The README claims the query engine is available offline, but only the
merge half was: no subcommand produced the path / frontmatter /
last_modified / etag / size projection that GET /query returns.
`anwesen query` takes the same --vault and --query flags as merge and
writes that document to stdout, compact, one line. src/merge.rs becomes
src/oneshot.rs and holds both: one parse-and-load path, so the two
subcommands cannot drift in grammar or in how strictly they read a
vault. cli.rs shares one VaultQueryArgs between them for the same
reason. rfc3339_z moves from http.rs to query.rs, next to the projection
whose timestamp dialect it is.
Assumed a trailing newline on stdout is wanted here even though merge
writes none: the JSON body is one line for jq and shell pipelines, and
the newline terminates it rather than joining the document. Flag if
wrong.
The hurl harness now compares `anwesen query` against GET /query on the
fixture vault for four queries before running the suite, so a drift
between the two surfaces fails CI.
Hydra's Application::run logs a start failure and returns normally, so
serve exited 0 after the tree never came up. systemd read that as a clean
start and Restart=on-failure never retried; anwesen was unreachable on ap
for 12 hours with NRestarts=0.
Anwesen carries a started flag that start() sets once the supervisor is
up, and main exits 1 when it is still false after run() returns.
Assumed the flag is the only seam available: run() consumes self and
returns (), so the Err is not observable at the call site. Flag if a newer
hydra exposes the result.
Delete the uptrace DSN parser and the {endpoint}/v1/* concatenation that
produced the dead export, and build the OTLP exporters with no endpoint,
headers, or protocol so the SDK reads OTEL_EXPORTER_OTLP_* itself.
The SDK concatenates as naively as we did: measured against a local sink,
a ?query base sends POST /?query/v1/metrics and a #fragment base sends
POST /. So one check stays -- a query or fragment on
OTEL_EXPORTER_OTLP_ENDPOINT fails at startup. Per-signal variables are
used verbatim and need none.
A protocol variable is the second startup check. The exporter picks its
transport at build time and this binary ships OTLP/HTTP alone, so
OTEL_EXPORTER_OTLP_PROTOCOL=grpc kept exporting over HTTP with nothing in
the log -- the same silence this change removes.
service.name stays a default rather than policy: an attribute set on the
resource builder wins over the SDK detectors that read OTEL_SERVICE_NAME
and OTEL_RESOURCE_ATTRIBUTES, so anwesen supplies its own only for keys
the environment leaves alone.
Telemetry-off is our own check on the endpoint variables: with none set
the exporter would still build and aim at the SDK default localhost:4318.
--uptrace-dsn, --otlp-endpoint, and --otlp-header stay parsed but hidden,
so a deployment upgrading with them set fails naming the OTEL_
replacement instead of going quiet.
Assumed hidden clap stubs are the right migration shape; clap rejecting
the flags as unknown would say nothing about the replacement. Flag if a
plain unknown-argument error is wanted instead.
Assumed rejecting http/json alongside grpc is right: neither has a
transport in this build. Flag if a build with both features is wanted
instead.
A directory can be named *.md too, and a vanished path cannot be probed,
so the suffix test alone stranded its notes in the index. Emit the note
delete and the prefix delete together; the two key sets are disjoint.
Optional OTLP export of request-level metrics (request and response-byte
counters keyed by route/status/conditional-GET, a duration histogram out
to 600s) plus server spans for slow or 5xx requests, nested under the
incoming W3C trace context. Off unless an OTLP endpoint or uptrace DSN is
configured; a config-less server is byte-for-byte unchanged.
The IT-86 release pipeline (release.yml) now builds both Linux targets
on GitHub and mirrors the binaries back to Forgejo, proven end-to-end by
the v0.0.0-ci.1 run (ANW-31, Verified). The on-demand Forgejo build from
ANW-30 was the fallback Linux CI until then; it is now redundant.
ap keeps every release and the raw, unstripped binaries are the bulk of
that storage. Strip symbols via the release profile and gzip -9 each
asset before upload (~2.7x smaller); the publish job and Forgejo
mirror-back pass the .gz assets through unchanged.
Assumed .gz per binary over .tar.gz -- the assets are single files, so a
tarball only adds a directory layer. Flag if a bundle (LICENSE/README) is
wanted. E2E needs a release run; piggyback on the next real tag.
Tag-triggered (v* on the Forgejo mirror) instead of a hand-made GitHub
release. Adds an aarch64-linux leg; a single publish job creates the GitHub
release and mirrors every binary back onto the matching Forgejo release via
FORGEJO_RELEASE_TOKEN. Keeps the hurl HTTP contract harness on x86_64-linux.
Assumed v* tags are the release trigger and the contract tests run once on the
deploy target (x86_64-linux); flag if either is wrong.
Collapse the matrix to one linear job on the amd64 :host runner: test
once, then build x86_64 natively and aarch64 by cross-compilation, then
publish both to a Forgejo release. Authenticate the upload with a real
write token (RELEASE_TOKEN, the carvers-ci bot) -- Forgejo's auto
GITHUB_TOKEN cannot upload release attachments. prerelease is a string
input, not boolean, so its value survives an API workflow_dispatch.
Forgejo-side counterpart to the GitHub release.yml (ANW-29). Driven by
workflow_dispatch (UI or API) rather than a release-published event, so a
build can be triggered for a test. Builds the release binary natively per
target, runs the test suite first, and publishes the assets to a Forgejo
release via actions/forgejo-release.
Assumed runner labels linux-amd64 / macos-arm64 must match the forge's
act_runner registration; arm64 needs a native Apple Silicon runner. Flagged
to aav -- correct if the topology differs.
Replace the placeholder README with the full public-facing README, and
switch licensing to BSD 3-Clause per aav's decision on ANW-28: add a single
LICENSE file, drop the dual LICENSE-MIT / LICENSE-APACHE, and update the
Cargo.toml license field to match. README/LICENSE content authored by kaa,
landed verbatim.
Assumed: dropping the old MIT/Apache files and updating Cargo.toml
license = BSD-3-Clause is the consistent completion of aav's BSD decision
(v0.0.0, nothing released relied on the old grant); flag if the dual
license was meant to stay.
On a published GitHub Release, for each target (amd64 Linux, arm macOS) run
the Rust test suite and the hurl HTTP contract tests, build the release
binary on a stable toolchain, and upload it as a target-named asset on the
triggering release. Trigger is release:published per ANW-29 -- aav creates
the release and tag in the UI; the workflow never creates them.
Thin local front-end over ANW-26's merge engine (query::execute_merge):
walk the vault once (vault::scan, same as doctor), evaluate the --query
string, assemble the merged document, write it to stdout. No HTTP, no
watcher, no index. Byte-identical to the HTTP merge mode for the same
vault and query.
Accept: text/markdown on /query concatenates matched note bodies into one
markdown document, each fragment preceded by an HTML-comment source marker.
Adds __anw-order (frontmatter-key ordering, asc/desc, missing-key-last) and
__anw-kind (homogeneity guard, 400 on a mixed merge). Both evaluate over the
full matched set before __anw-limit truncates. Merge mode uses its own
body-cloning projection; the JSON path is unchanged (no body clone).
