Document the new configuration and processing workflow

This commit is contained in:
2026-11-10 17:52:14 +00:00
parent 926fa26a31
commit ba1d5d2367
61 changed files with 832 additions and 765 deletions

View File

@@ -24,36 +24,27 @@ 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
may be the final transcript destination. The LLM runtime records successful reads
through PromptKit's profile filesystem and supplies those source paths to CLI;
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/profile inspection and before reading
inputs or dispatching model work. It does not create ordinary report parents.
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.
Debug is independent. Only `diagnostics.debug` creates a unique retained child
Debug is independent. The shared `--debug` flag enables a unique retained child
under `diagnostics.work_dir`. CLI passes empty diagnostics paths and nil writers
when debug is off. With debug on, source/normalized snapshots, invocation,
effective settings, origins, profiles, initial chunk summary, correction ledger,
indexed execution/utilization, and model interactions are written. Runner
candidate capture includes proposals later rejected or interrupted. Ledger
mapping uses optional string pointers so an empty proposed replacement is
serialized for every candidate and an empty applied replacement appears only
when that candidate applied; see the [public artifact contract](../integrations/output-files.md#debug-only-files).
Debug write failures reach the CLI through an explicit concurrency-safe diagnostic
callback and produce redacted stderr messages without aborting processing.
Completed validator decisions returned with a fatal batch error remain available
to the debug ledger; they never authorize applying an incompletely validated edit.
Runner decision records carry the validator's explicit approved, rejected, or
unavailable disposition into the ledger. Unavailable candidates have a distinct
top-level ledger disposition, while fatal interruptions remain interrupted when
no rejection or unavailable decision was returned. Successful validation
interaction payloads contain the full decoded response, final content, and
reconciliation outcomes; ignored indices do not become ledger candidates.
Every debug writer uses all configured backend secrets for redaction, including
unselected backend credentials and source snapshots. Debug can still contain
sensitive transcript and prompt content.
when debug is off. With debug enabled, it captures source and normalized
snapshots, invocation, effective settings, correction ledger, LLM execution
observations, and model interactions. Candidate capture includes proposals later
rejected or interrupted. Ledger mapping preserves empty proposed and applied
replacements and distinguishes unavailable candidates from interrupted work.
A concurrency-safe diagnostic callback reports optional write failures without
aborting processing. Diagnostic redaction uses credential source metadata; debug
artifacts can still contain sensitive transcript and prompt content. Execution
position and optional operator ID are carried separately from generated safe
artifact path coordinates.
## Key Tests