GPT-4 Session to Markdown

post

#!/usr/bin/env bash
#
# Copyright (c) 2023 Jegors Čemisovs
# License: MIT
# Repository: https://github.com/rabestro/gpt4-session-to-markdown
#
# Converts GPT-4 chat session JSON data to markdown format and
# adds YAML front matter to the top of the file to prepare it
# for publishing on the Jekyll site.
#
# Usage:
#   ./post <file.json>
#

# set these variables as needed for your environment
readonly POSTS_DIR="../_posts"
readonly TIMEZONE="+0300"
readonly CATEGORIES="AI"
readonly TAGS="ai"

create_file_name() {
  jq --raw-output '.history[0].name | ascii_downcase | gsub("[^a-z0-9]+"; "-")' "$1"
}

conversation() {
  jq --raw-output '.history[0].messages | map((
    if .role == "user" then "👤" else "🤖" end
    )+ " \(.content)\n") | join("\n")' "$1"
}

has_mermaid() {
  jq 'any(.history[].messages[]; .content | contains("```mermaid"))' "$1"
}

main() {
  local -r file_json="$1"
  local -r file_name=$(create_file_name "$file_json")
  local -r current_date=$(date '+%Y-%m-%d')
  local -r current_time=$(date '+%H:%M:%S')
  local -r file_post="${POSTS_DIR}/$current_date-$file_name.md"

  local -r title=$(jq --raw-output '.history[0].name' "$file_json")
  local -r prompt=$(jq --raw-output '.history[0].prompt' "$file_json")

  echo "Title: $title"
  echo "File: $file_name"

  cat <<EOF >"$file_post"
---
title: "$title"
date: $current_date $current_time $TIMEZONE
categories: [$CATEGORIES]
tags: [$TAGS]
mermaid: $(has_mermaid "$file_json")
---
{% raw %}

## The system prompt used

$prompt

## The AI Conversation Log

$(conversation "$file_json")

{% endraw %}
EOF
}

main "$1"