARGETS)

xclip -selection clipboard -o -t "${HTML_TARGET}${TEXT_TARGET}" >"${TEMP}.html"
URL=""
if [[ -n "$URL_TARGET" ]]; then
    URL=$(xclip -selection clipboard -o -t "$URL_TARGET")
fi

# 2. Sanitize
echo "" >"${TEMP}.md"
echo "<reference>" >>"${TEMP}.md"
echo "<number>${SERIAL}</number>" >>"${TEMP}.md"
if [[ -n "$URL" ]]; then
    echo "<url>${URL}</url>" >>"${TEMP}.md"
fi
echo "<content>" >>"${TEMP}.md"
if [[ -n "$HTML_TARGET" ]]; then
    html2markdown --images-to-alt --escape-all --no-automatic-links --mark-code <"${TEMP}.html" >>"${TEMP}.md"
else
    cat "${TEMP}.html" >>"${TEMP}.md"
fi
echo "</content>" >>"${TEMP}.md"
echo "</reference>" >>"${TEMP}.md"
echo "" >>"${TEMP}.md"

# 3. Replace into both clipboards
# (We NEVER know which one the user intends to use…)
# Note that unlike Chromium, xclip cannot set multiple targets, so we
# choose "STRING" since "TEXT" rarely worked.
xclip -selection primary -i -t STRING <"${TEMP}.md"
xclip -selection clipboard -i -t STRING <"${TEMP}.md"

MD_LINES=$(wc -l <"${TEMP}.md")
MD_WORDS=$(wc -w <"${TEMP}.md")
if [[ -n "$HTML_TARGET" ]]; then
    warn "HTML Clipboard is now LLM-friendly Markdown: ${MD_WORDS} words, ${MD_LINES} lines"
else
    warn "TEXT Clipboard is now LLM-friendly Markdown: ${MD_WORDS} words, ${MD_LINES} lines"
fi

# Willingly leaving the source and sanitized files in place for human
# inspection.