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.
Records operator profile as a snapshot event at session start (invariants #8/#9),
populates SessionState.boundProfile via reducer, and injects the about text as a
pinned L0 SYSTEM context entry in every stage. Gated by PersonalizationConfig.enabled.
Extract a SubagentRunner fun-interface (SubagentRunRequest/SubagentRunResult)
and an InSessionSubagentRunner default that delegates to the existing
executeStage path, with per-run effectives captured in the lambda closure so
the seam stays free of the internal RunEffectives type.
DefaultSessionOrchestrator.enterStage now dispatches through subagentRunner.run
instead of calling executeStage inline; effectives threading is dropped from
run/step/executeMove/enterStage/resume. ReplayOrchestrator supplies its own
runner. Behaviour is identical — RefinementLoopTest and full check stay green.
Stages declaring needs=[...] now receive each needed artifact's validated
JSON (from artifactContentCache) as an L1/USER neededArtifact context entry,
so a one-shot subagent sees its inputs.
Also emit llm-emitted artifact lifecycle events (Created/Validating/Validated)
on successful validation via emitLlmArtifacts, closing a latent gap where
verifyProduces would fail for stages producing only an llm-emitted artifact.
Events are emitted only after validation passes, preserving invariant #7.
Two issues surfaced during live QA once workflows began progressing past the
first stage:
1. Narration prompts went out with no user message, so the chat template
rejected them ("No user query found in messages"). buildNarrationContext
filed the trigger instruction (a user turn) under ContextLayer.L0, and
PromptRenderer folds every L0 entry into the system message regardless of
role. Move the trigger to L1 so it renders as the user turn.
2. TransitionExecutedEvent was emitted after the next stage had already run:
executeMove called enterStage (which executes the stage) before advanceStage
(which emits the transition), so the event log showed a stage running before
the event marking entry into it. Emit the transition before entering.
Adds a regression test asserting a rendered narration prompt carries a user
message.
Tool-calling loops were rendered as all-assistant-calls-then-all-tool-results
with the original task last, causing the model to re-issue the same tool call
indefinitely. Two stacked reorderings caused it:
- DefaultContextPackBuilder grouped entries by sourceType for per-type
compression, discarding a,t,a,t interleaving.
- PromptRenderer forced L1 (the live user turn) to render last, pushing the
task after the entire transcript.
- SessionOrchestrator's tool loop re-fed currentContext.layers.values.flatten()
(grouped by layer) each round, compounding the scramble.
Add a chronological `ordinal` to ContextEntry, stamped by the builder from
input order and restored after grouping/compression (the compressor preserves
entry identity, so ordinals survive). PromptRenderer now orders non-system
messages by ordinal, with the old L1-last layer priority kept only as a
tiebreak so router chat (ordinal 0) is unchanged. The orchestrator keeps a
running ordered accumulator instead of reading back the grouped pack.
Adds builder + renderer ordering regression tests.
Wires the workspace handshake end to end so a session's workspace is bound
from the client's cwd at connect time and is event-sourced for replay.
- Go TUI sends Hello{workingDir=os.Getwd()} as the first WS frame on connect
(protocol.go encoder + client.go connect path; golden test pins the wire
format against the Kotlin discriminator).
- Server adds ClientMessage.Hello, stashes the per-connection workingDir, and
on session start resolves it through WorkspaceResolver's trust pipeline,
emits SessionWorkspaceBoundEvent (invariant #9), and threads the resolved
workspace into OrchestrationConfig for the live run. A Hello after the first
StartSession is ignored (warn); a rejected path binds the resolver fallback.
- Replay derives the workspace from the recorded event: SessionState gains
boundWorkspace, DefaultSessionReducer fills it from SessionWorkspaceBoundEvent,
and ReplayOrchestrator uses it (Path.of only, no filesystem re-query —
invariant #8) with graceful fallback to config for pre-Phase-B logs.
Absent Hello / null resolver degrades to the prior config-workspace behavior.
A stage's allowedTools only shaped which tool definitions were offered to
the model; the dispatch path resolved returned tool calls straight from the
full registry and executed them without checking stage membership. A
hallucinated, jailbroken, or edited-transcript tool call for any registered
tool would run unchecked (violating invariant #5 / policy-is-absolute).
dispatchToolCalls now rejects any tool call whose name is not in the stage's
allowedTools (STAGE_COMPLETE_TOOL exempt), emitting ToolExecutionRejectedEvent
and feeding an error ContextEntry back to the model instead of executing.
Empty allowedTools means deny-all domain tools, consistent with offering
only STAGE_COMPLETE_TOOL at inference time.
A steering note attached to a tool approval was recorded but only
consumed by the NEXT stage's prompt build (buildSteeringNoteEntries runs
once, before the tool loop), so on a single-stage task it had no visible
effect. Two changes:
- Approve+note: capture userDecision.userSteering at the gate and inject
it as a context entry right after the tool result, so the same stage's
loop re-infers with the note and the model acts on it.
- Reject: buildSteeringNoteEntries now also emits anti-repeat feedback for
REJECTED decisions with no note, so the model does not re-propose the
identical call on the retryable re-run (a bare reject previously fed
back only "approval denied").
Integration test drives an approval with a steering note and asserts the
next inference's context carries it.
Compute the effective tool registry/executor and plane-2 WorkspacePolicy per run
from config.workspace (effectivesFor), threaded through the orchestrators; a null
workspace falls back to the boot instances byte-for-byte. Wire the concrete
WorkspaceToolRegistryProvider in Main (buildToolConfigForWorkspace). Session undo
unions the session's recorded workspace into its jail roots. (Axis 2 Phase A, tasks 3 + 7.)
Add a WorkspaceContext value type and a nullable OrchestrationConfig.workspace,
plus SessionWorkspaceBoundEvent registered in eventModule. Purely additive; no
behavior change when workspace is null. (Axis 2 Phase A, tasks 1-2.)
A stage declaring no `produces` passed verifyProduces vacuously, letting an
agent stage_complete with zero work. Gate on allowedTools: a produce-less stage
with tools available must have >=1 ToolInvocationRequestedEvent scoped to its
stageId, else Failure(retryable). Produce-less + no tools is exempt (only
stage_complete is ever offered), so pure-reasoning stages aren't bricked.
DefaultContextPackBuilder already enforced the token budget, but SessionOrchestrator discarded the entriesDropped signal at both build sites — overflow was enforced yet unobserved. Emit ContextTruncatedEvent when entriesDropped>0, matching the router path (open-threads 6.2 contract).
- LlamaCppInferenceProvider: check HTTP status and surface the real error body instead of masking it as a ChatCompletionResponse deserialization failure
- PromptRenderer: render the live conversation layer (L1) last so the user turn is the final message (fixes Qwen 'No user query found' 500 when L3 recall is present)
- TomlWorkflowLoader: resolve stage prompts relative to the workflow file's own dir (bundle), CWD-independent; config-dir fallback for globals
- SessionOrchestrator: hard-fail a stage whose declared prompt can't resolve (was a silent user-less request)
- move healthcheck prompts next to the example workflow (examples/workflows/prompts/)
StageConfig.modelId; InferenceRouter gains a backward-compatible model-aware
route() overload (default ignores modelId, so static routers and test fakes are
unaffected). New ManagedInferenceRouter (infra commons) resolves a target model
per stage (stage.modelId > capability match > default) and ensureLoaded()s it
before routing, returning the live managed provider — its id changes with the
resident model, so there is no stale health cache to invalidate on swap.
ModelManager.ensureLoaded default delegates to the idempotent load(). Main wires
the managed router on the [[models]] path; the static path keeps DefaultInferenceRouter.
Plan: docs/plans/2026-05-31-model-lifecycle-management.md (slice 2 of 5).
Tools can declare how their output is compressed before it enters the
context pack, replacing uniform handling. Open-threads 6.3 / ADR-0003.
- core:tools/compression: ToolOutputCompressor interface,
IdentityCompressor default, declarative OutputCompressionSpec +
CompressionRule (StripBlankLines, StripLeadingWhitespace,
DropMatching, HeadTail), and the pure DeclarativeCompressor engine
(regexes compiled at construction; bypassOnError passes raw on
non-zero exit; never throws on input).
- Tool gains `outputCompressor` (default IdentityCompressor). FileReadTool
strips blanks + leading whitespace; ShellTool strips blanks + head/tail.
- SessionOrchestrator applies the resolved tool's compressor on the
ToolResult -> ContextEntry path only. Raw output stays authoritative
in the event log (invariant #6); compression is a pure function of
(raw, exitCode, spec), so no new event and replay stays deterministic.
- Spec is @Serializable/config-shaped (JSON round-trip tested). No
custom-tool-from-config path exists today, so first-party tools wire
in code; TOML-declared compression rides along when that path lands.
Tier B (format-aware parsers: failures-only test output, git-log oneline)
is a documented per-tool extension behind the interface, not built here.
Add ParamRole and a defaulted Tool.paramRoles to the tool contract; built-in
tools declare which params carry effects: file_read/write/edit -> path=PATH,
shell -> argv=EXEC_COMMAND. PathContainmentRule / ExecInterpreterRule /
NetworkHostRule now inspect only the declared params when roles are present, and
fall back to the looksLikePath / leading-token / extractHost heuristic only for
un-enriched (e.g. custom) tools - so a custom FILE_WRITE tool is still
containment-checked. Shared extractParamStrings flattens String and List<*>
(argv) values. SessionOrchestrator.runPlane2Assessment resolves the tool once
and passes tool.paramRoles into the assessment input.
This retires the arg-guessing from the first-party path: a slashy edit 'content'
arg is no longer mis-flagged as a path.
Implements the full vertical slice for invariant #9 tool-call assessment:
- core:toolintent — new module with ToolCallRule seam, ToolCallAssessor,
WorldProbe/FileSystemWorldProbe, WorkspacePolicy, PathContainmentRule,
and RiskMapping (assessment → RiskSummary / AssessedIssue)
- core:tools — ToolCallAssessmentRecord + ToolInvocationRecord.assessment
field; DefaultToolReducer handles ToolCallAssessedEvent (replay proof)
- core:config — ToolsConfig gains workspaceRoot + privilegedLocations;
ConfigLoader parses both; DEFAULT_PRIVILEGED_LOCATIONS built-in
- core:kernel — OrchestratorEngines gains toolCallAssessor/workspacePolicy/
worldProbe fields; SessionOrchestrator.dispatchToolCalls runs
runPlane2Assessment before the tier gate: BLOCK → hard-reject without
executing; PROMPT_USER → elevates tier into approval path with plane2Risk
in the ApprovalRequestedEvent
- apps/server — constructs PathContainmentRule + ToolCallAssessor +
WorkspacePolicy from config and wires them into OrchestratorEngines
Assessment is recorded as ToolCallAssessedEvent (environment observed once,
facts stored, replay reads events — invariant #9). Assessor and WorldProbe
are only invoked on the live orchestrator path, never in replay.
Carry the full RiskSummary (level, signals, rationale) inline on
ApprovalRequestedEvent instead of discarding it after assessment. The two
server-side DTO builders (live ApprovalCoordinator path and replay
DomainEventMapper path) now render real risk level, recommended action,
signal factors, and the [code] rationale lines into the approval WebSocket
message, with a consistent enum-name fallback when no assessment ran (tool
path). RiskSummary/RiskSignal made @Serializable for event embedding;
shared RiskSummary.toDto() lives next to RiskSummaryDto.
Flags when a workflow stage's allowedTools references a tool name that
is not registered in the tool registry (config drift), surfaced as a
non-blocking WARNING.
- Add ValidationContext.availableTools (nullable; null = registry
unavailable, skip the check)
- MissingToolRule emits one MISSING_TOOL warning per (stage, tool) pair,
deterministically ordered
- Populate availableTools from the ToolRegistry at the orchestrator
construction site; wire the rule into the prod SemanticValidator
Extract message-rendering logic from LlamaCppInferenceProvider into a
provider-agnostic PromptRenderer in core:inference. SessionOrchestrator
now serializes the rendered ChatMessage list as JSON for the prompt
artifact, so the audit trail reflects what the LLM actually saw.
Resolves the TODO at SessionOrchestrator.kt:691.
- Add STAGE_COMPLETE_TOOL constant and meta-tool handler in executeStage
loop: when LLM calls stage_complete, break out of tool dispatch loop
and proceed to normal validation/transition
- Inject stage_complete tool definition in every inference request so
the LLM always has the option to signal stage completion
- Replace expand-all diff with scrollable diff window in approval
surface (ctrl+x toggles, up/down scrolls, shows line range)
P4-1: Wire steering to actually work (advisory-only):
- RouterFacade emits SteeringNoteAddedEvent via event store in STEERING mode
- SessionOrchestrator reads steering notes from event store and includes them
as ContextEntry objects (sourceType=steeringNote) in stage context packs
- Delete dead SteeringNote data class; single vocabulary = SteeringNoteAddedEvent
- Add optional validateSteering callback to DefaultRouterFacade for future
ValidationPipeline wiring (avoids cross-core dependency)
- Update RouterFacadeTest to assert event emission
P4-2: Router FACE — move hardcoded generation config:
- Add GenerationConfig field to RouterConfig with same defaults
- RouterFacade uses config.generationConfig instead of inline literals
- (Remaining P4-2 items: conversation persistence and coordination
semantics are design decisions requiring ADRs, deferred)
P1-1: Replace check-then-act with computeIfAbsent for activeSessionJobs to
prevent double-launch. Register pendingApprovals[requestId] BEFORE emitting
ApprovalRequestedEvent to close the approve-before-register window.
P1-2: ToolCompleted with a diff now appends msg.diff to routerMessages
instead of the static "--- diff from $toolName ---" marker.
P1-3: Change routerMessages from List<String> to Map<String, List<String>>
keyed by sessionId.value. RootReducer appends to the correct session's list;
RouterPanel renders only routerMessages[selectedId].
P1-4: On IDLE chat submit, SessionsReducer creates an optimistic
SessionSummary(status="STARTING") and sets selectedId.
processSessionStartedMessage removes any STARTING session and inserts the
real one on echo. RootReducer sets sessionEntered=true for the optimistic
session.
Tests: LaunchRegistrationRaceTest, ApprovalRegisterBeforeEmitTest,
RootReducerTest (diff/per-session/isolation/optimistic-submit),
SessionsReducerTest (optimistic/reconciliation). All P0/P1 tests pass.
./gradlew check green.
- Fix duplicate inference requests after approval resume by registering orchestrator
jobs in activeSessionJobs (ServerModule.launchSessionRun), so the guard in the
OrchestrationResumedEvent subscription prevents spurious duplicate resume.
- Replace blocking Files.readString in computeToolPreview with withContext(Dispatchers.IO)
by making the function suspend — no blocking I/O on the coroutine dispatcher.
- Guard orderedIds.add() in rebuildTools against duplicate invocation IDs on replayed
ToolInvocationRequestedEvent to prevent duplicate tool records in snapshots.
- Add unified-diff preview for file_write tools: computeToolPreview reads existing
file and generates ---/+++ diff before emitting ApprovalRequestedEvent.
- Render diff preview with color coding in ApprovalSurface (@@ yellow, +/- green/red).
- Include current-stage tool records in SessionSnapshot via rebuildTools event replay.
- Fix RouterContextBuilderTest: recursive buildPack helper and 2 tests using class-level
builder instead of their local builder instances (conversationKeepLast mismatch).
- Add suspend keyword to RouterFacadeTest mock, fix buildPack recursion.
- GlobalStreamHandler.handleCreateGrant validates scope (SESSION/STAGE)
and sends ProtocolError to client on validation failure instead of
silent log.warn + drop.
- Derive event stageId directly from GrantScope type, removing
redundant stageIdForEvent tuple.
- SessionOrchestrator integrates ApprovalEngine to check active grants
before requesting user approval for T2+ tool calls.
- ContextEntry gains EntryRole (SYSTEM/USER/ASSISTANT/TOOL) field for
proper chat message role mapping.
- RouterContextBuilder.build is now suspend; uses Tokenizer for
accurate token estimation with fallback to character-based estimate.
- LlamaCppInferenceProvider maps EntryRole to ChatMessage role instead
of heuristic layer-based inference.
Three coordinated changes:
1. DefaultOrchestrationReducer now handles TransitionExecutedEvent,
updating currentStageId so the projected state always reflects the
actual current stage rather than staying stuck at the start stage.
2. DefaultSessionOrchestrator.resume() re-enters the execution loop at
the current stage without re-emitting WorkflowStartedEvent. It reads
currentStageId from the projected state, creates an ExecutionContext
there, and calls enterStage → step exactly as run() does after its
first stage.
3. ServerModule.start() calls resumeAbandonedSessions(), which scans all
RUNNING/PAUSED sessions on startup and launches resume() for each one
whose orchestrator is no longer in memory. Sessions that have a live
deferred are unaffected; only sessions with no coroutine (e.g. the
server was restarted while a tool was executing or approval was pending)
are picked up and re-executed from their current stage.
When a TUI client connects at the exact moment between an
OrchestrationPausedEvent being stored and its corresponding
ApprovalRequestedEvent being stored, the approval projector sees empty
pendingApprovalRequests while orchState.pendingApproval=true. The
stuck-session fix would then append a spurious OrchestrationResumedEvent,
clearing the TUI's pendingApproval display and hiding the approval dialog.
Fix: count OrchestrationPausedEvents vs ApprovalRequestedEvents in the
event log. If there is an unpaired pause (more pauses than requests),
the session just entered the gate — skip the fix entirely.
Also remove spurious ToolExecutionCompleted/Failed emissions from the
kernel orchestrator; those events are emitted by SandboxedToolExecutor
in the infrastructure layer. Keep ToolExecutionRejectedEvent which the
orchestrator does own (rejection happens before executor.execute is
called).
ToolInvocationRequestedEvent was emitted but the corresponding outcome
events were never emitted, so the TUI permanently showed tools as
RUNNING. Now the orchestrator closes every tool invocation with the
appropriate event:
- ToolExecutionCompletedEvent (Success path)
- ToolExecutionFailedEvent (Failure path)
- ToolExecutionRejectedEvent (approval denied path)
Sessions that had an approval resolved (or never needed one) but never
received OrchestrationResumedEvent would show the last tool as RUNNING
forever with no subsequent events. Three coordinated fixes:
1. SessionOrchestrator now emits OrchestrationResumedEvent in both the
approved and rejected approval branches, so the paused flag clears
correctly going forward.
2. SessionEventBridge.replaySnapshot() detects the stuck pattern
(pendingApproval=true with no unresolved requests) and appends a
synthetic OrchestrationResumedEvent so historical sessions self-heal
on next TUI connect.
3. DomainEventMapper maps OrchestrationResumedEvent → SessionResumed;
TUI SessionsReducer/SnapshotPhaseReducer handle the new message,
clearing pendingApproval and setting status back to ACTIVE.
Also adds server-restart recovery path in DefaultSessionOrchestrator
(emit decision + resume events directly when no live deferred exists)
and pre-registers pending approvals in ServerModule.start() to close
the race between reconnect and ApprovalResponse routing.
Move ProcessResult artifact content storage into dispatchToolCalls so it
fires for all outcomes (success, ERROR, FATAL) — the spec mandates the
artifact MUST always be produced.
- dispatchToolCalls now stores ProcessResult JSON in CAS + cache for
every tool call result, using the current toolCall's function.arguments
as the command field (not firstOrNull)
- Replace storeFailedProcessResult with emitProcessResultEvents — event
emission only, content already stored by dispatchToolCalls
- FATAL and ERROR branches both now emit artifact lifecycle events,
fixing the gap where retryable ToolResult.Failure produced no artifact
- Add field to ProcessResultArtifact ("success"/"failed")
- Materialize ToolResult.Failure as a failed ProcessResult artifact
in the orchestrator, with artifact lifecycle events emitted
- Non-recoverable ToolResult.Failure now fails the stage ("FATAL:")
instead of being silently fed back as LLM context
- Fix enterStage to return Continue on non-retryable failures,
enabling back-edge transitions (fix/retry loops)
- Add ArtifactFieldEquals transition condition with flat and
nested dot-notation field access + FieldOperator enum
(eq, neq, gt, gte, lt, lte)
- Wire artifact_field_equals through ConditionFactory and
TomlWorkflowLoader (condition_field, condition_operator)
- Cache ProcessResult artifact content for evaluation context
- Add 16 tests covering all condition operators and error cases
- DefaultTransitionResolver catches condition evaluation errors
and returns Blocked instead of crashing
SessionOrchestratorIntegrationTest: wire approvalRepository into
OrchestratorRepositories constructor which gained the field in the kernel
refactor. TambouiKeyMapperTest: correct ctrl-h to alt-h — the mapper has
always placed ToggleApprovalOverlay under the Alt branch, not Ctrl.
SessionOrchestrator.emit: replace substring(0,7) with take(7) to avoid
StringIndexOutOfBoundsException when session IDs are shorter than 7 chars.
- fixed the tool overlay in tui.
- fixed tool calling - added schemas.
- getting all sessionIds via event store.
- instead of sending all events on the replay - there's a new way for replaying.
- session snapshot is now a thing getting sent for previously created sessions.
- added logging to important parts.
- revert workflowId from WorkflowCompletedEvent.
fixed detekt issues where possible.
fixed disttar failing build because tools is added twice in the server module.
added workflowId where required.
fixed some tests not being recognized because of runBlocking without explicit return type.
formatting + imports.
kami