deepline sessions

Usage: deepline sessions [options] [command]

Upload and render local agent session transcripts.

Options:
  -h, --help        display help for command

Commands:
  send [options]    Upload session transcript(s) or a local file to Deepline.
  render [options]  Render local session transcript(s) to an HTML viewer.
  help [command]    display help for command

Usage: deepline sessions send [options]

Upload session transcript(s) or a local file to Deepline.

Options:
  --session-id <uuid>           Claude or Codex session UUID. Repeat for
                                multiple sessions. (default: [])
  --label <label>               Label for the preceding session id (default: [])
  --agent <auto|claude|codex>   Limit transcript discovery to one agent runtime
                                (default: "auto")
  --current-session             Use the calling agent's session, else the newest
                                in this directory
  --file <path>                 Upload a raw local file instead of a session
  --rating <good|poor|neutral>  Tag the upload by outcome, not tone, so Deepline
                                can filter sessions: good = the user's goal was
                                met without product problems; poor = the goal
                                was missed, or a product failure, bug, or
                                avoidable friction (including a substantial
                                workaround) got in the way; neutral = partly met
                                or unclear.
  --json                        Emit JSON output
  -h, --help                    display help for command

Explanation:
  Purpose:
    Upload an agent session transcript, or a file, to Deepline so the team can
    see how a task went.
  When:
    Only after the user agrees. Ask once, at the end of the task and in its own
    message. After a user-requested feedback report, offer it alongside. When
    the upload succeeds, tell the user: "Session sent to Deepline. Rating:
    <rating>".
  Input:
    --current-session (the calling agent's session, else the newest one started
    in this directory), or --session-id <uuid> (repeatable, each with an
    optional --label), or --file <path> for a raw file. --agent
    auto|claude|codex limits where transcripts are looked for.
    --rating good|poor|neutral, judged by outcome and not tone: good when the
    user's goal was met without product problems; poor when it was missed, or
    a product failure, bug, or avoidable friction (including a substantial
    workaround) got in the way; neutral when partly met or unclear. A session
    that produced a feedback report is usually poor or neutral. --json.
  Output (JSON):
    A confirmation envelope with uploaded and session_ids; the text form names
    the session that was uploaded.
  Cost:
    Free.
  Avoid:
    The command is sessions, plural; the singular session form is the old one.
    In a Cowork sandbox the transcripts are mounted under
    ~/mnt/.claude/projects: when no session files are found and that directory
    exists, link it to ~/.claude/projects and retry, or pass the newest
    transcript there with --file.

Usage: deepline sessions render [options]

Render local session transcript(s) to an HTML viewer.

Options:
  --session-id <uuid>          Claude or Codex session UUID. Repeat for multiple
                               sessions. (default: [])
  --label <label>              Label for the preceding session id (default: [])
  --agent <auto|claude|codex>  Limit transcript discovery to one agent runtime
                               (default: "auto")
  --current-session            Use the calling agent's session, else the newest
                               in this directory
  --auto-refresh <seconds>     Add a browser auto-refresh interval
  -o, --output <path>          Output HTML path
  --json                       Emit JSON output
  -h, --help                   display help for command