Sessions

Each directory you launch kdeps from has its own conversation history. Nothing is written into the project directory. The other REPL behaviors stay on REPL features.

Where sessions are stored

Conversations and the agent's memory are stored in ~/.kdeps/, partitioned by the directory you launch kdeps from - ~/.kdeps/sessions/ and ~/.kdeps/memory/ each hold a subfolder per working directory. cd into a project and kdeps sees that folder's history; a different folder starts clean. Nothing is written into the project directory itself.

The resume picker

When you start kdeps in a folder that has saved sessions - and you did not pass --resume or --new - a picker lists them:

kdeps                     # in a folder with history -> resume picker
kdeps --new               # skip the picker, start a clean session
kdeps --resume <id>       # resume a specific session directly (no picker)

Picker rules

• The picker only appears in an interactive terminal with at least one saved session for this folder and no --resume / --new. Piped input starts clean.
• Resuming, continuing, then exiting updates the same session - it does not fork a new one. The session is also saved after every turn, so a crash or kill still leaves it in the picker.
• Settings (/refine, /theme, the default model, tool tuning), the model cache, and everything else under ~/.kdeps/ are unchanged.
• ctrl+d on a highlighted session deletes it in place (the "Start a new session" row cannot be deleted) - same as /session delete <id>, just reachable from the picker itself.
• Multiple kdeps instances can point at the same folder at once, each with its own picker and its own chosen (or new) session - the underlying session file is opened only for the moment of each read/write, not held open for an instance's whole run, so instances do not lock each other out.

Session commands

/session list                  # this folder's saved sessions
/session save [name]           # save a named snapshot of the current session
/session load <id>             # restore a saved session
/session delete <id>           # delete a saved session
/session checkpoint            # print the current entry ID (for /session goto)
/session goto <entry-id>       # restore session to the turn at that entry ID
/session branches              # list stashed (pruned) turns from prior /session goto calls
/session import <path>         # load a JSONL session file exported from another run

/session goto is non-destructive: the pruned tail is stashed. Use /session branches to see stashed entry IDs, then /session goto <id> again to navigate back.