Publish outputs to checked destinations and preserve failure evidence

This commit is contained in:
2026-11-16 03:52:35 +00:00
parent cea7e44232
commit b63c8af467
11 changed files with 656 additions and 336 deletions

View File

@@ -20,16 +20,19 @@ validator entries yield `validation_decisions_filtered` only for a successful
operation; they do not count as filtered correction proposals. Failed operations
retain their execution counts and never produce a recovery record.
`internal/core/diagnostics.OutputSession` owns active destination checks,
transcript preflight, report initialization, regular-file replacement, and direct
stream publication. Transcript input is protected from early report writes but
`internal/core/diagnostics.OutputSession` owns destination selection, report
initialization, and publication. Transcript and report files share a staging
primitive that writes beside the resolved destination. Streams are opened for
delivery. Transcript input is protected from early report writes but
may be the final transcript destination. The LLM runtime records consumed PromptKit sources and effective host message
files and supplies those source paths to CLI;
no second filename-based discovery rule is used.
CLI composes it after configuration, transcript acquisition, and preflight
readiness and before dispatching model work. It does not create ordinary report parents.
On a final write error it attempts other enabled reports and one failed-status
pass; a published transcript is never erased by a later report failure.
Final report rendering uses the delivery and cleanup errors already observed.
If a report write first fails, the remaining writes use the failed bundle and
one pass attempts to update the reports still missing failed evidence. A
published transcript remains available after a later report failure.
Debug is independent. The shared `--debug` flag enables a unique retained child
under `diagnostics.work_dir`. The CLI creates one interaction sink rooted in

View File

@@ -34,15 +34,20 @@ untouched until successful publication. Pipes and devices are opened only for
delivery, and a pipe can wait for its reader. If a transcript destination changes
between a regular file and a pipe or device during processing, Audita can report a
delivery failure while preserving the new destination. Parent directories for
required outputs must already exist.
required outputs must already exist. Publication uses the destinations resolved
during preparation. Give concurrent invocations separate destinations, and
choose separate report directories when retaining reports from multiple runs.
Enabled reports are initialized as `in_progress` before processing. The
transcript is published before final reports; the files do not form a single
transaction. A late report failure may therefore leave a published transcript.
Fatal errors and cancellation clean unpublished temporary resources and attempt
failed-status final reports. An uncatchable termination can leave `in_progress`
evidence. Read exit status and reports together rather than treating a file's
existence as proof of success.
After a required write failure or cancellation, Audita attempts to clean
unpublished temporary resources and publish failed-status reports. A cleanup
failure is reported with the earlier error. Cancellation observed after a
stream write still fails the invocation, even when some or all bytes reached
the reader. If failed reports cannot be rendered, available report evidence
remains. An uncatchable termination can leave `in_progress` evidence. Use the
exit status and report status to determine the outcome.
## Debug evidence
@@ -50,7 +55,9 @@ The shared `--debug` flag enables retained artifacts in a run directory beneath
`diagnostics.work_dir`; selecting a work directory alone does not enable capture.
Optional debug preparation or writes can produce warnings without
changing correction decisions or the required output result. Debug material can
include transcript content, prompts, responses, and correction decisions. Protect this material according to the sensitivity of the inputs and responses. Repair interactions
include transcript content, prompts, responses, and correction decisions.
Protect this material according to the sensitivity of the inputs and responses
and manage its retention. Repair interactions
and revalidation artifacts include proposal and attempt coordinates; the
correction ledger connects them to the original edit.

View File

@@ -186,6 +186,13 @@ after I/O, cleanup errors, initialized evidence when rendering fails, and
write behavior after known failure. Observe publication attempts as well as
final files. Leave processing fact and ledger projection to Stage 17.
**Complete.** Required outputs use the destinations resolved during selection
and share regular-file staging. Publication checks cancellation, retains and
reports cleanup failures, and uses failed-status reports when required delivery
cannot complete. Tests verified publication attempts, retained evidence when
rendering failed, symlink destinations, stream cancellation, and Markdown path
escaping. The CLI build and full offline Go suite passed.
## Stage 5 — Consolidate typed transcript data
Audit #13, #31, #51, #52, #54, #62 (file helper).