smts:export

Export the current Claude Code session transcript to a single self-contained HTML file with a solarized-light theme, then open it in the browser.

CLI Options

Process

1. Parse arguments: Extract any flags from the user's input:

   • Look for -o or --output followed by a path
   • Look for --split flag
   • Look for --page-size followed by a number
   • Check for pick keyword

2. Determine the session file:

   If pick was specified:

   • Run the session lister script scoped to the current project, showing the last 5 sessions:

     python3 "${CLAUDE_PLUGIN_ROOT}/scripts/list-sessions.py" --project "${SMTS_PROJECT_DIR:-$(pwd)}" --limit 5
     

   • Each output line is tab-separated: date \t first-message \t session-id(short) \t /full/path
   • Use AskUserQuestion to present each session as [date] session-id(short): first message…
   • From the chosen entry:
     • Take the last tab field as SESSION_FILE (full path)
     • Extract the full session ID from the filename: SESSION_ID=$(basename "$SESSION_FILE" .jsonl)
     • Extract the project directory from the path (parent directory of the session file)

   Otherwise (default: auto-detect current session):

   • Locate the session file:

     # Convert cwd to the project directory slug
     # e.g., /Users/user/code/myproject -> -Users-user-code-myproject
     PROJECT_DIR=$(echo "${SMTS_PROJECT_DIR:-$(pwd)}" | sed 's|^/||; s|/|-|g' | sed 's|^|-|')
     
     # Try the session ID captured at session start (set by SessionStart hook)
     SESSION_ID="${SMTS_SESSION_ID:-}"
     if [ -n "$SESSION_ID" ]; then
       SESSION_FILE=$(ls ~/.claude/projects/${PROJECT_DIR}/${SESSION_ID}.jsonl 2>/dev/null | head -1)
     fi
     
     # Fallback: most recently modified non-agent JSONL in the project directory
     if [ -z "${SESSION_FILE:-}" ]; then
       SESSION_FILE=$(ls -t ~/.claude/projects/${PROJECT_DIR}/*.jsonl 2>/dev/null | grep -v '/agent-' | head -1)
     fi
     

3. Determine output path:

   If -o was specified:

   • Use the provided path

   Otherwise:

   • Default output directory: docs/sessions/ (create if needed)
   • Default filename: <full-session-id>.html (or <full-session-id>-pages/ for split mode)
   • The full session ID is the complete UUID (e.g., 12c5520e-5003-423a-b494-15eb204b4572)
   • Ask user: "Export session to docs/sessions/<full-session-id>.html? [Y/n]"

4. Check message count and suggest split for large sessions:

   Count messages in the session:

   wc -l "$SESSION_FILE"
   

   If message count > 100 and --split was NOT specified:

   • Ask user: "This session has X messages. Split into multiple pages for easier viewing? [y/N]"
   • If yes, enable split mode

5. Generate the HTML:

   For single HTML:

   python3 "${CLAUDE_PLUGIN_ROOT}/scripts/generate-html.py" "$SESSION_FILE" -o "$OUTPUT_PATH"
   

   For split HTML:

   python3 "${CLAUDE_PLUGIN_ROOT}/scripts/generate-html.py" "$SESSION_FILE" -o "$OUTPUT_DIR" --split --page-size "$PAGE_SIZE"
   

6. Open in browser:

   open "$OUTPUT_PATH"  # or open "$OUTPUT_DIR/index.html" for split mode
   

7. Report to user: Tell them the output path and that it's been opened in their browser.

Important Notes

• The generated HTML is fully self-contained (inline CSS + JS, no external dependencies)
• Uses the Solarized Light color scheme for comfortable reading
• All content blocks are rendered: user messages, assistant text, thinking (collapsible), tool calls, tool results
• Long content sections are truncatable with "Show more" buttons
• Split mode creates keyboard navigation (arrow keys) between pages
• Timestamps are converted to local time via JavaScript