relayburn-cli: burn state subcommands (#248 D4)#313
Merged
willwashburn merged 6 commits intomainfrom May 6, 2026
Merged
Conversation
Wire `burn state` as a typed clap subcommand over `relayburn-sdk`:
- `state status` (default) reports per-table row counts in `burn.sqlite`,
the row count in `content.sqlite`, the `archive_state` schema /
last-built / last-rebuild fields, and the resolved retention config.
`--json` emits a structured `StateStatus` payload.
- `state rebuild {index,content,archive,all}` drives
`Ledger::rebuild_derivable`. The four 1.x subtargets collapse onto the
same SQL transaction in 2.0; standalone `rebuild classify` is stubbed
with a follow-up note (the Rust ingest pipeline classifies at append
time per #274).
- `state prune` drives `Ledger::prune_content_older_than`, sourcing the
retention window from `--days` or the resolved config.
- `state reset` is stubbed (drops + re-ingest is filed for follow-up
under #240); it returns exit 1 with a typed message pointing at
`state rebuild all` + `burn ingest` as a workaround.
The TS sibling reports against the 1.x JSONL ledger layout
(ledger.idx, archive.sqlite, content/ sidecars). The 2.0 Rust port has
two SQLite databases with no separate index/archive files, so the
`state-status` golden snapshots are reshaped to match the 2.0 layout —
deliberate divergence tracked under #240. File sizes are omitted from
both human and JSON output because WAL checkpointing makes them
non-deterministic on a logically-empty ledger.
SDK additions:
- `state_status(ledger_home: Option<PathBuf>) -> Result<StateStatus>`
free fn + `LedgerHandle::state_status()` method, with row-count,
archive_state, and config blocks. Three new unit tests cover the
fresh-ledger zero-row path, the appended-turn count path, and the
free-fn `ledger_home` round-trip.
- `RawLedger::read_archive_state_json()` to surface the single-row
`archive_state` snapshot without forcing callers to bind to rusqlite.
CLI tests:
- Two `state-status*` golden invocations flipped to `enabled: true`;
snapshots regenerated.
- Smoke test's `each_stub_exits_one` filters out the now-real `state`
command via a `REAL_COMMANDS` allowlist.
- Fixture-dir `.gitignore` extended to cover `content.sqlite*` /
`*-journal` so a stray run doesn't dirty the working tree.
Follow-ups filed for: standalone `state rebuild classify` reclassify
pass, `state reset` (drop + re-ingest), and a dedicated 2.0-format
fixture corpus for the read-path golden snapshots.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2922e0fb03
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Comment on lines
+336
to
+337
| let cutoff_ms = now_ms.saturating_sub(cutoff_ms); | ||
| let cutoff = format!("ts:{:013}.000", cutoff_ms); |
There was a problem hiding this comment.
run_prune computes cutoff as ts:{:013}.000 from epoch milliseconds, but content.created_at values are written by writer::now_iso as ts:{:020}.{:09} and compared lexically in prune_older_than. With the current format, normal rows like ts:000... are lexicographically less than cutoffs like ts:174..., so burn state prune can delete effectively all content rows regardless of the intended retention window.
Useful? React with 👍 / 👎.
Member
Author
There was a problem hiding this comment.
Fixed in 99ecb8f. run_prune now formats the cutoff via a new format_cutoff_ts helper that mirrors writer::now_iso's ts:{:020}.{:09} shape exactly. Added three regression tests under commands::state::tests: cutoff_matches_writer_format_byte_for_byte, cutoff_is_lex_comparable_against_writer_rows, cutoff_padding_widths_are_stable — the last one pins the 33-char total width so a future format tweak can't silently flip the lex ordering again.
Comment on lines
+212
to
+216
| #[arg(long)] | ||
| pub full: bool, | ||
| /// Reclaim unused SQLite pages after the apply. | ||
| #[arg(long)] | ||
| pub vacuum: bool, |
There was a problem hiding this comment.
archive vacuum positional compatibility
The typed clap shape for state rebuild archive only accepts --vacuum/--full flags, so burn state rebuild archive vacuum is now rejected as an extra argument. The TypeScript command still accepts archive vacuum as a positional action (runArchiveTarget), so this breaks existing scripts/automation that use that invocation style.
Useful? React with 👍 / 👎.
Member
Author
There was a problem hiding this comment.
Fixed in 99ecb8f. Added a typed ArchiveAction ValueEnum and an Option<ArchiveAction> positional on StateRebuildArchiveArgs, so burn state rebuild archive vacuum now parses cleanly. Verified burn state rebuild archive bogus still rejects with [possible values: vacuum]. The dispatch is unchanged: both the positional and --vacuum flag route through rebuild_derivable because 2.0 collapses archive.sqlite into burn.sqlite — kept the surface so 1.x automation doesn't break, with a doc comment explaining the no-op equivalence.
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughImplements a typed ChangesBurn State CLI + SDK
Sequence DiagramsequenceDiagram
participant User
participant CLI as relayburn-cli
participant SDK as relayburn-sdk
participant DB as LedgerDB
User->>CLI: burn state status
CLI->>SDK: state_status(opts)
SDK->>DB: query burn.sqlite (row counts, turns, etc.)
DB-->>SDK: burn results
SDK->>DB: query content.sqlite (row count)
DB-->>SDK: content results
SDK->>DB: query archive_state row
DB-->>SDK: archive_state JSON
SDK->>SDK: assemble StateStatus (burn, content, archive, config)
SDK-->>CLI: StateStatus
CLI->>CLI: format_status() (JSON or human)
CLI-->>User: formatted output
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related issues
Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (2)
crates/relayburn-cli/src/commands/state.rs (1)
389-391: 💤 Low valueNegative
--dayssilently converts to "forever" — consider returning an error instead.A user passing
--days -5likely made an input mistake, but this code silently treats it as "forever" rather than surfacing an error. This could mask user errors and lead to unexpected data retention behavior.🔧 Suggested fix: return None to trigger the existing error path
if !n.is_finite() { return None; } if n < 0.0 { - return Some(relayburn_sdk::Retention::Forever); + return None; // Negative values are invalid; let caller report the error } Some(relayburn_sdk::Retention::Days(n))🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@crates/relayburn-cli/src/commands/state.rs` around lines 389 - 391, The CLI currently treats a negative --days value as "forever" instead of surfacing an error; update the code that parses the --days flag (the branch that converts negative days into the "forever" sentinel) to return None (or Err) when days < 0 so the existing error path is triggered and an error is returned to the user; specifically modify the parsing logic that handles the --days flag/variable to validate days >= 0 and bail out on negative input rather than silently mapping it to forever.crates/relayburn-cli/src/cli.rs (1)
226-231: ⚡ Quick winParse
--daysinto a typed value instead ofString.
Option<String>accepts arbitrary tokens even though this flag only allows a day count orforever. Moving that validation into clap would keep the new state surface fully typed and fail fast on bad input.Possible direction
+pub enum StatePruneDays { + Forever, + Days(u32), +} + #[derive(Debug, Clone, ClapArgs, Default)] pub struct StatePruneArgs { /// Override the configured retention window. Accepts a number /// (days) or the literal `forever`. - #[arg(long)] - pub days: Option<String>, + #[arg(long, value_parser = parse_state_prune_days)] + pub days: Option<StatePruneDays>, /// Delete sidecars even when the source session file still exists. #[arg(long)] pub force: bool, }fn parse_state_prune_days(raw: &str) -> Result<StatePruneDays, String> { if raw.eq_ignore_ascii_case("forever") { return Ok(StatePruneDays::Forever); } raw.parse::<u32>() .map(StatePruneDays::Days) .map_err(|_| "expected a non-negative integer day count or `forever`".to_string()) }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@crates/relayburn-cli/src/cli.rs` around lines 226 - 231, The --days flag is currently parsed into Option<String>, allowing invalid tokens; add a typed parser and wire it into clap so the flag yields Option<StatePruneDays> instead. Implement a parser function (e.g. parse_state_prune_days) that maps "forever" (case-insensitive) to StatePruneDays::Forever and parses a non-negative integer into StatePruneDays::Days(u32) with a clear error message, then use clap's .value_parser / .value_parser_fn or .value_hint to attach that parser to the --days argument and change the CLI field type from Option<String> to Option<StatePruneDays> so validation is done by clap and the rest of the code receives a strongly typed value.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@crates/relayburn-cli/src/cli.rs`:
- Around line 242-245: The --reingest flag is currently allowed without --force;
update the clap arg definition for the "reingest" option (the argument that
defines "--reingest" in crates/relayburn-cli/src/cli.rs) to require the "force"
argument by adding the clap requires dependency (e.g., using .requires("force")
on the builder or requires = "force" in the derive attribute) so the parser
enforces that --reingest can only be used together with --force.
---
Nitpick comments:
In `@crates/relayburn-cli/src/cli.rs`:
- Around line 226-231: The --days flag is currently parsed into Option<String>,
allowing invalid tokens; add a typed parser and wire it into clap so the flag
yields Option<StatePruneDays> instead. Implement a parser function (e.g.
parse_state_prune_days) that maps "forever" (case-insensitive) to
StatePruneDays::Forever and parses a non-negative integer into
StatePruneDays::Days(u32) with a clear error message, then use clap's
.value_parser / .value_parser_fn or .value_hint to attach that parser to the
--days argument and change the CLI field type from Option<String> to
Option<StatePruneDays> so validation is done by clap and the rest of the code
receives a strongly typed value.
In `@crates/relayburn-cli/src/commands/state.rs`:
- Around line 389-391: The CLI currently treats a negative --days value as
"forever" instead of surfacing an error; update the code that parses the --days
flag (the branch that converts negative days into the "forever" sentinel) to
return None (or Err) when days < 0 so the existing error path is triggered and
an error is returned to the user; specifically modify the parsing logic that
handles the --days flag/variable to validate days >= 0 and bail out on negative
input rather than silently mapping it to forever.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 11812568-fa57-4cd1-8f6c-2e0ec3fedb0e
📒 Files selected for processing (11)
CHANGELOG.mdcrates/relayburn-cli/src/cli.rscrates/relayburn-cli/src/commands/state.rscrates/relayburn-cli/src/main.rscrates/relayburn-cli/tests/smoke.rscrates/relayburn-sdk/src/ledger.rscrates/relayburn-sdk/src/query_verbs.rstests/fixtures/cli-golden/invocations.jsontests/fixtures/cli-golden/ledger/.gitignoretests/fixtures/cli-golden/snapshots/state-status-json.stdout.txttests/fixtures/cli-golden/snapshots/state-status.stdout.txt
- Codex P1 (state.rs): prune cutoff was formatted as `ts:{:013}.000` from
epoch ms but content rows are stamped by `writer::now_iso` as
`ts:{:020}.{:09}` and compared lexically — the width mismatch made
every stamped row sort lex-greater than the cutoff, so prune was
effectively a no-op (or, after a rollover, a full wipe). Move the
formatting into a `format_cutoff_ts` helper that mirrors the writer's
shape exactly, with three regression tests covering byte-equivalence,
lex ordering against writer-stamped rows, and constant-width invariant.
- Codex P2 (cli.rs): accept `burn state rebuild archive vacuum` (legacy
TS positional shape) by adding a typed `ArchiveAction` ValueEnum.
Today both the positional and `--vacuum` flag route through the same
`rebuild_derivable` path in 2.0 (no separate archive.sqlite to vacuum),
but existing scripts that target the 1.x surface no longer error out
on the positional.
- CodeRabbit minor (cli.rs): `--reingest` now `requires = "force"` so
the parser rejects bare `--reingest` instead of silently no-opping.
Folds D1+D2+D3 (PRs #315/#312/#314) into the D4 state branch. Resolutions: - CHANGELOG.md: union all four bullets, alphabetized. - crates/relayburn-cli/src/cli.rs: keep all five typed Command variants (Summary/Hotspots/Overhead/Compare/State); merge ValueEnum import; keep D4's state-related types (StateArgs et al.) plus D2/D3's OverheadArgs / CompareArgs / OverheadKind. - crates/relayburn-cli/src/main.rs: auto-merged — five typed dispatch arms. - crates/relayburn-cli/tests/smoke.rs: drop REAL_COMMANDS allow-list and early-continue; remove "state" from UNIMPLEMENTED_SUBCOMMANDS, leaving ["run", "ingest", "mcp-server"]. - tests/fixtures/cli-golden/ledger/.gitignore: superset of patterns. - Cargo.lock: regenerated from origin/main. - tests/fixtures/cli-golden/snapshots/state-status*.stdout.txt: keep D4's 2.0 SQLite layout (deliberate divergence) but refresh row counts to reflect the bootstrapped fixture (golden.rs on main now replays ledger.jsonl into burn.sqlite before running, so the fixture has 18 rows instead of 0).
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@crates/relayburn-cli/src/cli.rs`:
- Around line 337-349: The handler in crates/relayburn-cli/src/commands/state.rs
currently ignores the flags in StateRebuildArchiveArgs (action, full, vacuum)
and StateRebuildAllArgs (force) and always calls run_rebuild_derivable(globals);
update the handler to inspect these fields and either (A) pass them through to
the underlying rebuild routine (extend run_rebuild_derivable or its caller to
accept an options struct with action/full/vacuum/force and apply the requested
behavior), or (B) return a clear, typed "not supported" error when any of those
flags are set so callers don't get a silent no-op; reference
StateRebuildArchiveArgs, StateRebuildAllArgs and run_rebuild_derivable to locate
where to validate and propagate or reject the flags.
In `@crates/relayburn-sdk/src/query_verbs.rs`:
- Around line 1356-1362: The computed total_rows (variable total_rows) only sums
tracked event fields on the rows struct and omits the separate archive_state row
fetched from burn.sqlite, so update the calculation to include the archive_state
count (e.g., add rows.archive_state or the archive_state value fetched later) or
change the variable/label (total_rows -> tracked_rows or similar) and adjust any
human/JSON output code that prints total_rows to avoid claiming a full DB total;
ensure you update both the numeric sum and the displayed key names for
consistency.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 79b89a0c-6861-4533-a3ea-873423e585bf
📒 Files selected for processing (9)
CHANGELOG.mdcrates/relayburn-cli/src/cli.rscrates/relayburn-cli/src/main.rscrates/relayburn-cli/tests/smoke.rscrates/relayburn-sdk/src/query_verbs.rstests/fixtures/cli-golden/invocations.jsontests/fixtures/cli-golden/ledger/.gitignoretests/fixtures/cli-golden/snapshots/state-status-json.stdout.txttests/fixtures/cli-golden/snapshots/state-status.stdout.txt
🚧 Files skipped from review as they are similar to previous changes (1)
- tests/fixtures/cli-golden/invocations.json
- state rebuild archive --full / --vacuum (and the legacy `vacuum` positional) and state rebuild all --force now print stderr breadcrumbs explaining the no-op in the 2.0 layout, instead of silently discarding the flag. Adds a small `report_advisory` helper in render/error.rs; stderr in both human and --json modes so JSON stdout stays single-shape. - StateStatus.burn renames `total_rows` (`totalRows` JSON) to `tracked_rows` (`trackedRows`) — the field sums per-table event rows and excludes the `archive_state` metadata row, so the name now reflects the scope. Human format updates from "rows: N total" to "tracked rows: N". Snapshots refreshed. - state_status now plumbs the active ledger home into config loading via a new `load_config_with_home` helper in ledger/config.rs. `--ledger-path foo state status` previously reported the env-default home's retention/store settings; now it reads `foo/config.json`. resolve_config_summary returns Result so IO errors propagate through the typed reporter instead of being unwrapped to defaults.
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@crates/relayburn-cli/src/commands/state.rs`:
- Around line 307-331: The config is being loaded with load_config() which uses
the env home and mismatches the ledger path passed via globals.ledger_path;
update run_prune to load the config using the explicit ledger/home from
GlobalArgs instead of load_config() so the same path is used for retention logic
and database opening (mirror the fix applied in state_status). Locate
load_config call in run_prune and replace it with the equivalent config-loading
function that accepts the home/ledger path (e.g. a load_config_with_home or
load_config_from_path variant) or otherwise pass globals.ledger_path into
load_config so retention_days comes from the same ledger path used later. Ensure
any error handling and report_ledger_error usage remains unchanged.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: bbedfb70-265f-4330-a79f-b94319e621f6
📒 Files selected for processing (7)
crates/relayburn-cli/src/commands/state.rscrates/relayburn-cli/src/render/error.rscrates/relayburn-sdk/src/ledger.rscrates/relayburn-sdk/src/ledger/config.rscrates/relayburn-sdk/src/query_verbs.rstests/fixtures/cli-golden/snapshots/state-status-json.stdout.txttests/fixtures/cli-golden/snapshots/state-status.stdout.txt
🚧 Files skipped from review as they are similar to previous changes (1)
- crates/relayburn-sdk/src/ledger.rs
Follow-up to 344e899. The prior fixup wired `state_status` to load config from the ledger's active home, but `run_prune` was missed — it was still calling bare `load_config()`, so `burn --ledger-path /foo state prune` would read retention_days from `$RELAYBURN_HOME`'s config.json against `/foo`'s content DB. Mirror the state_status fix: thread `globals.ledger_path` into `load_config_with_home`. Re-export `load_config_with_home` and `config_path_at_home` from `relayburn-sdk::lib` so external callers (CLI today; future embedders) have access via the canonical surface rather than reaching into the `ledger` module.
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@crates/relayburn-cli/src/commands/state.rs`:
- Around line 145-150: The rel_to_home function currently rewrites any
byte-prefix match of home (path.starts_with(home)), which can mislabel paths
like "/x/home2/..." when home="/x/home"; fix by normalizing/trimming trailing
slashes from home and only treating path as inside home when either path == home
or path starts with home + "/" (i.e., check path == home ||
path.starts_with(&(home.to_owned() + "/"))), then compute the rest as before;
reference: rel_to_home, path, home.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 70f27b7d-f372-4290-8bf7-dbf2fd1af8e0
📒 Files selected for processing (2)
crates/relayburn-cli/src/commands/state.rscrates/relayburn-sdk/src/lib.rs
`rel_to_home` operates on `&str`, so the prior `path.starts_with(home)`
was a byte-prefix match — `/x/home2/foo` against `home="/x/home"` would
get rewritten to `${RELAYBURN_HOME}/2/foo`. Switch to a separator-
bounded check via `strip_prefix` + `'/'` peek; normalize a trailing
slash on `home` so `home="/x/home"` and `home="/x/home/"` behave the
same; bail to a passthrough for degenerate inputs (`home=""`,
`home="/"`).
Five new unit tests cover the fix path, the byte-prefix collision the
guard prevents, the trailing-slash normalization, the degenerate-home
inputs, and the `path == home` boundary case (preserves the prior
trailing-slash output shape so downstream callers don't see a
behavioral shift).
willwashburn
added a commit
that referenced
this pull request
May 7, 2026
…et hook scope) - Merge origin/main (D6 codex landed); CHANGELOG additive, registry/mod auto-merged with codex slot from main. - Codex P2: route burn ingest one-shot summary to stdout (matches TS runIngestOnce at packages/cli/src/commands/ingest.ts:121-126); --watch banner + --hook breadcrumbs stay on stderr. - CodeRabbit Minor: scope --quiet to --hook via clap requires = "hook" (mirrors the --reingest requires = "force" pattern from #313). Standalone --quiet and --watch --quiet are now rejected at parse time.
willwashburn
added a commit
that referenced
this pull request
May 7, 2026
…319) * relayburn-cli: burn ingest + burn mcp-server (#248 D8, closes #210) Wire `burn ingest` as a thin presenter over `relayburn_sdk::ingest_all` plus the SDK's `start_watch_loop` controller. Three modes mirror the TS sibling at `packages/cli/src/commands/ingest.ts`: - No flags: one full sweep, then exit. Logs `[burn] ingest: ingested N session(s) (+M turn(s))` on stderr. - `--watch [--interval MS]`: foreground poll loop. Skips empty-tick log lines (TS shape); SIGINT / SIGTERM stop drains in-flight ticks before returning. - `--hook claude [--quiet]`: stdin-driven Claude hook entrypoint. Validates the `session_id` / `transcript_path` payload shape, then drives a full sweep (cursor short-circuit means cost is bounded by new turns, not the hook payload). Failure paths log + exit 0 to avoid blocking the parent Claude tool call. Wire `burn mcp-server` as a hand-rolled JSON-RPC 2.0 stdio server, mirroring `packages/mcp/src/server.ts` rather than depending on `rmcp`. The on-wire surface (`initialize`, `ping`, `tools/list`, `tools/call`) is small enough that a tight in-tree implementation beats freezing a heavy SDK version. Single tool ships in this PR: `burn__sessionCost`, a 1:1 port of `packages/mcp/src/tools/session-cost.ts` delegating to `LedgerHandle::session_cost`. Future tools (summary, hotspots, …) land separately. Closes #210. Smoke test drops `ingest` and `mcp-server` from `UNIMPLEMENTED_SUBCOMMANDS`; `--help` for both subcommands still passes the existing help-shape assertion. Manual stdio handshake verified: `initialize` echoes the client's `protocolVersion`, `tools/list` returns `burn__sessionCost`, and `tools/call burn__sessionCost` returns the SDK payload as both stringified `text` content and a `structuredContent` mirror. * relayburn-cli: address PR #319 review (rebase + stdout report + --quiet hook scope) - Merge origin/main (D6 codex landed); CHANGELOG additive, registry/mod auto-merged with codex slot from main. - Codex P2: route burn ingest one-shot summary to stdout (matches TS runIngestOnce at packages/cli/src/commands/ingest.ts:121-126); --watch banner + --hook breadcrumbs stay on stderr. - CodeRabbit Minor: scope --quiet to --hook via clap requires = "hook" (mirrors the --reingest requires = "force" pattern from #313). Standalone --quiet and --watch --quiet are now rejected at parse time.
This was referenced May 7, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
burn stateas a typed clap subcommand overrelayburn-sdkwithstatus(default),rebuild {index,content,archive,all},prune, andresetverbs.--jsonflips every read path onto a structured payload.state_statusto the SDK: free fn +LedgerHandle::state_status()method returning per-table row counts, archive_state metadata, and resolved retention config. Three unit tests cover the fresh / populated / free-fn round-trip paths.state-status*golden invocations and reshape the snapshots to the 2.0 SQLite layout.2.0 vs 1.x — golden snapshot divergence
The TS state-status output references the 1.x JSONL ledger artifacts (
ledger.idx,ledger.content.idx,archive.sqlite,content/sidecars). The 2.0 Rust ledger has two SQLite databases (burn.sqlite+content.sqlite) with no separate index files and the archive_state row collapsed insideburn.sqlite. The Ruststate statustherefore reports the 2.0-shape view: per-table row counts in the events DB, a row count in the content DB, and thearchive_stateschema/last-built/last-rebuild fields. File sizes are omitted from both human and JSON output because WAL checkpointing makes them non-deterministic on a logically-empty ledger. This is a deliberate divergence from "byte-equivalent with TS" tracked under #240; sibling D-PRs (D1/D2/D3) will hit the same architectural drift on their read-path snapshots.Stubs + follow-ups
state reset(full wipe + optional--reingest) returns exit 1 with a typed message — the 1.x implementation walked$RELAYBURN_HOMEand unlinked individual JSONL files, none of which exist in 2.0. The 2.0 equivalent (drop + recreate the two SQLite DBs, then re-ingest) is filed for follow-up under Epic: rewrite burn in Rust (v2 — supersedes #222) #240. Workaround:burn state rebuild all && burn ingest.state rebuild classify(standalone reclassify pass) returns exit 1 with a typed message — the 2.0 ingest pipeline classifies at append time per [Rust port] relayburn-analyze: hotspots aggregator #274. Filed for follow-up.index/content/archive/all) collapse ontoLedger::rebuild_derivablein 2.0 (single SQL transaction over the derivable tables); they're kept as separate clap subcommands so existing automation doesn't break.Files touched
crates/relayburn-cli/src/cli.rs—Command::State(StateArgs)with nestedStateSubcommandenum and per-verb args structs (StateStatusArgs,StateRebuildArgs,StateRebuildTarget,StatePruneArgs,StateResetArgs).crates/relayburn-cli/src/commands/state.rs— real presenter dispatching on the subcommand; human + JSON renderers;format_intmatching TSIntl.NumberFormat('en-US'); typed-error mapping that pullsLedgerErrorout ofanyhow::ErrorsoEXIT_LEDGER_ERROR(3) still surfaces.crates/relayburn-cli/src/main.rs— dispatch arm.crates/relayburn-cli/tests/smoke.rs—REAL_COMMANDSallowlist for theeach_stub_exits_onetest.crates/relayburn-sdk/src/query_verbs.rs—StateStatus+ supporting types,LedgerHandle::state_statusmethod,state_status(opts)free fn, three unit tests.crates/relayburn-sdk/src/ledger.rs—RawLedger::read_archive_state_json()helper.tests/fixtures/cli-golden/invocations.json— flipstate-status/state-status-jsontoenabled: true.tests/fixtures/cli-golden/snapshots/state-status*.stdout.txt— regenerated against the 2.0 layout.tests/fixtures/cli-golden/ledger/.gitignore— extended to covercontent.sqlite*and-journalso a stray run doesn't dirty the tree.CHANGELOG.md—[Unreleased]bullet.Test plan
cargo test --workspace(608+ pass)BURN_GOLDEN=1 cargo test --test golden -p relayburn-cli—state-status+state-status-jsonpassburn state --help/burn state status --help/burn state rebuild --help/burn state prune --help/burn state reset --helpmirror the TS flag setburn state statuson a fresh$RELAYBURN_HOMEproduces stable output across consecutive runs (verified — file sizes omitted)🤖 Generated with Claude Code