Feature: Trajectory reporting (whole.log, enhanced history, HTML report)

[kirahan]

Overview

When save-trajectory is enabled, each run now produces a richer audit bundle under ~/.mano/trajectory/, plus an interactive HTML report—without new config keys beyond the existing save-trajectory flag (default remains false).

What's New

1. Full terminal capture — whole.log

• Pure stdout/stderr tee for the entire task lifecycle
• Early buffer captures output before the trajectory folder exists (e.g. Session created:)
• Flushed into whole.log when the session directory is created; closed after the task summary is printed

2. Enhanced structured history — history.jsonl

Each line now includes:

Field	Purpose
phase	init / action / terminal
action_kind	Report styling (open_app, key, done, …)
screenshot	Relative path when a step PNG exists
status	Agent terminal state when applicable (e.g. DONE)

Also records step 0 (Initializing) and DONE/FAIL steps; supports multiple lines per step to mirror the terminal timeline.

3. Visual report — report.html

• Generated from session.json + history.jsonl + screenshots only (does not parse whole.log)
• English UI: stats header, step groups, collapsible screenshots, action-type color badges
• Expand/collapse all screenshots; works offline via file://
• Footer link to whole.log when present

4. End-of-run convenience

Task summary prints:

Report: <absolute path to report.html>
Preview: file:///.../report.html

Cloud runs optionally store cloud_session_id in session.json alongside the local trajectory session_id.

5. Documentation

README Configuration section documents save-trajectory, artifacts, and mano-cua config usage.

Session folder layout

~/.mano/trajectory/sess-<timestamp>-<id>/
├── session.json
├── history.jsonl
├── screenshots/
├── whole.log      ← new
└── report.html    ← new

Implementation

Area	Change
visual/trajectory/	New module: log_tee, history, action_styles, report_generator, unit tests
visual/vla.py	Early tee on run; cloud_session_id in session meta
visual/model/task_model.py	Tee install, history append, report generation, summary URIs
visual/view_model/task_view_model.py	Pass cloud_session_id
visual/config/user_config.py	Help text; default save-trajectory = false
README.md	Configuration docs

Enable

mano-cua config --set save-trajectory true
mano-cua run "your task"
# Open Report / Preview paths from the terminal summary

Files Changed

• README.md
• visual/config/user_config.py
• visual/model/task_model.py
• visual/model/tests/__init__.py
• visual/view_model/task_view_model.py
• visual/vla.py
• visual/trajectory/ (module + tests)

Testing

• PYTHONPATH=. python -m unittest discover -s visual/trajectory/tests -v
• Enable save-trajectory, run a task, verify all five artifact types exist
• whole.log includes early logs, step output, summary, and report paths
• report.html opens in browser with correct steps and screenshots
• With save-trajectory false, no trajectory folder or tee (unchanged behavior)

---

Suggested PR title: Feature: trajectory reporting with whole.log and HTML report