Review history

Every miucr review (local and --pr) auto-saves a full record to a local store: the findings + stats, the PR/repo context, the provider/model, a per-turn transcript of the reviewer’s tool calls, and the raw prompt and raw final response. That makes a review auditable and replayable after the fact: miucr history show <id> reconstructs exactly what the reviewer saw and said.

Auto-save is on by default. Nothing leaves your machine: the store is the same local SQLite DB at ~/.config/miu/cr/state.db (or your configured Postgres backend), and it is gitignored. No credentials are ever persisted: tokens never enter the prompt, the diff, or the record.

What is saved

Field	Meaning
id, created_at	Record identity + timestamp
target	owner/repo#N for a PR, else the local repo dir
mode	staged / range / commit (a --pr review is stored as range; it is identified by target, not a distinct pr mode)
provider, model	The LLM profile used
status, max_severity	Terminal status + worst finding severity
findings, stats	The full review result
transcript	Per-turn tool calls the reviewer made
raw_prompt, raw_response	The verbatim LLM I/O (audit trail)
trace	The full redacted trace (system prompt, diff meta, selected files, injected rules, prompts, response); view with miucr trace <id>

Opting out per run

miucr review --staged --no-save     # review, but persist nothing

The stdout envelope is unchanged either way; it gains an additive review_id field (the saved record id, or empty with --no-save).

Browsing history

miucr history                                   # recent reviews, newest first
miucr history --repo owner/repo                 # filter by repo (PR) or repo dir (local)
miucr history --pr owner/repo#7                 # filter to one PR
miucr history --since 7d                        # 7d / 24h / 2026-06-01
miucr history --limit 50                        # cap rows (default 20; 0 = no limit)
miucr history -o pretty                          # human table

miucr history show <id>                          # one full record (findings + stats + transcript + raw I/O)
miucr history show <id> -o pretty --raw          # pretty, with the raw prompt/response inline

history emits the history.list envelope kind; show emits history.record (404s with a typed history.not_found on an unknown id).

Inspecting the trace

Alongside the record, every review keeps a redacted trace: the ordered steps of the pipeline: the system prompt, the diff identification (base/head + how it was computed), the selected files, the injected rules (stem + provenance), the user prompt, the model/provider, the raw response, and the tool calls. View any past review’s trace:

miucr trace <id>                 # ordered steps (kind: trace.show)
miucr trace <id> -o pretty       # a readable per-step view

trace.show data is {id, steps:[{step, payload}]}. An unknown id returns a typed trace.not_found; an old review with no trace renders empty.

For a live trace, pass --trace to review: each capture seam streams one NDJSON line ({"step":...,"payload":...}) to stderr as the run proceeds, distinct from --verbose progress. The stdout result envelope is unchanged.

miucr review --staged --trace 2> trace.ndjson    # live steps on stderr; envelope on stdout

The trace holds the prompt (your own code), so it is local only: read from the local store, never re-fetched from a provider, never posted, and never in the review.result envelope. Secrets (tokens, DSNs) are redacted at persist and in the live stream.

Pruning

History grows unbounded by default. Trim it explicitly:

miucr history prune --keep 200 --yes             # keep the newest 200, delete the rest
miucr history prune --older-than 30d --yes       # delete records older than 30 days

At least one of --keep / --older-than is required, and --yes confirms the destructive delete. prune emits the history.prune kind with the deleted count.

Auto-prune cap

Set an optional retention cap in config.toml to auto-trim the oldest records after every save:

[history]
enabled = true      # default; set false to disable auto-save globally
max_records = 500    # 0 (default) = no cap