[[models]] config (ModelConfig/ModelsSettings) parsed by ConfigLoader;
InfrastructureModule.createModelManager + modelConfigToDescriptor; Main.kt
managed boot path spawns the default llama-server when [[models]] is present
(static [[providers]] path preserved when absent) and kills it on shutdown.
Plan: docs/plans/2026-05-31-model-lifecycle-management.md (slice 1 of 5).
TurboVec persists vectors via the sidecar but kept entry metadata
(sessionId/turnId/text) only in an in-memory map, so restarting with
backend=turbovec silently dropped every query hit (the id→metadata
lookup missed). Rebuild that map from the ChatTurnEvent log at startup.
- RehydratableL3MemoryStore: capability interface for stores whose
vectors persist independently of the JVM. TurboVec implements it;
InMemory deliberately does not (its vectors are volatile too).
- L3MetadataRehydrator replays ChatTurnEvents into the metadata map.
No embedder: the query path never reads entry.vector, so vectors
stay in the sidecar and metadata comes from the event log (the
source of truth, invariant #1). Short-circuits for non-rehydratable
backends before scanning the log.
- Wired into Main before the server accepts queries; no sidecar spawn.
Backfill of never-embedded history (needs the embedder) is a separate
follow-up. The in_memory non-durability warning already exists.
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.
Final plane-2 step: the tool-call intent assessment was decided
server-side but never reached the client — DomainEventMapper dropped
ToolCallAssessedEvent through the else/null branch.
- ServerMessage.ToolAssessed (tool.assessed): disposition as a plain
String + AssessedIssueDto list; observations stay off the wire
(internal replay facts, invariant #9).
- DomainEventMapper maps the event 1:1 (no filtering — rendering
decisions belong in the client).
- Go TUI consumes tool.assessed, records an event entry only for
non-PROCEED or issue-bearing assessments (noise control), and is
added to IsEventBearing() so it buffers correctly during snapshot
replay like its sibling tool.* events.
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.
DefaultMaterializingArtifactWriter and its MaterializingArtifactWriter interface
were test-only orphans after the sandbox dual-write retirement (95e79bf) - no
production wiring, never instantiated outside their own test. Delete all three.
Add healthcheck_*.txt to .gitignore (one-off generated output).
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.
Remove the wall-clock approval timeout that auto-rejected pending approvals.
The timeout modeled machine latency, but approvals are human latency
(unbounded); it also created an intent/outcome divergence where a human
approving at the same instant the timeout fired was silently overridden by
the auto-reject. Approvals now block until the operator decides; the
resolved-request dedup still guards against double-submit.
Add ServerMessage.ApprovalResolved and map ApprovalDecisionResolvedEvent to
it in DomainEventMapper so clients are notified when any approval is resolved
(by anyone) — letting a second connected client clear its prompt. This is the
event-sourced replacement for the removed timeout notification path.
Drop the now-dead ApprovalConfig (timeout_ms) and its loader/test/sample-config
references. ApprovalStatus.TIMED_OUT is retained for replay of historical events.
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.
Record L3 retrieval as an event carrying hit text (invariant #9), then
rebuild router state and inject recalled memories as a SYSTEM L3 layer in
the context pack. Apply token budget: protected frames (L0 immutable +
current user turn) are never dropped; honest budgetUsed is reported and
'BudgetExceeded' is flagged in appliedStrategies when they overflow. L1/L2
fit newest-first so oldest entries drop first. Emit ContextTruncatedEvent
when entries are dropped. L3MemoryRetrievedEvent is emitted on every CHAT
turn (empty hits reset recalled memory).
When a user sends input, embed it, query L3 across all sessions, dedup
against in-session turns, and record the retrieval as an event so replay
is deterministic (invariant #9). Hits land in RouterState; context
injection follows in a later slice.
- Add L3MemoryRetrievedEvent + L3RetrievedHit (registered in eventModule)
- RouterState.lastRetrievedMemory + reducer case; RouterTurn carries turnId
- RouterConfig.retrievalK (default 5)
- Harden L3 write path: runCatching + visible logging, cancellation
re-thrown; event append stays ahead of the L3 write
- Warn prominently when the non-durable in_memory L3 backend is selected
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
Risk tier is now derived from the worst validation issue severity and
the issues that drove it travel with the summary, so consumers can show
operators why a tier was assigned instead of an opaque level.
- Add RiskSummary.rationale (defaulted, backward-compatible) holding
human-readable '[code] message' lines from the validation report
- Pin severity -> risk mapping: ERROR->HIGH, WARNING->MEDIUM, INFO->LOW
- DefaultRiskAssessor folds the validation-driven level into the summary
Wire SemanticValidator into the production validator pipeline and
consolidate cycle-policy types under core:validation.
- Add SemanticValidator(CycleExitRule(requirePolicyForCycles=false))
to the prod ValidationPipeline (no-op default, preserves behavior)
- Move CyclePolicy/Binding/Signature/Factory from core:transitions.policy
to core:validation.policy (validation already owns ValidationContext)
- Rename CyclePolicyBindingRule -> CycleExitRule (issue code unchanged)
- Delete dead CyclePolicyResolver + PolicyValidation
Adds [router.embedder] and [router.l3] sections to CorrexConfig with
backend selectors. Ships LlamaCppEmbedder that hits llama.cpp's
/embedding endpoint (handles OpenAI-compatible, simple, and array
response shapes; validates dimension). InfrastructureModule gains
createEmbedderFromConfig and createL3MemoryStoreFromConfig that
dispatch on backend value.
Defaults preserve current behavior (noop embedder + in-memory L3).
Switching to "llamacpp" / "turbovec" is a config-only change — no code
edits required. For turbovec backend, the bundled python sidecar
script is extracted from classpath to ~/.cache/correx/ on first use.
Adds ProviderConfig + providers list and per-tool enabled flags to
CorrexConfig, and rewires apps/server/Main to build the provider list
from config instead of the hardcoded buildLlamaProvider() factory. Env
vars (CORREX_MODEL_ID, CORREX_MODEL_PATH, CORREX_LLAMA_URL) remain a
fallback only when no providers are declared in config.
Completes slice A of the config layer alongside commit 0834c70 which
already shipped the TOML parser upgrade and sample config — those
parser changes referenced these types but were committed prematurely
in isolation; this commit makes HEAD buildable.
- Add support for nested table headers ([tools.shell], [tools.file_read], etc.)
- Add support for inline tables (capabilities = { General = 1.0, Coding = 0.7 })
- Add support for JSON-style arrays (shell_allowed_executables = ["bash", "sh"])
- Maintain backward compatibility with old CSV formats (deprecated, logged as warnings)
- Refactor buildConfig() to handle new value types (Map, List, Boolean, Number)
- Update docs/sample-config.toml to use clean TOML shapes
- Add comprehensive tests for inline tables, nested sections, JSON arrays, and fallback parsing
- Remove ugly workarounds: capabilities no longer a CSV string, tool flags now in proper nested sections
Introduces Embedder contract in core:inference with a NoopEmbedder
default (zero vector) and wires both Embedder and L3MemoryStore into
DefaultRouterFacade. Each ChatTurnEvent emission now embeds the turn
content and writes a corresponding L3MemoryEntry, so cross-session
memory is captured at the source of truth (the event emission point).
InfrastructureModule defaults to NoopEmbedder + InMemoryL3MemoryStore
so the system runs without an external embedding model or vector store
wired. The TurboVec adapter stays unwired for now; switching to it is
a configuration concern (Epic 12 config layer).
Also gitignores apps/tui/logs/ (Kotlin TUI runtime logs, pre-Go rewrite).
Adds ChatTurnEvent (role USER|ROUTER) so router conversation history
survives restart and is rebuilt from the event log via projection,
restoring Hard Invariant #1 for the CHAT path. Both CHAT and STEERING
now emit user + router turn events; STEERING additionally keeps its
existing SteeringNoteAddedEvent so the structured directive record is
preserved alongside the conversation turns.
RouterFacade drops the in-memory ConcurrentHashMap of histories and
rebuilds state through the repository after each emission. Reducer
appends turns to conversationHistory. ChatTurnEvent is registered in
eventModule.
Defines the L3MemoryStore contract in core:router (entry, query, hit,
in-memory test default) and a new infrastructure:router:turbovec module
that adapts turbovec via a Python sidecar over stdin/stdout JSON-Lines.
Adapter manages the subprocess via ProcessBuilder with lazy start,
mutex-serialized requests, and shutdown via withTimeout. Wire-format
metadata that turbovec doesn't store (sessionId, turnId, text) is kept
in an in-memory map for now — durable persistence ships with the
upcoming ChatTurnEvent.
Not yet wired into RouterFacade, InfrastructureModule, or DI. Embedding
generation source is a separate concern.
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.
Replace raw string routerMessages with structured RouterEntry(role, content)
entries. User messages are stored locally on SubmitInput (both IDLE and
IN_SESSION). Router responses arrive as 'router' entries via
RouterResponseMessage. Tool diffs arrive as 'tool' entries.
RouterPanel renders as a styled conversation log with role prefixes:
▸ user msgs in accent, router replies in strong body, tool output in dim.
Multi-line content (diffs, long responses) is split on \n for proper display.
Panel title changed from 'router' to 'output' to reflect the log role.
LlamaCppInferenceProvider was created with empty capabilities, causing
NoEligibleProviderException when the router requested ModelCapability.General.
FirstAvailableRoutingStrategy requires declared capabilities to cover all
requested ones — empty set satisfied nothing.
Add a DEFAULT_LLAMA_CAPABILITIES constant with General, Coding, Reasoning,
Summarization, and ToolCalling scores, and pass it through createLlamaCppProvider()
to ModelDescriptor. Existing callers get sensible defaults automatically.
This is the root cause of the 'router not connected — epic 14' placeholder:
the server-side error (capability mismatch) sent a ProtocolError instead of a
RouterResponseMessage, so the TUI never set routerConnected=true.
Complete TUI visual redesign: Theme.kt with full palette + 10 styles,
3-row grid layout with 1.7:1 horizontal split, redesigned StatusBar
with context meter, EventStreamPanel, FooterBar with contextual
keybinds, WelcomePanel for IDLE state, InputBar as left panel footer.
All existing components updated to use Theme colors. RouterPanel keeps
'not connected — epic 14' placeholder.
The diffScrollOffset in NavUp/NavDown has no upper bound — it can
scroll past the last line. The ApprovalSurface clamps the viewport
with coerceIn/minOf, so it doesn't crash, but the offset keeps
growing. Marked with TODO(diff-scroll) in both handlers.
- 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)
P3-1: Delete 17 empty Gradle modules (core:agents/observability/policies/stages,
infrastructure:scheduler/security/telemetry, all 7 plugins/*, all 3
interfaces/*). Zero source files, zero dependents, identical build.gradle
stubs. Remove from settings.gradle. Keep .adoc spec docs as roadmap refs.
P3-2: Fix docs drift — rename :core:orchestration to :core:kernel in
core-orchestration-submodule-spec.md, rewrite stale chat-transcript
modules-and-spec.md as proper doc.
(CLAUDE.md Router context isolation clarification is local-only,
file is gitignored.)
P3-3: Reduce detekt maxIssues from 999999 to 120 (actual current count ~107).
(CLAUDE.md detekt description updated locally; file is gitignored.)
P3-4: Remove stale locked worktree agent-a98d45277ce4b0040 (contained only
cosmetic test style changes and an unused kotlinx-datetime dep).
P2-1: Guard SessionId NPE with safe-call in SessionsReducer
P2-2: Remove 16 dead event classes + reducer branches; replace with live orchestration events in tests
P2-3: Standardize ChatInput divergences — guard CancelSession, single-arg ProtocolError
P2-4: Remove dead TuiToolRecord.diff and ToolDisplayStatus.REQUESTED
P2-5: Remove dead streamLive from SessionEventBridge
P2-6: Extract SessionsReducerContext data class for 6+ param method
P2-7: Replace StageId("none") sentinel with TypeId.NONE
P2-9: Add TypeId.random() factory on all type-alias IDs
P2-10: Add KDoc on schemaVersion documenting reserved-for-migration
P2-8: Verified RouterReducer string-template already correct (TypeId value class toString)
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.
Event log integrity (sole source of truth):
- SqliteEventStore: serialize append/appendAll under a Mutex, enable WAL +
busy_timeout + synchronous=NORMAL, roll back on any Throwable. Replaces the
app-computed `SELECT MAX(seq)+1` read-modify-write that dropped events under
concurrent appends (race was invisible to CI: only the in-memory store and
single-threaded :memory: tests were ever exercised).
- Serialization: encodeDefaults + ignoreUnknownKeys, explicit @SerialName on all
46 EventPayload subclasses and event-referenced enum constants, @Serializable on
RiskLevel/RiskAction. Guards the append-only log against silent corruption from
future class renames, field removals, or default-value changes.
Approval/grant security (Invariant: approvals cannot widen authority):
- Tier authorization is now a ceiling (request.tier.level <= max granted) instead
of set membership; SESSION grants bind to a specific toolName instead of matching
everything; handleCreateGrant rejects empty/over-tier/scopeless grants.
Router chat wiring (Phase 0-2): ChatInput round-trip, StartChatSession from IDLE,
router-panel layout, workflows behind Ctrl+W.
Tests: SqliteEventStore concurrency, serialization round-trip + discriminator
stability + unknown-key tolerance, adversarial grant scope/tier; updated existing
approval and reducer tests for the new grant semantics and StartChatSession effect.
- 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)
kami