Problem Statement
If a user exits the terminal or the process is interrupted, the entire state of a multi-turn session is lost. There is currently no way to "pause" a long-running debate and resume it later with the exact same agent states, history, and configuration.
Proposed Solution
- Implement
to_dict() and from_dict() methods for the Session class in session.py.
- This should capture the full SessionConfig, the state of all Agent instances, the entire conversation
history, and metadata like turn_count.
- Add a CLI command or a post-session option to "Save Snapshot" which writes this state to a JSON file.
- Update the CLI wizard to allow "Resuming from Snapshot" by loading one of these JSON files.
Alternatives Considered
Database-backed persistence, but JSON serialization matches our current local-first, file-based storage philosophy.
Priority
Low 🟢
Additional Context
Crucial for complex projects, multi-day simulations, or research tasks that span more than one sitting.
Problem Statement
If a user exits the terminal or the process is interrupted, the entire state of a multi-turn session is lost. There is currently no way to "pause" a long-running debate and resume it later with the exact same agent states, history, and configuration.
Proposed Solution
to_dict()andfrom_dict()methods for the Session class in session.py.history, and metadata liketurn_count.Alternatives Considered
Database-backed persistence, but JSON serialization matches our current local-first, file-based storage philosophy.
Priority
Low 🟢
Additional Context
Crucial for complex projects, multi-day simulations, or research tasks that span more than one sitting.