First slice of the plan-generation pipeline epic. A pure-Kotlin PlanLinter runs over the
compiled freestyle ExecutionPlan (WorkflowGraph) before it is locked, complementing
ExecutionPlanCompiler (which throws on unreachable/unknown-kind/bad-edge) with the checks
it does NOT make:
- HARD unproduced_need — a stage needs an artifact no stage produces. Real gap: only the
TOML loader checked this; the freestyle compiler did not, so such plans compiled and
failed at runtime.
- HARD trap_state — a stage with no path to the terminal (inescapable loop / dead end).
Uses terminal-reachability, NOT "any cycle", so legitimate verdict-gated review loops pass.
- SOFT stage_count / fan_out / empty_brief / duplicate_brief — scored, never blocking.
FreestyleDriver lints after compile and records PlanLintCompletedEvent (pure function of
the recorded plan → replay-safe, no env observation in v1); a hard failure rejects the plan
(ExecutionPlanRejectedEvent source="lint") before the operator is asked or the plan locks,
so a broken plan never executes. Pays off now for single-plan freestyle; becomes the
per-candidate filter when multi-candidate generation (Slice 2) lands.
Deferred to later slices: file-path grounding + symbol resolution (needs an index),
token-budget/ADR-bypass checks, and a dedicated :core:planning module.
Run operator-configured static-analysis commands (compiler/detekt/formatters) as a
deterministic harness step on the producing stage, before the LLM reviewer. A stage
declares `static_analysis = [commands]`; after it produces its artifact, each command
runs in the workspace root and the results are recorded in a StaticAnalysisCompletedEvent
(invariant #9 env observation — recorded live, folded on replay, never re-run). A
non-clean command fails the stage retryably with its output fed back verbatim (§2:
compiler/test output is ground truth), so the implementer fixes it and only static-clean
code ever reaches the reviewer. This makes §5's "reviewer context excludes static findings"
mechanical — the findings are resolved upstream, so the reviewer can't waste inference on
what detekt catches for free; no output-parsing, no context plumbing.
- StaticAnalysisCompletedEvent + StaticAnalysisFinding (core:events) + registration.
- StaticAnalysisRunner seam + ProcessStaticAnalysisRunner (whitespace-split argv, stderr
merged, timeout→nonzero exit) in core:kernel; wired (nullable) into OrchestratorEngines
and the server. Null runner / no workspace root → logged no-op, never blocks the run.
- StageConfig.staticAnalysis + `static_analysis` TOML field; runStaticAnalysis folded into
a runPostStageGates sequence (produces → grounding → echo → static analysis).
- Documented `static_analysis` on the role_pipeline implementer stage (commented; commands
are workspace-specific). The reviewer prompt half shipped in 3467826.
New brief_echo stage before the planner in role_pipeline: the model restates
the analyst brief as a structured artifact, and a deterministic gate
(checkBriefEcho, opt-in via brief_echo=true) diffs the restatement against the
original brief. On divergence — dropped requirements (Jaccard coverage < 0.34)
or hallucinated files — it emits BriefEchoMismatchEvent and fails the stage
retryably, so a misread brief never reaches plan generation.
The diff (BriefEchoDiff) is a pure function of two recorded artifacts, so it is
replay-safe and records no observation; the event fires only on mismatch, for
audit. Symbols are recorded but non-blocking in v1. Mirrors the groundBrief-
References gate. role_pipeline only; freestyle_planning wiring is a follow-up.
Runtime install (like research): copy brief_echo.json + brief_echo.md and the
[[artifacts]] block into ~/.config/correx.
Third HealthProbe (after disk + llama): times a cheap lastGlobalSequence()
read as an event-store responsiveness heartbeat — DEGRADED on throw or when
latency exceeds eventStoreLatencyWarnMs (default 500). Read-only by design
(no probe writes into the source-of-truth log; Invariant #1). Plugs into the
existing HealthMonitor; no new events. Observability spec §4.
Also wraps a long line in LlamaServerHealthProbeTest to clear a detekt
MaxLineLength finding introduced in aaf896d.
Second HealthProbe (after disk-watermark): liveness via GET /health on the
configured llama base URL + tokens/sec degradation watch derived from recorded
InferenceCompletedEvents (windowed average vs configurable threshold, opt-in via
llamaTpsWarnBelow). Plugs into the existing HealthMonitor edge-detection loop;
no new events. Observability spec §4.
The rubber-duck path emits a fenced {"ideas":[…]} block; Ideas.parse strips it
from the chat turn and onUserInput records an IdeaCapturedEvent each (CHAT only).
IdeaReader folds eventStore.allEvents() into the active cross-session board, and
DefaultRouterContextBuilder injects a capped "## Ideas" entry — router-only, so
workflow stages never see it. Extended SYSTEM_PROMPT option (d) to capture ideas.
IdeaCapturedEvent(ideaId, sessionId, text, timestampMs) + IdeaDiscardedEvent
(ideaId, sessionId) tombstone, registered in eventModule. The cross-session idea
board is rebuilt from these via eventStore.allEvents(); discard hides without
deleting (invariant #1).
The triage system prompt now appends a fenced json propose_workflows block;
WorkflowProposals.parse strips it from the visible chat turn and onUserInput
records a WorkflowProposedEvent (CHAT only). Candidate ids are validated against
the registered workflows (invariant #7) so an invented id never reaches the
operator. Bumped a token-budget test for the larger prompt.
ProposedWorkflow + WorkflowProposedEvent(sessionId, proposalId, prompt,
candidates, originalRequest), registered in the polymorphic eventModule.
Records the router's triage suggestion of candidate workflows as a
non-authoritative fact the operator acts on.
When a stage produces an LLM artifact carrying a non-empty questions
array, the orchestrator parks the run, records ClarificationRequested,
awaits the operator's answers (submitClarification seam), records
ClarificationAnswered, injects the answered Q&A as an L2 USER entry, and
re-runs the same stage so the questions are resolved by the role that
asked them. Runs on a dedicated pendingClarifications map and the
enterStage success-branch loop, so a clarification round never consumes
the failure-retry budget; capped at three rounds. Integration test covers
the happy path and the cap.
Adds the vocabulary for a stage raising open questions to the operator:
ClarificationQuestion/ClarificationAnswer value types, the
ClarificationRequested/ClarificationAnswered events (registered in the
event module), a ClarificationRequestId, and ClarificationQuestions.parse
— a lenient, fence-tolerant reader that pulls an optional questions array
out of an LLM artifact (shared by orchestrator and server).
The schema declared sources as an array of objects, which correx's flat
JsonSchema model can't represent (JsonSchemaProperty has no nested
properties). The strict config loader rejected it at startup, silently
disabling the source_dossier artifact kind. Flatten sources to an array
of strings (URL-prefixed per-source notes), update the gather prompt to
match, and add ResearchSchemaLoadTest to guard every shipped research
schema against the strict loader.
[tools.research].enabled now defaults true, matching shell/file tools. Safe because
web_fetch is T2 — nothing leaves the machine without operator approval regardless of
the flag; the flag only controls whether the tools are registered. web_search (T1) hits
only the configured SearXNG endpoint.
Makes the research feature runnable end-to-end, off by default.
- config: [tools.research] (enabled, searxng_url, max_results, max_fetch_bytes).
- registration: web_search/web_fetch are built into BOTH the default and per-workspace
tool registries when research.enabled, sharing one HTTP client threaded from Main
(none built on the static path). Egress stays harness-enforced: web_fetch is T2
(operator-approved) and the existing NetworkHostRule still applies.
- workflow: examples/workflows/research.toml — decompose → gather → report, with the
three artifact schemas and prompts. Fan-out (search per sub-question, fetch per source)
runs as repeated tool calls inside the gather stage (Correx has no parallel agents);
per-source synthesis into the dossier is the compression step, so the report stage
consumes summaries, never raw pages. ResearchWorkflowTest validates the graph contract.
To run: set [tools.research].enabled, register the 3 [[artifacts]], copy research.toml +
prompts + schemas into the workflows dir, start SearXNG. Launch like any workflow (the
T2 fetch approval surfaces as an approval card; the report opens in the artifact viewer).
Follow-ups (noted, not blocking): batch fetch-approval at the source-list level (§3),
a dedicated SourceFetched/LowQualityExtraction event (quality + content hash are already
in tool-result metadata), dynamic per-session egress allowlist, and the web approval client (§6).
Health states are events. A HealthMonitor polls registered HealthProbes on an
interval and appends HealthDegraded/HealthRestored only on a status transition
(hysteresis), so health is observable, replayable, and renderable like every
other fact in the log. The probe reads nondeterministic environment state and
records it at observe-time (Invariant #9); HealthProjection rebuilds current
health by replay, never re-probing (Invariant #8). Monitor is live-only, started
in ServerModule.start alongside narration; seeded from the projection so a
restart does not re-emit a degraded the log already records.
First concrete probe: DiskWatermarkProbe — event-log + CAS size and growth rate,
warning before the 2am surprise. Read surface mirrors `correx stats`:
GET /health/checks replays the system session and `correx health` renders it.
Health events ride the reserved __system__ session, filtered from operator
session listings.
LLAMA_SERVER (liveness + tokens/sec trend) and EVENT_STORE (append/fsync latency)
probes plug into the same HealthProbe framework as follow-ups.
Local analysts hallucinate file paths the same as planners. A stage can now
opt in with `ground_references = true`: after it produces its brief, every
workspace-relative file path the brief named is checked for existence, and a
path that doesn't exist fails the stage (retryable) with the misses fed back
("Reference only files you have read") so the model re-grounds.
Grounding probes the filesystem directly rather than the repo map — the map
is recency-ranked and depth/extension-limited, not a complete existence
oracle, so "absent from the map" can't be a hard failure. Existence is a
nondeterministic observation, so the result is recorded as a
BriefGroundingCheckedEvent (invariant #9); replay reads it back and never
re-probes.
- BriefGroundingCheckedEvent + GroundingReference, registered in eventModule
- BriefReferenceExtractor: pure, deterministic pull of relative file-path
references from a brief artifact JSON (needs a '/' and a dotted extension;
skips modules, bare names, URLs, absolute/parent paths — so coarse
"subsystem" mentions the brief is allowed to make aren't false-flagged)
- SessionOrchestrator.groundBriefReferences on the post-verifyProduces path,
gated by stage metadata; uses the existing worldProbe
- TomlWorkflowLoader `ground_references` → metadata; role_pipeline analyst
opts in
Scoped to file references with a '/' + extension (not bare basenames, which
a depth-limited probe can't resolve), so a flagged miss is high-confidence.
The symbol-grounding half of §3 is left for a real symbol index.
Scope creep — an implementer writing files it never declared — is the
dominant local-model failure that compiler/tests don't catch. A stage can
now declare a `writes` manifest (workspace-relative globs); a FILE_WRITE
that resolves inside the workspace but outside the declared set is raised
as PATH_OUTSIDE_MANIFEST → BLOCK, so the model gets the violation as tool
feedback and retries within scope (the §2 bounded-retry signal).
Implemented as a plane-2 rule beside PathContainmentRule (same effect-based
dispatch on FILE_WRITE, same symlink-safe WorldProbe resolution, same
recorded observations so replay reads them back — invariant #9). An empty
manifest is unrestricted; out-of-workspace targets stay PathContainmentRule's
concern, so the two rules don't double-flag.
- StageConfig.writeManifest + TomlWorkflowLoader `writes` parsing
- ToolCallAssessmentInput.writeManifest, threaded from the active stage via
SessionOrchestrator.runPlane2Assessment
- new ManifestContainmentRule, registered in the server assessor
- shared candidatePathStrings extracted so both path rules judge the same
target set
- role_pipeline.toml documents the option on the implementer stage
The dedicated ManifestViolationEvent the spec names is not added: the
violation is already recorded replayably in ToolCallAssessedEvent (issue +
observations + BLOCK) and surfaced in the TUI rationale band. A first-class
event is worth adding only once reconciliation consumes it.
The JSON-schema artifact validator enforced `required` keys but not array
contents, so a model could emit `"requirements": []` (or `"steps": []`,
`"components": []`) and pass — a misread brief sailing through as a valid
artifact. That empty-list gap is the hole under role-reliability §3's
keystone ("force the analyst to define done; verifiability propagates
downstream").
- JsonSchemaProperty gains `minItems` and `items` (recursive), letting a
schema declare a required-non-empty / element-typed array.
- JsonSchemaValidator checks array length >= minItems and each element's
declared item type (with index in the message). Pure + deterministic;
single-return to stay within the detekt ReturnCount budget.
- Role-pipeline schemas tightened: analysis.requirements, design.components,
impl_plan.steps → minItems 1 + items string. Prompt-free — these arrays
were already `required`; this only forbids them being empty.
§4 (force a non-empty `alternativesConsidered` on the architect) is
deliberately NOT taken: it contradicts the standing "don't raise
alternatives unnecessarily" preference the architect prompt already encodes.
Layer 1 of tui-requirements §5: a no-LLM analyzer that shell-lexes a
proposed command, classifies binaries against a known-command table, and
mechanically flags the constructs that widen blast radius — sudo/doas,
recursive delete, pipe-to-shell (skipping wrappers so `| sudo sh` counts),
redirects outside the workspace, network binaries, base64/eval, and
unparseable input (itself a red flag). Pure + replay-stable: same preview
+ workspaceRoot → same analysis, no environment reads.
The approval band renders command approvals as a card — worst-first flag
chips over the reconstructed command line, each token colored by severity
(T0–T4 ramp). ^x opens a scrollable fullscreen view so a long command is
never truncated. Diff and plain-preview paths are unchanged.
Server: computeToolPreview now renders the shell tool's argv as a full,
shell-quoted command line instead of the 200-char JSON-args fallback,
honouring §3's "full untruncated command". The TUI parser accepts both
the rendered line and the older `{"argv":[...]}` shape for replay.
Layer 2 (model annotation) deliberately deferred per the spec.
The router matched only the adapter-prefixed registry id
(llama-cpp:narrator) while the operator's [router.narration] model_id
naturally repeats the bare [[providers]] id (narrator), so the match
always failed and narration silently fell back to the slow main model.
Match either form.
Every TUI config-editor save regenerates the TOML through
CorrexConfigWriter, whose [router.narration] section never wrote
model_id — one save and narration silently fell back to the main
model, queueing every narration behind slow 9B inferences (live QA:
approval narrations arrived minutes late, after the workflow had
already failed). The writer now persists it and the field is editable
in the TUI (blank clears it).
#2: approval pauses narrate from ApprovalRequestedEvent (carries
tool/tier/preview/stage directly) instead of OrchestrationPaused +
a pending-approvals map the next event had not yet populated — the
narrator always lost that race and produced generic text.
#3: NarrationTrigger now carries the triggering event's stageId; the
narration status line and RouterNarrationEvent.stageId use it instead
of the router projection's currentStageId, which terminal events have
already cleared ('Stage: none').
#9: ExecutionPlanLocked -> plan.locked and ArtifactValidated ->
artifact.validated mapped to the TUI (Go: feed lines, plan shows the
compiled stage chain); ArtifactValidating explicitly dropped.
#10: blank router turn guard — a length-finish with empty text now
emits an explanatory turn naming the max_tokens budget instead of a
silent blank; blank steering responses no longer emit steering notes.
#12: resumeAbandoned only auto-resumes sessions whose last event is
within orchestration.resume_abandoned_max_age_minutes (default 24h,
0 disables); staler sessions stay parked for POST /resume. Knob is
TUI-editable and persisted.
ExecutionPlanCompiler treated the plan's single 'produces' string as a
registry kind id, while the architect prompt described it as a unique
artifact id referenced by needs/edges — the model followed the prompt,
invented descriptive ids (files_created), and every freestyle plan
failed to compile. Collapsing both onto one string is also unsound:
two stages producing the same kind would collide on slot name and make
artifact_validated edge conditions ambiguous.
Plan stages now carry an optional 'kind' selecting the registered kind
(mirroring TOML's produces = { name, kind }); 'produces' stays the
unique slot name. Missing 'kind' falls back to the old produces-as-kind
reading so existing plans still compile. Vocabulary entry, architect
prompt, and execution_plan schemas updated to match.
LlamaCppTokenizer posted {"tokens":[text]} but llama.cpp's /tokenize expects
{"content":text}; the server silently answered {"tokens":[]}, so countTokens
returned 0 for every string. Every context entry carried tokenEstimate=0 and
the whole budget/trim pipeline was blind — live QA saw a 34k-token prompt
sail through a 16384 stage budget (three raw file_read results of plan docs),
crater t/s, degrade output, and fail the stage on validation 3x.
Also guard estimateTokens: a zero count for non-blank content falls back to
the chars/4 heuristic so a lying tokenizer can never blind budgeting again.
Three gaps found by live QA (kill between planning-complete and plan lock):
- launchSessionResumeWithRehydrate never handed off to FreestyleDriver, so a
resumed freestyle session stranded at planning-complete with no plan lock
and no phase 2; now locks+runs when planning completed and no
ExecutionPlanLockedEvent exists yet.
- resume route 400'd on phase-2 sessions (workflowId freestyle-<sid> is
compiled, never registered); now recompiles the locked plan after
rehydrating the artifact cache.
- orchestrator.resume threw when currentStageId was terminal ('done') or
unknown to the graph; now returns WorkflowResult.Completed.
The line-based parser treated 'conventions = [' as the complete value,
binding conventions to listOf("[") and dropping every entry on the
following lines. Accumulate value lines until the top-level bracket
closes (quote-aware), skipping comments inside the array.
Strict chat templates (Qwen3.5) reject any system message after index 0.
Recalled L3 memory, L2 stage summaries, and retrieval entries carry
EntryRole.SYSTEM and rendered as standalone mid-conversation system turns,
making llama-server 400 on every router turn once L3 recall returned hits.
- Add existsByTurnIdPrefix(prefix) to L3MemoryStore interface; implemented in
InMemoryL3MemoryStore (mutex-guarded) and TurboVecL3MemoryStore (metadata map).
All test fakes (TrackingL3MemoryStore, CapturingStore) updated.
- Introduce RepoMapIndexerPort interface; RepoMapIndexer implements it. Allows
injection of test doubles without framework overhead.
- ProjectMemoryService.indexAndRecord: reads latest WorkspaceStateObservedEvent
from event store (invariant #9 — recorded fact, not re-probe); skips scan+embed
when L3 already holds entries for "repomap:<root>:<stateKey>"; passes stateKey
to RepoMapComputedEvent; embeds each scanned entry after append; per-entry
failures warn-logged (CancellationException rethrown).
- Tests: InMemoryL3MemoryStoreTest +2 cases; new ProjectMemoryServiceReuseTest
covers same-key skip, changed-key re-index, no-observation always-reindex,
and embedding turnId tag verification.
After a freestyle plan compiles successfully, requestPlanApproval is invoked
before ExecutionPlanLockedEvent is emitted. Rejection emits
ExecutionPlanRejectedEvent(source="operator") with no lock and no phase-2 run.
Adds public SessionOrchestrator.requestPlanApproval seam delegating to
requestStageApproval, wired in Main.kt. Two new tests cover the reject and
approve paths; existing tests stay green via the default passthrough param.
Both silent-failure branches in FreestyleDriver.lockAndRun now emit
ExecutionPlanRejectedEvent (source=compile or missing_content) before
returning, satisfying invariant #1. Event registered in eventModule for
correct polymorphic serialization.
emitLlmArtifacts previously emitted ArtifactCreated/Validating/Validated for every
llm-emitted slot unconditionally, fabricating "validated" artifacts when the model
produced no content. Gate now mirrors the F-015 file_written check: slots with null
or blank cached content are skipped with a warn log. Regression test added in
NeedsArtifactInjectionTest verifies no artifact events appear for a blank-response stage.
[router.narration] model_id selects which provider handles narration turns
instead of generic capability routing — lets a small/fast model own narration
while the main model keeps CHAT/STEERING.
- NarrationSettings.modelId parsed from TOML, threaded via RouterConfig
.narrationModelId into DefaultRouterFacade.narrate (3-arg route)
- DefaultInferenceRouter: exact-id match against healthy providers, with a
post-select health check; any miss logs WARN and falls back to capability
routing (never fails the narration turn)
- onUserInput unchanged — only narrate uses the pinned model (tested)
Follow-ups spotted while fixing the artifact-content bug:
- #1 robustness: listArtifacts resolves content via runCatching, degrading to null
instead of failing the whole listing on one bad CAS read (resolveContent helper)
- #2 eviction visibility: WARN when a referenced artifact hash resolves to null
(most likely evicted/compacted while still referenced by the event log)
- #3 ordering trap: document on ArtifactCreated/ArtifactContentStored that 'Created'
fires at validation, AFTER ContentStored at inference — the latent fold hazard
- #4 reopen gap: SessionSnapshot now carries workspaceRoot (derived from
SessionWorkspaceBound), so a reopened session restores its workspace label
- #5 orchestrator headroom: isBackEdge externalised to a top-level function so
DefaultSessionOrchestrator drops below the TooManyFunctions threshold (was at it)
#6 (orphaned empty F-010-era artifacts) needs no code change — empty content is
already normalised to null ('(no content stored)'); emission is prevented going
forward by the shipped F-010/F-018/F-020 fixes.
Implements the replay-safe half of preemptive redirect (graph re-routing). An
operator-confirmed PreemptRedirectEvent overrides the resolver's edge at the next
transition boundary; the pure resolver stays untouched.
- PreemptRedirectEvent / PreemptRedirectBlockedEvent (+ serialization registration)
- TransitionExecutedEvent + TransitionDecision.Move gain optional redirectId — the
durable 'consumed once' marker
- PreemptRedirect.decide: pure decision over the event log (override / block / none),
so replay reaches the identical outcome without re-classifying (invariant #8)
- override wired in DefaultSessionOrchestrator.step above resolveTransition; back-edge
jumps count against the existing maxRetries cap via executeMove
- blocks jumps to unknown stages or targets with unsatisfied needs
- DomainEventMapper: redirect events mapped to null (operator surface ships with the
LLM-proposal + approval-confirm front-half)
- tests: override+consume-once, block-on-needs, block-on-unknown, ignore-consumed
Front-half (LLM proposal -> approval-gate confirm -> emit PreemptRedirectEvent) is
deferred; needs live LLM verification. Until it ships no redirect event is emitted in
production, so the override is inert. Spec: docs/plans/2026-06-10-freestyle-graph-rerouting.md
- generic SystemResourceProbe: owner-agnostic /proc/meminfo system RAM
overlaid on the vendor GPU probe, wired on both managed and static paths
so the VRAM/GPU/RAM gauge renders with an external llama-server (was dormant)
- F-002: classify provider 4xx (e.g. llama.cpp 400) as terminal, not retryable
(except 408/429); stops retrying deterministic request failures to exhaustion
- F-003: FileSystemPromptLoader expands leading ~ / ~/ to user home
- F-004: map InferenceFailedEvent + RetryAttemptedEvent to new
ServerMessage.InferenceFailed / RetryAttempted, decoded+rendered in tui-go;
ArtifactContentStoredEvent explicitly mapped to null (internal, no warn)
- tests: tilde expansion, SystemResourceProbe (static/managed/fail-soft)
Two operator features over the existing WebSocket protocol, plus a tui-go nav fix.
Config editor (g / palette "config"):
- core:config gains a thin registry stack: OrchestrationKnobs ([orchestration]
block), CorrexConfigWriter (round-trip TOML serializer; parseToml(write(c))==c),
ConfigHolder (AtomicReference live config), and EditableConfig (allowlist of
editable fields + patch applier; security fields are absent so they can never
be patched).
- ConfigService validates -> persists (atomic temp+move) -> swaps the holder ->
rebuilds config-derived services. Live-apply is scoped to safe seams: stage
timeout, compaction threshold (now a supplier), router knobs (routerFacade
rebuild), and personalization/project toggles all take effect on the next
session/turn; in-flight sessions keep the config they started with.
server.host/port are persisted + badged "restart required", not hot-swapped.
- Protocol: GetConfig / UpdateConfig / ConfigSnapshot; TUI overlay with
type-aware editing (bool toggle, enum cycle, inline int/string edit) and save.
Artifact viewer (v / palette "artifacts"):
- Server assembles a session's artifacts from the event log and resolves bytes
from the CAS store (StreamQueries.listArtifacts) — a replay-neutral snapshot
read. TUI overlay lists artifacts and shows scrollable content.
tui-go fix: workflow failure no longer clears selectedID (which yanked the user
to the list); listNav lands on the first session when nothing is selected
instead of wrapping to a random one.
Config is not event-sourced (it lives in TOML); editing stays outside the log.
rehydrate() now scans ArtifactContentStoredEvent (the durable slot->CAS-hash
bridge) and fetches stored bytes by content hash, keying the cache by the
logical slot name. The slot name is never passed to the hash-keyed CAS, so the
odd-length-hex crash is gone and artifact content is restored on cold-start
resume. RehydrateTest reworked to use a content-addressed fake store (slot name
!= content hash) so it actually exercises the bug. Replay (#8) is unaffected —
rehydrate is live-resume-only.
Also bump router narration maxTokens 1024 -> 4096: reasoning models were
spending the whole budget thinking and emitting empty content (finishReason=
length), so narrations never reached the TUI.
Removes the stale TUI-refactor progress doc.
Live QA audit of role_pipeline against a sample repo surfaced and fixed a
chain of defects that prevented any tool+artifact workflow from running
end-to-end. The pipeline now completes cleanly with a real, validated,
reviewed file change.
- F-008 workflow: propagate stage token_budget -> generationConfig.maxTokens
(TomlWorkflowLoader) so large stages stop truncating at the 2048 default.
- F-009 tools/kernel: file_read missing-file is recoverable; recoverable tool
errors feed back into the loop instead of aborting the stage.
- F-010/F-018 kernel: reject premature stage_complete and nudge the model to
emit the owed artifact / invoke a write tool (bounded by MAX_TOOL_ROUNDS).
- F-011 schemas: list-typed artifact fields string -> array (analysis/design/
impl_plan) to match model output.
- F-012 kernel: strip an outer markdown code fence from LLM artifacts.
- F-013 validation: payload-validation failures are retryable; structural
failures stay terminal (wires the invariant-#7 reject+retry contract).
- F-014 tools: file_edit schema matches its real params; edits emit a diff.
- F-015 kernel: record tool-execution events + materialise a real file_written
artifact from the on-disk write; gate validation so a no-write stage cannot
be rubber-stamped (no more false-success runs).
- F-016 server: raise stageTimeoutMs 60s -> 180s for slow local models.
- F-019 artifacts/kernel: file_written artifact carries the unified diff so the
reviewer can verify the change (shared DiffUtil; file_write emits a diff too).
- F-020 tools: model-correctable file_edit failures are recoverable + steer
toward replace/file_write over the fragile patch op.
Full findings register (F-001..F-021, incl. open F-021 resume rehydrate bug)
in qa/audit-report-2026-06-08.md. All module tests green.
ArtifactPayloadValidator passed the logical slot name into the
content-hash-keyed CAS store, crashing every workflow with a typed
produces slot. Validator now reads payloads from a content map threaded
through ValidationContext (populated from the orchestrator's per-session
artifact cache across all stages) and no longer depends on ArtifactStore.
Also records the slot->CAS-hash mapping as a new ArtifactContentStoredEvent
(registered for serialization), emitted at both CAS write sites, so
artifact content is recoverable from CAS by hash after a cold start.
kami