Aster Chat Sessions

Chat, sessions, and memory

aster chat talks to the review agent with read/search tools over the repo. In a terminal it opens a TUI; from scripts and agents always use the one-shot forms.

One-shot answers

aster chat -p "why is finding 2 critical?"       # plain text (default when piped)
aster chat --json "summarize the last review"    # {"reply", "edits", "usage"}
aster chat --no-tools -p "explain this error"    # plain LLM turn, no repo tools

--model overrides the model for the turn (else ASTER_MODEL, else aster.yaml).

History and persistence

aster chat --continue -p "and the second one?"   # seed the most recent session's history
aster chat --session <ID> -p "..."               # persist this turn into a session by id
aster chat --messages-json msgs.json -p "..."    # caller-owned history (or `-` for stdin)

--messages-json takes a JSON array of {"role","content"} with roles user | assistant | system. Combined with --session, the session only records the turn (the caller owns history); --session alone also seeds the session's prior history.

Edits

--allow-edits exposes an edit_file tool to the agent, gated by the permissions block in aster.yaml (deny-first globs, protected paths). Leave it off unless the task is explicitly to change files, and diff the working tree afterward.

Inspecting sessions

aster sessions                 # list saved sessions for this repo
aster sessions show <id>       # full transcript
aster sessions --json          # machine-readable (also on `show`)

Durable memory

aster memory                        # list stored memory
aster memory add "we use pnpm"      # append a fact to project memory
aster memory add --title deploy "…" # save a titled block
aster memory --json                 # machine-readable

Memory is per-repo and fed to future chats and reviews; store durable facts (conventions, constraints), not one-off context.