fix: resolve CSRF exempt endpoint before execution (#626)

Co-authored-by: Kimi Agent <kimi@timmy.local>
Co-committed-by: Kimi Agent <kimi@timmy.local>

scripts/cycle_retro.py

@@ -54,6 +54,7 @@ REPO_ROOT = Path(__file__).resolve().parent.parent
 RETRO_FILE = REPO_ROOT / ".loop" / "retro" / "cycles.jsonl"
 SUMMARY_FILE = REPO_ROOT / ".loop" / "retro" / "summary.json"
 EPOCH_COUNTER_FILE = REPO_ROOT / ".loop" / "retro" / ".epoch_counter"
+CYCLE_RESULT_FILE = REPO_ROOT / ".loop" / "cycle_result.json"
 
 # How many recent entries to include in rolling summary
 SUMMARY_WINDOW = 50
@@ -246,9 +247,37 @@ def update_summary() -> None:
     SUMMARY_FILE.write_text(json.dumps(summary, indent=2) + "\n")
 
 
+def _load_cycle_result() -> dict:
+    """Read .loop/cycle_result.json if it exists; return empty dict on failure."""
+    if not CYCLE_RESULT_FILE.exists():
+        return {}
+    try:
+        raw = CYCLE_RESULT_FILE.read_text().strip()
+        # Strip hermes fence markers (```json ... ```) if present
+        if raw.startswith("```"):
+            lines = raw.splitlines()
+            lines = [l for l in lines if not l.startswith("```")]
+            raw = "\n".join(lines)
+        return json.loads(raw)
+    except (json.JSONDecodeError, OSError):
+        return {}
+
+
 def main() -> None:
     args = parse_args()
 
+    # Backfill from cycle_result.json when CLI args have defaults
+    cr = _load_cycle_result()
+    if cr:
+        if args.issue is None and cr.get("issue"):
+            args.issue = int(cr["issue"])
+        if args.type == "unknown" and cr.get("type"):
+            args.type = cr["type"]
+        if args.tests_passed == 0 and cr.get("tests_passed"):
+            args.tests_passed = int(cr["tests_passed"])
+        if not args.notes and cr.get("notes"):
+            args.notes = cr["notes"]
+
     # Auto-detect issue from branch when not explicitly provided
     if args.issue is None:
         args.issue = detect_issue_from_branch()

src/config.py

@@ -84,6 +84,7 @@ class Settings(BaseSettings):
     # Only used when explicitly enabled and query complexity warrants it.
     grok_enabled: bool = False
     xai_api_key: str = ""
+    xai_base_url: str = "https://api.x.ai/v1"
     grok_default_model: str = "grok-3-fast"
     grok_max_sats_per_query: int = 200
     grok_free: bool = False  # Skip Lightning invoice when user has own API key

src/dashboard/app.py

@@ -46,6 +46,7 @@ from dashboard.routes.tasks import router as tasks_router
 from dashboard.routes.telegram import router as telegram_router
 from dashboard.routes.thinking import router as thinking_router
 from dashboard.routes.tools import router as tools_router
+from dashboard.routes.tower import router as tower_router
 from dashboard.routes.voice import router as voice_router
 from dashboard.routes.work_orders import router as work_orders_router
 from dashboard.routes.world import router as world_router
@@ -376,73 +377,78 @@ def _startup_background_tasks() -> list[asyncio.Task]:
     ]
 
 
+def _try_prune(label: str, prune_fn, days: int) -> None:
+    """Run a prune function, log results, swallow errors."""
+    try:
+        pruned = prune_fn()
+        if pruned:
+            logger.info(
+                "%s auto-prune: removed %d entries older than %d days",
+                label,
+                pruned,
+                days,
+            )
+    except Exception as exc:
+        logger.debug("%s auto-prune skipped: %s", label, exc)
+
+
+def _check_vault_size() -> None:
+    """Warn if the memory vault exceeds the configured size limit."""
+    try:
+        vault_path = Path(settings.repo_root) / "memory" / "notes"
+        if vault_path.exists():
+            total_bytes = sum(f.stat().st_size for f in vault_path.rglob("*") if f.is_file())
+            total_mb = total_bytes / (1024 * 1024)
+            if total_mb > settings.memory_vault_max_mb:
+                logger.warning(
+                    "Memory vault (%.1f MB) exceeds limit (%d MB) — consider archiving old notes",
+                    total_mb,
+                    settings.memory_vault_max_mb,
+                )
+    except Exception as exc:
+        logger.debug("Vault size check skipped: %s", exc)
+
+
 def _startup_pruning() -> None:
     """Auto-prune old memories, thoughts, and events on startup."""
     if settings.memory_prune_days > 0:
-        try:
-            from timmy.memory_system import prune_memories
+        from timmy.memory_system import prune_memories
 
-            pruned = prune_memories(
+        _try_prune(
+            "Memory",
+            lambda: prune_memories(
                 older_than_days=settings.memory_prune_days,
                 keep_facts=settings.memory_prune_keep_facts,
-            )
-            if pruned:
-                logger.info(
-                    "Memory auto-prune: removed %d entries older than %d days",
-                    pruned,
-                    settings.memory_prune_days,
-                )
-        except Exception as exc:
-            logger.debug("Memory auto-prune skipped: %s", exc)
+            ),
+            settings.memory_prune_days,
+        )
 
     if settings.thoughts_prune_days > 0:
-        try:
-            from timmy.thinking import thinking_engine
+        from timmy.thinking import thinking_engine
 
-            pruned = thinking_engine.prune_old_thoughts(
+        _try_prune(
+            "Thought",
+            lambda: thinking_engine.prune_old_thoughts(
                 keep_days=settings.thoughts_prune_days,
                 keep_min=settings.thoughts_prune_keep_min,
-            )
-            if pruned:
-                logger.info(
-                    "Thought auto-prune: removed %d entries older than %d days",
-                    pruned,
-                    settings.thoughts_prune_days,
-                )
-        except Exception as exc:
-            logger.debug("Thought auto-prune skipped: %s", exc)
+            ),
+            settings.thoughts_prune_days,
+        )
 
     if settings.events_prune_days > 0:
-        try:
-            from swarm.event_log import prune_old_events
+        from swarm.event_log import prune_old_events
 
-            pruned = prune_old_events(
+        _try_prune(
+            "Event",
+            lambda: prune_old_events(
                 keep_days=settings.events_prune_days,
                 keep_min=settings.events_prune_keep_min,
-            )
-            if pruned:
-                logger.info(
-                    "Event auto-prune: removed %d entries older than %d days",
-                    pruned,
-                    settings.events_prune_days,
-                )
-        except Exception as exc:
-            logger.debug("Event auto-prune skipped: %s", exc)
+            ),
+            settings.events_prune_days,
+        )
 
     if settings.memory_vault_max_mb > 0:
-        try:
-            vault_path = Path(settings.repo_root) / "memory" / "notes"
-            if vault_path.exists():
-                total_bytes = sum(f.stat().st_size for f in vault_path.rglob("*") if f.is_file())
-                total_mb = total_bytes / (1024 * 1024)
-                if total_mb > settings.memory_vault_max_mb:
-                    logger.warning(
-                        "Memory vault (%.1f MB) exceeds limit (%d MB) — consider archiving old notes",
-                        total_mb,
-                        settings.memory_vault_max_mb,
-                    )
-        except Exception as exc:
-            logger.debug("Vault size check skipped: %s", exc)
+        _check_vault_size()
 
 
 async def _shutdown_cleanup(
@@ -583,6 +589,7 @@ app.include_router(system_router)
 app.include_router(experiments_router)
 app.include_router(db_explorer_router)
 app.include_router(world_router)
+app.include_router(tower_router)
 
 
 @app.websocket("/ws")

src/dashboard/middleware/csrf.py

@@ -175,18 +175,12 @@ class CSRFMiddleware(BaseHTTPMiddleware):
             return await call_next(request)
 
         # Token validation failed and path is not exempt
-        # We still need to call the app to check if the endpoint is decorated
-        # with @csrf_exempt, so we'll let it through and check after routing
-        response = await call_next(request)
-
-        # After routing, check if the endpoint is marked as exempt
-        endpoint = request.scope.get("endpoint")
+        # Resolve the endpoint from routes BEFORE executing to avoid side effects
+        endpoint = self._resolve_endpoint(request)
         if endpoint and is_csrf_exempt(endpoint):
-            # Endpoint is marked as exempt, allow the response
-            return response
+            return await call_next(request)
 
-        # Endpoint is not exempt and token validation failed
-        # Return 403 error
+        # Endpoint is not exempt and token validation failed — reject without executing
         return JSONResponse(
             status_code=403,
             content={
@@ -196,6 +190,42 @@ class CSRFMiddleware(BaseHTTPMiddleware):
             },
         )
 
+    def _resolve_endpoint(self, request: Request) -> Callable | None:
+        """Resolve the endpoint for a request without executing it.
+
+        Walks the app chain to find routes, then matches against the request
+        scope. This allows checking @csrf_exempt before the handler runs
+        (avoiding side effects on CSRF rejection).
+
+        Returns:
+            The endpoint callable if found, None otherwise.
+        """
+        try:
+            from starlette.routing import Match
+
+            # Walk the middleware/app chain to find something with routes
+            routes = None
+            current = self.app
+            for _ in range(10):  # Safety limit
+                routes = getattr(current, "routes", None)
+                if routes:
+                    break
+                current = getattr(current, "app", None)
+                if current is None:
+                    break
+
+            if not routes:
+                return None
+
+            scope = dict(request.scope)
+            for route in routes:
+                match, child_scope = route.matches(scope)
+                if match == Match.FULL:
+                    return child_scope.get("endpoint")
+        except Exception:
+            logger.debug("Failed to resolve endpoint for CSRF check")
+        return None
+
     def _is_likely_exempt(self, path: str) -> bool:
         """Check if a path is likely to be CSRF exempt.
 

src/dashboard/models/calm.py

@@ -1,4 +1,4 @@
-from datetime import date, datetime
+from datetime import UTC, date, datetime
 from enum import StrEnum
 
 from sqlalchemy import JSON, Boolean, Column, Date, DateTime, Index, Integer, String
@@ -40,8 +40,13 @@ class Task(Base):
     deferred_at = Column(DateTime, nullable=True)
 
     # Timestamps
-    created_at = Column(DateTime, default=datetime.utcnow, nullable=False)
-    updated_at = Column(DateTime, default=datetime.utcnow, onupdate=datetime.utcnow, nullable=False)
+    created_at = Column(DateTime, default=lambda: datetime.now(UTC), nullable=False)
+    updated_at = Column(
+        DateTime,
+        default=lambda: datetime.now(UTC),
+        onupdate=lambda: datetime.now(UTC),
+        nullable=False,
+    )
 
     __table_args__ = (Index("ix_task_state_order", "state", "sort_order"),)
 
@@ -59,4 +64,4 @@ class JournalEntry(Base):
     gratitude = Column(String(500), nullable=True)
     energy_level = Column(Integer, nullable=True)  # User-reported, 1-10
 
-    created_at = Column(DateTime, default=datetime.utcnow, nullable=False)
+    created_at = Column(DateTime, default=lambda: datetime.now(UTC), nullable=False)

src/dashboard/routes/calm.py

@@ -1,5 +1,5 @@
 import logging
-from datetime import date, datetime
+from datetime import UTC, date, datetime
 
 from fastapi import APIRouter, Depends, Form, HTTPException, Request
 from fastapi.responses import HTMLResponse
@@ -38,6 +38,56 @@ def get_later_tasks(db: Session) -> list[Task]:
     )
 
 
+def _create_mit_tasks(db: Session, titles: list[str | None]) -> list[int]:
+    """Create MIT tasks from a list of titles, return their IDs."""
+    task_ids: list[int] = []
+    for title in titles:
+        if title:
+            task = Task(
+                title=title,
+                is_mit=True,
+                state=TaskState.LATER,
+                certainty=TaskCertainty.SOFT,
+            )
+            db.add(task)
+            db.commit()
+            db.refresh(task)
+            task_ids.append(task.id)
+    return task_ids
+
+
+def _create_other_tasks(db: Session, other_tasks: str):
+    """Create non-MIT tasks from newline-separated text."""
+    for line in other_tasks.split("\n"):
+        line = line.strip()
+        if line:
+            task = Task(
+                title=line,
+                state=TaskState.LATER,
+                certainty=TaskCertainty.FUZZY,
+            )
+            db.add(task)
+
+
+def _seed_now_next(db: Session):
+    """Set initial NOW/NEXT states when both slots are empty."""
+    if get_now_task(db) or get_next_task(db):
+        return
+    later_tasks = (
+        db.query(Task)
+        .filter(Task.state == TaskState.LATER)
+        .order_by(Task.is_mit.desc(), Task.sort_order)
+        .all()
+    )
+    if later_tasks:
+        later_tasks[0].state = TaskState.NOW
+        db.add(later_tasks[0])
+        db.flush()
+        if len(later_tasks) > 1:
+            later_tasks[1].state = TaskState.NEXT
+            db.add(later_tasks[1])
+
+
 def promote_tasks(db: Session):
     """Enforce the NOW/NEXT/LATER state machine invariants.
 
@@ -114,63 +164,19 @@ async def post_morning_ritual(
     other_tasks: str = Form(""),
 ):
     """Process morning ritual: create MITs, other tasks, and set initial states."""
-    # Create Journal Entry
-    mit_task_ids = []
     journal_entry = JournalEntry(entry_date=date.today())
     db.add(journal_entry)
     db.commit()
     db.refresh(journal_entry)
 
-    # Create MIT tasks
-    for mit_title in [mit1_title, mit2_title, mit3_title]:
-        if mit_title:
-            task = Task(
-                title=mit_title,
-                is_mit=True,
-                state=TaskState.LATER,  # Initially LATER, will be promoted
-                certainty=TaskCertainty.SOFT,
-            )
-            db.add(task)
-            db.commit()
-            db.refresh(task)
-            mit_task_ids.append(task.id)
-
-    journal_entry.mit_task_ids = mit_task_ids
+    journal_entry.mit_task_ids = _create_mit_tasks(db, [mit1_title, mit2_title, mit3_title])
     db.add(journal_entry)
 
-    # Create other tasks
-    for task_title in other_tasks.split("\n"):
-        task_title = task_title.strip()
-        if task_title:
-            task = Task(
-                title=task_title,
-                state=TaskState.LATER,
-                certainty=TaskCertainty.FUZZY,
-            )
-            db.add(task)
-
+    _create_other_tasks(db, other_tasks)
     db.commit()
 
-    # Set initial NOW/NEXT states
-    # Set initial NOW/NEXT states after all tasks are created
-    if not get_now_task(db) and not get_next_task(db):
-        later_tasks = (
-            db.query(Task)
-            .filter(Task.state == TaskState.LATER)
-            .order_by(Task.is_mit.desc(), Task.sort_order)
-            .all()
-        )
-        if later_tasks:
-            # Set the highest priority LATER task to NOW
-            later_tasks[0].state = TaskState.NOW
-            db.add(later_tasks[0])
-            db.flush()  # Flush to make the change visible for the next query
-
-            # Set the next highest priority LATER task to NEXT
-            if len(later_tasks) > 1:
-                later_tasks[1].state = TaskState.NEXT
-                db.add(later_tasks[1])
-    db.commit()  # Commit changes after initial NOW/NEXT setup
+    _seed_now_next(db)
+    db.commit()
 
     return templates.TemplateResponse(
         request,
@@ -220,7 +226,7 @@ async def post_evening_ritual(
     )
     for task in active_tasks:
         task.state = TaskState.DEFERRED  # Or DONE, depending on desired archiving logic
-        task.deferred_at = datetime.utcnow()
+        task.deferred_at = datetime.now(UTC)
         db.add(task)
 
     db.commit()
@@ -273,7 +279,7 @@ async def start_task(
         raise HTTPException(status_code=404, detail="Task not found")
 
     task.state = TaskState.NOW
-    task.started_at = datetime.utcnow()
+    task.started_at = datetime.now(UTC)
     db.add(task)
     db.commit()
 
@@ -303,7 +309,7 @@ async def complete_task(
         raise HTTPException(status_code=404, detail="Task not found")
 
     task.state = TaskState.DONE
-    task.completed_at = datetime.utcnow()
+    task.completed_at = datetime.now(UTC)
     db.add(task)
     db.commit()
 
@@ -332,7 +338,7 @@ async def defer_task(
         raise HTTPException(status_code=404, detail="Task not found")
 
     task.state = TaskState.DEFERRED
-    task.deferred_at = datetime.utcnow()
+    task.deferred_at = datetime.now(UTC)
     db.add(task)
     db.commit()
 

src/dashboard/routes/system.py

@@ -16,52 +16,11 @@ router = APIRouter(tags=["system"])
 
 @router.get("/lightning/ledger", response_class=HTMLResponse)
 async def lightning_ledger(request: Request):
-    """Ledger and balance page."""
-    # Mock data for now, as this seems to be a UI-first feature
-    balance = {
-        "available_sats": 1337,
-        "incoming_total_sats": 2000,
-        "outgoing_total_sats": 663,
-        "fees_paid_sats": 5,
-        "net_sats": 1337,
-        "pending_incoming_sats": 0,
-        "pending_outgoing_sats": 0,
-    }
+    """Ledger and balance page backed by the in-memory Lightning ledger."""
+    from lightning.ledger import get_balance, get_transactions
 
-    # Mock transactions
-    from collections import namedtuple
-    from enum import Enum
-
-    class TxType(Enum):
-        incoming = "incoming"
-        outgoing = "outgoing"
-
-    class TxStatus(Enum):
-        completed = "completed"
-        pending = "pending"
-
-    Tx = namedtuple(
-        "Tx", ["tx_type", "status", "amount_sats", "payment_hash", "memo", "created_at"]
-    )
-
-    transactions = [
-        Tx(
-            TxType.outgoing,
-            TxStatus.completed,
-            50,
-            "hash1",
-            "Model inference",
-            "2026-03-04 10:00:00",
-        ),
-        Tx(
-            TxType.incoming,
-            TxStatus.completed,
-            1000,
-            "hash2",
-            "Manual deposit",
-            "2026-03-03 15:00:00",
-        ),
-    ]
+    balance = get_balance()
+    transactions = get_transactions()
 
     return templates.TemplateResponse(
         request,
@@ -70,7 +29,7 @@ async def lightning_ledger(request: Request):
             "balance": balance,
             "transactions": transactions,
             "tx_types": ["incoming", "outgoing"],
-            "tx_statuses": ["completed", "pending"],
+            "tx_statuses": ["pending", "settled", "failed", "expired"],
             "filter_type": None,
             "filter_status": None,
             "stats": {},

src/dashboard/routes/tasks.py

@@ -5,7 +5,7 @@ import sqlite3
 import uuid
 from collections.abc import Generator
 from contextlib import closing, contextmanager
-from datetime import datetime
+from datetime import UTC, datetime
 from pathlib import Path
 
 from fastapi import APIRouter, Form, HTTPException, Request
@@ -219,7 +219,7 @@ async def create_task_form(
         raise HTTPException(status_code=400, detail="Task title cannot be empty")
 
     task_id = str(uuid.uuid4())
-    now = datetime.utcnow().isoformat()
+    now = datetime.now(UTC).isoformat()
     priority = priority if priority in VALID_PRIORITIES else "normal"
 
     with _get_db() as db:
@@ -287,7 +287,7 @@ async def modify_task(
 async def _set_status(request: Request, task_id: str, new_status: str):
     """Helper to update status and return refreshed task card."""
     completed_at = (
-        datetime.utcnow().isoformat() if new_status in ("completed", "vetoed", "failed") else None
+        datetime.now(UTC).isoformat() if new_status in ("completed", "vetoed", "failed") else None
     )
     with _get_db() as db:
         db.execute(
@@ -316,7 +316,7 @@ async def api_create_task(request: Request):
         raise HTTPException(422, "title is required")
 
     task_id = str(uuid.uuid4())
-    now = datetime.utcnow().isoformat()
+    now = datetime.now(UTC).isoformat()
     priority = body.get("priority", "normal")
     if priority not in VALID_PRIORITIES:
         priority = "normal"
@@ -358,7 +358,7 @@ async def api_update_status(task_id: str, request: Request):
         raise HTTPException(422, f"Invalid status. Must be one of: {VALID_STATUSES}")
 
     completed_at = (
-        datetime.utcnow().isoformat() if new_status in ("completed", "vetoed", "failed") else None
+        datetime.now(UTC).isoformat() if new_status in ("completed", "vetoed", "failed") else None
     )
     with _get_db() as db:
         db.execute(

src/dashboard/routes/tower.py

@@ -0,0 +1,108 @@
+"""Tower dashboard — real-time Spark visualization via WebSocket.
+
+GET  /tower     — HTML Tower dashboard (Thinking / Predicting / Advising)
+WS   /tower/ws  — WebSocket stream of Spark engine state updates
+"""
+
+import asyncio
+import json
+import logging
+
+from fastapi import APIRouter, Request, WebSocket
+from fastapi.responses import HTMLResponse
+
+from dashboard.templating import templates
+from spark.engine import spark_engine
+
+logger = logging.getLogger(__name__)
+
+router = APIRouter(prefix="/tower", tags=["tower"])
+
+_PUSH_INTERVAL = 5  # seconds between state broadcasts
+
+
+def _spark_snapshot() -> dict:
+    """Build a JSON-serialisable snapshot of Spark state."""
+    status = spark_engine.status()
+
+    timeline = spark_engine.get_timeline(limit=10)
+    events = []
+    for ev in timeline:
+        entry = {
+            "event_type": ev.event_type,
+            "description": ev.description,
+            "importance": ev.importance,
+            "created_at": ev.created_at,
+        }
+        if ev.agent_id:
+            entry["agent_id"] = ev.agent_id[:8]
+        if ev.task_id:
+            entry["task_id"] = ev.task_id[:8]
+        try:
+            entry["data"] = json.loads(ev.data)
+        except (json.JSONDecodeError, TypeError):
+            entry["data"] = {}
+        events.append(entry)
+
+    predictions = spark_engine.get_predictions(limit=5)
+    preds = []
+    for p in predictions:
+        pred = {
+            "task_id": p.task_id[:8] if p.task_id else "?",
+            "accuracy": p.accuracy,
+            "evaluated": p.evaluated_at is not None,
+            "created_at": p.created_at,
+        }
+        try:
+            pred["predicted"] = json.loads(p.predicted_value)
+        except (json.JSONDecodeError, TypeError):
+            pred["predicted"] = {}
+        preds.append(pred)
+
+    advisories = spark_engine.get_advisories()
+    advs = [
+        {
+            "category": a.category,
+            "priority": a.priority,
+            "title": a.title,
+            "detail": a.detail,
+            "suggested_action": a.suggested_action,
+        }
+        for a in advisories
+    ]
+
+    return {
+        "type": "spark_state",
+        "status": status,
+        "events": events,
+        "predictions": preds,
+        "advisories": advs,
+    }
+
+
+@router.get("", response_class=HTMLResponse)
+async def tower_ui(request: Request):
+    """Render the Tower dashboard page."""
+    snapshot = _spark_snapshot()
+    return templates.TemplateResponse(
+        request,
+        "tower.html",
+        {"snapshot": snapshot},
+    )
+
+
+@router.websocket("/ws")
+async def tower_ws(websocket: WebSocket) -> None:
+    """Stream Spark state snapshots to the Tower dashboard."""
+    await websocket.accept()
+    logger.info("Tower WS connected")
+
+    try:
+        # Send initial snapshot
+        await websocket.send_text(json.dumps(_spark_snapshot()))
+
+        while True:
+            await asyncio.sleep(_PUSH_INTERVAL)
+            await websocket.send_text(json.dumps(_spark_snapshot()))
+    except Exception:
+        logger.debug("Tower WS disconnected")

src/dashboard/routes/work_orders.py

@@ -5,7 +5,7 @@ import sqlite3
 import uuid
 from collections.abc import Generator
 from contextlib import closing, contextmanager
-from datetime import datetime
+from datetime import UTC, datetime
 from pathlib import Path
 
 from fastapi import APIRouter, Form, HTTPException, Request
@@ -144,7 +144,7 @@ async def submit_work_order(
     related_files: str = Form(""),
 ):
     wo_id = str(uuid.uuid4())
-    now = datetime.utcnow().isoformat()
+    now = datetime.now(UTC).isoformat()
     priority = priority if priority in PRIORITIES else "medium"
     category = category if category in CATEGORIES else "suggestion"
 
@@ -211,7 +211,7 @@ async def active_partial(request: Request):
 
 async def _update_status(request: Request, wo_id: str, new_status: str, **extra):
     completed_at = (
-        datetime.utcnow().isoformat() if new_status in ("completed", "rejected") else None
+        datetime.now(UTC).isoformat() if new_status in ("completed", "rejected") else None
     )
     with _get_db() as db:
         sets = ["status=?", "completed_at=COALESCE(?, completed_at)"]

src/dashboard/templates/mission_control.html

@@ -138,6 +138,47 @@
     </div>
 </div>
 
+<!-- Spark Intelligence -->
+{% from "macros.html" import panel %}
+<div class="mc-card-spaced">
+  <div class="card">
+    <div class="card-header">
+      <h2 class="card-title">Spark Intelligence</h2>
+      <div>
+        <span class="badge" id="spark-status-badge">Loading...</span>
+      </div>
+    </div>
+    <div class="grid grid-3">
+      <div class="stat">
+        <div class="stat-value" id="spark-events">-</div>
+        <div class="stat-label">Events</div>
+      </div>
+      <div class="stat">
+        <div class="stat-value" id="spark-memories">-</div>
+        <div class="stat-label">Memories</div>
+      </div>
+      <div class="stat">
+        <div class="stat-value" id="spark-predictions">-</div>
+        <div class="stat-label">Predictions</div>
+      </div>
+    </div>
+  </div>
+  <div class="grid grid-2 mc-section-gap">
+    {% call panel("SPARK TIMELINE", id="spark-timeline-panel",
+                  hx_get="/spark/timeline",
+                  hx_trigger="load, every 10s") %}
+      <div class="spark-timeline-scroll">
+        <p class="chat-history-placeholder">Loading timeline...</p>
+      </div>
+    {% endcall %}
+    {% call panel("SPARK INSIGHTS", id="spark-insights-panel",
+                  hx_get="/spark/insights",
+                  hx_trigger="load, every 30s") %}
+      <p class="chat-history-placeholder">Loading insights...</p>
+    {% endcall %}
+  </div>
+</div>
+
 <!-- Chat History -->
 <div class="card mc-card-spaced">
     <div class="card-header">
@@ -428,7 +469,34 @@ async function loadGrokStats() {
     }
 }
 
+// Load Spark status
+async function loadSparkStatus() {
+    try {
+        var response = await fetch('/spark');
+        var data = await response.json();
+        var st = data.status || {};
+
+        document.getElementById('spark-events').textContent = st.total_events || 0;
+        document.getElementById('spark-memories').textContent = st.total_memories || 0;
+        document.getElementById('spark-predictions').textContent = st.total_predictions || 0;
+
+        var badge = document.getElementById('spark-status-badge');
+        if (st.total_events > 0) {
+            badge.textContent = 'Active';
+            badge.className = 'badge badge-success';
+        } else {
+            badge.textContent = 'Idle';
+            badge.className = 'badge badge-warning';
+        }
+    } catch (error) {
+        var badge = document.getElementById('spark-status-badge');
+        badge.textContent = 'Offline';
+        badge.className = 'badge badge-danger';
+    }
+}
+
 // Initial load
+loadSparkStatus();
 loadSovereignty();
 loadHealth();
 loadSwarmStats();
@@ -442,5 +510,6 @@ setInterval(loadHealth, 10000);
 setInterval(loadSwarmStats, 5000);
 setInterval(updateHeartbeat, 5000);
 setInterval(loadGrokStats, 10000);
+setInterval(loadSparkStatus, 15000);
 </script>
 {% endblock %}

src/dashboard/templates/tower.html

@@ -0,0 +1,180 @@
+{% extends "base.html" %}
+
+{% block title %}Timmy Time — Tower{% endblock %}
+
+{% block extra_styles %}{% endblock %}
+
+{% block content %}
+<div class="container-fluid tower-container py-3">
+
+  <div class="tower-header">
+    <div class="tower-title">TOWER</div>
+    <div class="tower-subtitle">
+      Real-time Spark visualization &mdash;
+      <span id="tower-conn" class="tower-conn-badge tower-conn-connecting">CONNECTING</span>
+    </div>
+  </div>
+
+  <div class="row g-3">
+
+    <!-- Left: THINKING (events) -->
+    <div class="col-12 col-lg-4 d-flex flex-column gap-3">
+      <div class="card mc-panel tower-phase-card">
+        <div class="card-header mc-panel-header tower-phase-thinking">// THINKING</div>
+        <div class="card-body p-3 tower-scroll" id="tower-events">
+          <div class="tower-empty">Waiting for Spark data&hellip;</div>
+        </div>
+      </div>
+    </div>
+
+    <!-- Middle: PREDICTING (EIDOS) -->
+    <div class="col-12 col-lg-4 d-flex flex-column gap-3">
+      <div class="card mc-panel tower-phase-card">
+        <div class="card-header mc-panel-header tower-phase-predicting">// PREDICTING</div>
+        <div class="card-body p-3" id="tower-predictions">
+          <div class="tower-empty">Waiting for Spark data&hellip;</div>
+        </div>
+      </div>
+      <div class="card mc-panel">
+        <div class="card-header mc-panel-header">// EIDOS STATS</div>
+        <div class="card-body p-3">
+          <div class="tower-stat-grid" id="tower-stats">
+            <div class="tower-stat"><span class="tower-stat-label">EVENTS</span><span class="tower-stat-value" id="ts-events">0</span></div>
+            <div class="tower-stat"><span class="tower-stat-label">MEMORIES</span><span class="tower-stat-value" id="ts-memories">0</span></div>
+            <div class="tower-stat"><span class="tower-stat-label">PREDICTIONS</span><span class="tower-stat-value" id="ts-preds">0</span></div>
+            <div class="tower-stat"><span class="tower-stat-label">ACCURACY</span><span class="tower-stat-value" id="ts-accuracy">—</span></div>
+          </div>
+        </div>
+      </div>
+    </div>
+
+    <!-- Right: ADVISING -->
+    <div class="col-12 col-lg-4 d-flex flex-column gap-3">
+      <div class="card mc-panel tower-phase-card">
+        <div class="card-header mc-panel-header tower-phase-advising">// ADVISING</div>
+        <div class="card-body p-3 tower-scroll" id="tower-advisories">
+          <div class="tower-empty">Waiting for Spark data&hellip;</div>
+        </div>
+      </div>
+    </div>
+
+  </div>
+</div>
+
+<script>
+(function() {
+  var ws = null;
+  var badge = document.getElementById('tower-conn');
+
+  function setConn(state) {
+    badge.textContent = state.toUpperCase();
+    badge.className = 'tower-conn-badge tower-conn-' + state;
+  }
+
+  function esc(s) { var d = document.createElement('div'); d.textContent = s; return d.innerHTML; }
+
+  function renderEvents(events) {
+    var el = document.getElementById('tower-events');
+    if (!events || !events.length) { el.innerHTML = '<div class="tower-empty">No events captured yet.</div>'; return; }
+    var html = '';
+    for (var i = 0; i < events.length; i++) {
+      var ev = events[i];
+      var dots = ev.importance >= 0.8 ? '\u25cf\u25cf\u25cf' : ev.importance >= 0.5 ? '\u25cf\u25cf' : '\u25cf';
+      html += '<div class="tower-event tower-etype-' + esc(ev.event_type) + '">'
+        + '<div class="tower-ev-head">'
+        + '<span class="tower-ev-badge">' + esc(ev.event_type.replace(/_/g, ' ').toUpperCase()) + '</span>'
+        + '<span class="tower-ev-dots">' + dots + '</span>'
+        + '</div>'
+        + '<div class="tower-ev-desc">' + esc(ev.description) + '</div>'
+        + '<div class="tower-ev-time">' + esc((ev.created_at || '').slice(0, 19)) + '</div>'
+        + '</div>';
+    }
+    el.innerHTML = html;
+  }
+
+  function renderPredictions(preds) {
+    var el = document.getElementById('tower-predictions');
+    if (!preds || !preds.length) { el.innerHTML = '<div class="tower-empty">No predictions yet.</div>'; return; }
+    var html = '';
+    for (var i = 0; i < preds.length; i++) {
+      var p = preds[i];
+      var cls = p.evaluated ? 'tower-pred-done' : 'tower-pred-pending';
+      var accTxt = p.accuracy != null ? Math.round(p.accuracy * 100) + '%' : 'PENDING';
+      var accCls = p.accuracy != null ? (p.accuracy >= 0.7 ? 'text-success' : p.accuracy < 0.4 ? 'text-danger' : 'text-warning') : '';
+      html += '<div class="tower-pred ' + cls + '">'
+        + '<div class="tower-pred-head">'
+        + '<span class="tower-pred-task">' + esc(p.task_id) + '</span>'
+        + '<span class="tower-pred-acc ' + accCls + '">' + accTxt + '</span>'
+        + '</div>';
+      if (p.predicted) {
+        var pr = p.predicted;
+        html += '<div class="tower-pred-detail">';
+        if (pr.likely_winner) html += '<span>Winner: ' + esc(pr.likely_winner.slice(0, 8)) + '</span> ';
+        if (pr.success_probability != null) html += '<span>Success: ' + Math.round(pr.success_probability * 100) + '%</span> ';
+        html += '</div>';
+      }
+      html += '<div class="tower-ev-time">' + esc((p.created_at || '').slice(0, 19)) + '</div>'
+        + '</div>';
+    }
+    el.innerHTML = html;
+  }
+
+  function renderAdvisories(advs) {
+    var el = document.getElementById('tower-advisories');
+    if (!advs || !advs.length) { el.innerHTML = '<div class="tower-empty">No advisories yet.</div>'; return; }
+    var html = '';
+    for (var i = 0; i < advs.length; i++) {
+      var a = advs[i];
+      var prio = a.priority >= 0.7 ? 'high' : a.priority >= 0.4 ? 'medium' : 'low';
+      html += '<div class="tower-advisory tower-adv-' + prio + '">'
+        + '<div class="tower-adv-head">'
+        + '<span class="tower-adv-cat">' + esc(a.category.replace(/_/g, ' ').toUpperCase()) + '</span>'
+        + '<span class="tower-adv-prio">' + Math.round(a.priority * 100) + '%</span>'
+        + '</div>'
+        + '<div class="tower-adv-title">' + esc(a.title) + '</div>'
+        + '<div class="tower-adv-detail">' + esc(a.detail) + '</div>'
+        + '<div class="tower-adv-action">' + esc(a.suggested_action) + '</div>'
+        + '</div>';
+    }
+    el.innerHTML = html;
+  }
+
+  function renderStats(status) {
+    if (!status) return;
+    document.getElementById('ts-events').textContent = status.events_captured || 0;
+    document.getElementById('ts-memories').textContent = status.memories_stored || 0;
+    var p = status.predictions || {};
+    document.getElementById('ts-preds').textContent = p.total_predictions || 0;
+    var acc = p.avg_accuracy;
+    var accEl = document.getElementById('ts-accuracy');
+    if (acc != null) {
+      accEl.textContent = Math.round(acc * 100) + '%';
+      accEl.className = 'tower-stat-value ' + (acc >= 0.7 ? 'text-success' : acc < 0.4 ? 'text-danger' : 'text-warning');
+    } else {
+      accEl.textContent = '\u2014';
+    }
+  }
+
+  function handleMsg(data) {
+    if (data.type !== 'spark_state') return;
+    renderEvents(data.events);
+    renderPredictions(data.predictions);
+    renderAdvisories(data.advisories);
+    renderStats(data.status);
+  }
+
+  function connect() {
+    var proto = location.protocol === 'https:' ? 'wss:' : 'ws:';
+    ws = new WebSocket(proto + '//' + location.host + '/tower/ws');
+    ws.onopen = function() { setConn('live'); };
+    ws.onclose = function() { setConn('offline'); setTimeout(connect, 3000); };
+    ws.onerror = function() { setConn('offline'); };
+    ws.onmessage = function(e) {
+      try { handleMsg(JSON.parse(e.data)); } catch(err) { console.error('Tower WS parse error', err); }
+    };
+  }
+
+  connect();
+})();
+</script>
+{% endblock %}

src/infrastructure/error_capture.py

@@ -149,6 +149,52 @@ def _log_error_event(
         logger.debug("Failed to log error event: %s", log_exc)
 
 
+def _build_report_description(
+    exc: Exception,
+    source: str,
+    context: dict | None,
+    error_hash: str,
+    tb_str: str,
+    affected_file: str,
+    affected_line: int,
+    git_ctx: dict,
+) -> str:
+    """Build the markdown description for a bug report task."""
+    parts = [
+        f"**Error:** {type(exc).__name__}: {str(exc)}",
+        f"**Source:** {source}",
+        f"**File:** {affected_file}:{affected_line}",
+        f"**Git:** {git_ctx.get('branch', '?')} @ {git_ctx.get('commit', '?')}",
+        f"**Time:** {datetime.now(UTC).isoformat()}",
+        f"**Hash:** {error_hash}",
+    ]
+
+    if context:
+        ctx_str = ", ".join(f"{k}={v}" for k, v in context.items())
+        parts.append(f"**Context:** {ctx_str}")
+
+    parts.append(f"\n**Stack Trace:**\n```\n{tb_str[:2000]}\n```")
+    return "\n".join(parts)
+
+
+def _log_bug_report_created(source: str, task_id: str, error_hash: str, title: str) -> None:
+    """Log a BUG_REPORT_CREATED event (best-effort)."""
+    try:
+        from swarm.event_log import EventType, log_event
+
+        log_event(
+            EventType.BUG_REPORT_CREATED,
+            source=source,
+            task_id=task_id,
+            data={
+                "error_hash": error_hash,
+                "title": title[:100],
+            },
+        )
+    except Exception as exc:
+        logger.warning("Bug report event log error: %s", exc)
+
+
 def _create_bug_report(
     exc: Exception,
     source: str,
@@ -164,25 +210,20 @@ def _create_bug_report(
         from swarm.task_queue.models import create_task
 
         title = f"[BUG] {type(exc).__name__}: {str(exc)[:80]}"
-
-        description_parts = [
-            f"**Error:** {type(exc).__name__}: {str(exc)}",
-            f"**Source:** {source}",
-            f"**File:** {affected_file}:{affected_line}",
-            f"**Git:** {git_ctx.get('branch', '?')} @ {git_ctx.get('commit', '?')}",
-            f"**Time:** {datetime.now(UTC).isoformat()}",
-            f"**Hash:** {error_hash}",
-        ]
-
-        if context:
-            ctx_str = ", ".join(f"{k}={v}" for k, v in context.items())
-            description_parts.append(f"**Context:** {ctx_str}")
-
-        description_parts.append(f"\n**Stack Trace:**\n```\n{tb_str[:2000]}\n```")
+        description = _build_report_description(
+            exc,
+            source,
+            context,
+            error_hash,
+            tb_str,
+            affected_file,
+            affected_line,
+            git_ctx,
+        )
 
         task = create_task(
             title=title,
-            description="\n".join(description_parts),
+            description=description,
             assigned_to="default",
             created_by="system",
             priority="normal",
@@ -190,24 +231,9 @@ def _create_bug_report(
             auto_approve=True,
             task_type="bug_report",
         )
-        task_id = task.id
 
-        try:
-            from swarm.event_log import EventType, log_event
-
-            log_event(
-                EventType.BUG_REPORT_CREATED,
-                source=source,
-                task_id=task_id,
-                data={
-                    "error_hash": error_hash,
-                    "title": title[:100],
-                },
-            )
-        except Exception as exc:
-            logger.warning("Bug report screenshot error: %s", exc)
-
-        return task_id
+        _log_bug_report_created(source, task.id, error_hash, title)
+        return task.id
 
     except Exception as task_exc:
         logger.debug("Failed to create bug report task: %s", task_exc)

src/infrastructure/events/bus.py

@@ -64,7 +64,7 @@ class EventBus:
 
         @bus.subscribe("agent.task.*")
         async def handle_task(event: Event):
-            logger.debug(f"Task event: {event.data}")
+            logger.debug("Task event: %s", event.data)
 
         await bus.publish(Event(
             type="agent.task.assigned",

src/infrastructure/router/__init__.py

@@ -2,6 +2,7 @@
 
 from .api import router
 from .cascade import CascadeRouter, Provider, ProviderStatus, get_router
+from .history import HealthHistoryStore, get_history_store
 
 __all__ = [
     "CascadeRouter",
@@ -9,4 +10,6 @@ __all__ = [
     "ProviderStatus",
     "get_router",
     "router",
+    "HealthHistoryStore",
+    "get_history_store",
 ]

src/infrastructure/router/api.py

@@ -8,6 +8,7 @@ from fastapi import APIRouter, Depends, HTTPException
 from pydantic import BaseModel
 
 from .cascade import CascadeRouter, get_router
+from .history import HealthHistoryStore, get_history_store
 
 logger = logging.getLogger(__name__)
 router = APIRouter(prefix="/api/v1/router", tags=["router"])
@@ -199,6 +200,17 @@ async def reload_config(
         raise HTTPException(status_code=500, detail=f"Reload failed: {exc}") from exc
 
 
+@router.get("/history")
+async def get_history(
+    hours: int = 24,
+    store: Annotated[HealthHistoryStore, Depends(get_history_store)] = None,
+) -> list[dict[str, Any]]:
+    """Get provider health history for the last N hours."""
+    if store is None:
+        store = get_history_store()
+    return store.get_history(hours=hours)
+
+
 @router.get("/config")
 async def get_config(
     cascade: Annotated[CascadeRouter, Depends(get_cascade_router)],

src/infrastructure/router/cascade.py

@@ -221,65 +221,56 @@ class CascadeRouter:
                 raise RuntimeError("PyYAML not installed")
 
             content = self.config_path.read_text()
-            # Expand environment variables
             content = self._expand_env_vars(content)
             data = yaml.safe_load(content)
 
-            # Load cascade settings
-            cascade = data.get("cascade", {})
-
-            # Load fallback chains
-            fallback_chains = data.get("fallback_chains", {})
-
-            # Load multi-modal settings
-            multimodal = data.get("multimodal", {})
-
-            self.config = RouterConfig(
-                timeout_seconds=cascade.get("timeout_seconds", 30),
-                max_retries_per_provider=cascade.get("max_retries_per_provider", 2),
-                retry_delay_seconds=cascade.get("retry_delay_seconds", 1),
-                circuit_breaker_failure_threshold=cascade.get("circuit_breaker", {}).get(
-                    "failure_threshold", 5
-                ),
-                circuit_breaker_recovery_timeout=cascade.get("circuit_breaker", {}).get(
-                    "recovery_timeout", 60
-                ),
-                circuit_breaker_half_open_max_calls=cascade.get("circuit_breaker", {}).get(
-                    "half_open_max_calls", 2
-                ),
-                auto_pull_models=multimodal.get("auto_pull", True),
-                fallback_chains=fallback_chains,
-            )
-
-            # Load providers
-            for p_data in data.get("providers", []):
-                # Skip disabled providers
-                if not p_data.get("enabled", False):
-                    continue
-
-                provider = Provider(
-                    name=p_data["name"],
-                    type=p_data["type"],
-                    enabled=p_data.get("enabled", True),
-                    priority=p_data.get("priority", 99),
-                    url=p_data.get("url"),
-                    api_key=p_data.get("api_key"),
-                    base_url=p_data.get("base_url"),
-                    models=p_data.get("models", []),
-                )
-
-                # Check if provider is actually available
-                if self._check_provider_available(provider):
-                    self.providers.append(provider)
-                else:
-                    logger.warning("Provider %s not available, skipping", provider.name)
-
-            # Sort by priority
-            self.providers.sort(key=lambda p: p.priority)
+            self.config = self._parse_router_config(data)
+            self._load_providers(data)
 
         except Exception as exc:
             logger.error("Failed to load config: %s", exc)
 
+    def _parse_router_config(self, data: dict) -> RouterConfig:
+        """Build a RouterConfig from parsed YAML data."""
+        cascade = data.get("cascade", {})
+        cb = cascade.get("circuit_breaker", {})
+        multimodal = data.get("multimodal", {})
+
+        return RouterConfig(
+            timeout_seconds=cascade.get("timeout_seconds", 30),
+            max_retries_per_provider=cascade.get("max_retries_per_provider", 2),
+            retry_delay_seconds=cascade.get("retry_delay_seconds", 1),
+            circuit_breaker_failure_threshold=cb.get("failure_threshold", 5),
+            circuit_breaker_recovery_timeout=cb.get("recovery_timeout", 60),
+            circuit_breaker_half_open_max_calls=cb.get("half_open_max_calls", 2),
+            auto_pull_models=multimodal.get("auto_pull", True),
+            fallback_chains=data.get("fallback_chains", {}),
+        )
+
+    def _load_providers(self, data: dict) -> None:
+        """Load, filter, and sort providers from parsed YAML data."""
+        for p_data in data.get("providers", []):
+            if not p_data.get("enabled", False):
+                continue
+
+            provider = Provider(
+                name=p_data["name"],
+                type=p_data["type"],
+                enabled=p_data.get("enabled", True),
+                priority=p_data.get("priority", 99),
+                url=p_data.get("url"),
+                api_key=p_data.get("api_key"),
+                base_url=p_data.get("base_url"),
+                models=p_data.get("models", []),
+            )
+
+            if self._check_provider_available(provider):
+                self.providers.append(provider)
+            else:
+                logger.warning("Provider %s not available, skipping", provider.name)
+
+        self.providers.sort(key=lambda p: p.priority)
+
     def _expand_env_vars(self, content: str) -> str:
         """Expand ${VAR} syntax in YAML content.
 
@@ -564,6 +555,7 @@ class CascadeRouter:
                 messages=messages,
                 model=model or provider.get_default_model(),
                 temperature=temperature,
+                max_tokens=max_tokens,
                 content_type=content_type,
             )
         elif provider.type == "openai":
@@ -604,6 +596,7 @@ class CascadeRouter:
         messages: list[dict],
         model: str,
         temperature: float,
+        max_tokens: int | None = None,
         content_type: ContentType = ContentType.TEXT,
     ) -> dict:
         """Call Ollama API with multi-modal support."""
@@ -614,13 +607,15 @@ class CascadeRouter:
         # Transform messages for Ollama format (including images)
         transformed_messages = self._transform_messages_for_ollama(messages)
 
+        options = {"temperature": temperature}
+        if max_tokens:
+            options["num_predict"] = max_tokens
+
         payload = {
             "model": model,
             "messages": transformed_messages,
             "stream": False,
-            "options": {
-                "temperature": temperature,
-            },
+            "options": options,
         }
 
         timeout = aiohttp.ClientTimeout(total=self.config.timeout_seconds)
@@ -764,7 +759,7 @@ class CascadeRouter:
 
         client = openai.AsyncOpenAI(
             api_key=provider.api_key,
-            base_url=provider.base_url or "https://api.x.ai/v1",
+            base_url=provider.base_url or settings.xai_base_url,
             timeout=httpx.Timeout(300.0),
         )
 

src/infrastructure/router/history.py

@@ -0,0 +1,152 @@
+"""Provider health history — time-series snapshots for dashboard visualization."""
+
+import asyncio
+import logging
+import sqlite3
+from datetime import UTC, datetime, timedelta
+from pathlib import Path
+
+logger = logging.getLogger(__name__)
+
+_store: "HealthHistoryStore | None" = None
+
+
+class HealthHistoryStore:
+    """Stores timestamped provider health snapshots in SQLite."""
+
+    def __init__(self, db_path: str = "data/router_history.db") -> None:
+        self.db_path = db_path
+        if db_path != ":memory:":
+            Path(db_path).parent.mkdir(parents=True, exist_ok=True)
+        self._conn = sqlite3.connect(db_path, check_same_thread=False)
+        self._conn.row_factory = sqlite3.Row
+        self._init_schema()
+        self._bg_task: asyncio.Task | None = None
+
+    def _init_schema(self) -> None:
+        self._conn.execute("""
+            CREATE TABLE IF NOT EXISTS snapshots (
+                id INTEGER PRIMARY KEY AUTOINCREMENT,
+                timestamp TEXT NOT NULL,
+                provider_name TEXT NOT NULL,
+                status TEXT NOT NULL,
+                error_rate REAL NOT NULL,
+                avg_latency_ms REAL NOT NULL,
+                circuit_state TEXT NOT NULL,
+                total_requests INTEGER NOT NULL
+            )
+        """)
+        self._conn.execute("""
+            CREATE INDEX IF NOT EXISTS idx_snapshots_ts
+            ON snapshots(timestamp)
+        """)
+        self._conn.commit()
+
+    def record_snapshot(self, providers: list[dict]) -> None:
+        """Record a health snapshot for all providers."""
+        ts = datetime.now(UTC).isoformat()
+        rows = [
+            (
+                ts,
+                p["name"],
+                p["status"],
+                p["error_rate"],
+                p["avg_latency_ms"],
+                p["circuit_state"],
+                p["total_requests"],
+            )
+            for p in providers
+        ]
+        self._conn.executemany(
+            """INSERT INTO snapshots
+               (timestamp, provider_name, status, error_rate,
+                avg_latency_ms, circuit_state, total_requests)
+               VALUES (?, ?, ?, ?, ?, ?, ?)""",
+            rows,
+        )
+        self._conn.commit()
+
+    def get_history(self, hours: int = 24) -> list[dict]:
+        """Return snapshots from the last N hours, grouped by timestamp."""
+        cutoff = (datetime.now(UTC) - timedelta(hours=hours)).isoformat()
+        rows = self._conn.execute(
+            """SELECT timestamp, provider_name, status, error_rate,
+                      avg_latency_ms, circuit_state, total_requests
+               FROM snapshots WHERE timestamp >= ? ORDER BY timestamp""",
+            (cutoff,),
+        ).fetchall()
+
+        # Group by timestamp
+        snapshots: dict[str, list[dict]] = {}
+        for row in rows:
+            ts = row["timestamp"]
+            if ts not in snapshots:
+                snapshots[ts] = []
+            snapshots[ts].append(
+                {
+                    "name": row["provider_name"],
+                    "status": row["status"],
+                    "error_rate": row["error_rate"],
+                    "avg_latency_ms": row["avg_latency_ms"],
+                    "circuit_state": row["circuit_state"],
+                    "total_requests": row["total_requests"],
+                }
+            )
+
+        return [{"timestamp": ts, "providers": providers} for ts, providers in snapshots.items()]
+
+    def prune(self, keep_hours: int = 168) -> int:
+        """Remove snapshots older than keep_hours. Returns rows deleted."""
+        cutoff = (datetime.now(UTC) - timedelta(hours=keep_hours)).isoformat()
+        cursor = self._conn.execute("DELETE FROM snapshots WHERE timestamp < ?", (cutoff,))
+        self._conn.commit()
+        return cursor.rowcount
+
+    def close(self) -> None:
+        """Close the database connection."""
+        if self._bg_task and not self._bg_task.done():
+            self._bg_task.cancel()
+        self._conn.close()
+
+    def _capture_snapshot(self, cascade_router) -> None:  # noqa: ANN001
+        """Capture current provider state as a snapshot."""
+        providers = []
+        for p in cascade_router.providers:
+            providers.append(
+                {
+                    "name": p.name,
+                    "status": p.status.value,
+                    "error_rate": round(p.metrics.error_rate, 4),
+                    "avg_latency_ms": round(p.metrics.avg_latency_ms, 2),
+                    "circuit_state": p.circuit_state.value,
+                    "total_requests": p.metrics.total_requests,
+                }
+            )
+        self.record_snapshot(providers)
+
+    async def start_background_task(
+        self,
+        cascade_router,
+        interval_seconds: int = 60,  # noqa: ANN001
+    ) -> None:
+        """Start periodic snapshot capture."""
+
+        async def _loop() -> None:
+            while True:
+                try:
+                    self._capture_snapshot(cascade_router)
+                    logger.debug("Recorded health snapshot")
+                except Exception:
+                    logger.exception("Failed to record health snapshot")
+                await asyncio.sleep(interval_seconds)
+
+        self._bg_task = asyncio.create_task(_loop())
+        logger.info("Health history background task started (interval=%ds)", interval_seconds)
+
+
+def get_history_store() -> HealthHistoryStore:
+    """Get or create the singleton history store."""
+    global _store  # noqa: PLW0603
+    if _store is None:
+        _store = HealthHistoryStore()
+    return _store

src/lightning/__init__.py

@@ -0,0 +1 @@
+"""Lightning Network integration for tool-usage micro-payments."""

src/lightning/factory.py

@@ -0,0 +1,69 @@
+"""Lightning backend factory.
+
+Returns a mock or real LND backend based on ``settings.lightning_backend``.
+"""
+
+from __future__ import annotations
+
+import hashlib
+import logging
+import secrets
+from dataclasses import dataclass
+
+from config import settings
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class Invoice:
+    """Minimal Lightning invoice representation."""
+
+    payment_hash: str
+    payment_request: str
+    amount_sats: int
+    memo: str
+
+
+class MockBackend:
+    """In-memory mock Lightning backend for development and testing."""
+
+    def create_invoice(self, amount_sats: int, memo: str = "") -> Invoice:
+        """Create a fake invoice with a random payment hash."""
+        raw = secrets.token_bytes(32)
+        payment_hash = hashlib.sha256(raw).hexdigest()
+        payment_request = f"lnbc{amount_sats}mock{payment_hash[:20]}"
+        logger.debug("Mock invoice: %s sats — %s", amount_sats, payment_hash[:12])
+        return Invoice(
+            payment_hash=payment_hash,
+            payment_request=payment_request,
+            amount_sats=amount_sats,
+            memo=memo,
+        )
+
+
+# Singleton — lazily created
+_backend: MockBackend | None = None
+
+
+def get_backend() -> MockBackend:
+    """Return the configured Lightning backend (currently mock-only).
+
+    Raises ``ValueError`` if an unsupported backend is requested.
+    """
+    global _backend  # noqa: PLW0603
+    if _backend is not None:
+        return _backend
+
+    kind = settings.lightning_backend
+    if kind == "mock":
+        _backend = MockBackend()
+    elif kind == "lnd":
+        # LND gRPC integration is on the roadmap — for now fall back to mock.
+        logger.warning("LND backend not yet implemented — using mock")
+        _backend = MockBackend()
+    else:
+        raise ValueError(f"Unknown lightning_backend: {kind!r}")
+
+    logger.info("Lightning backend: %s", kind)
+    return _backend

src/lightning/ledger.py

@@ -0,0 +1,146 @@
+"""In-memory Lightning transaction ledger.
+
+Tracks invoices, settlements, and balances per the schema in
+``docs/adr/018-lightning-ledger.md``.  Uses a simple in-memory list so the
+dashboard can display real (ephemeral) data without requiring SQLite yet.
+"""
+
+from __future__ import annotations
+
+import logging
+import uuid
+from dataclasses import dataclass
+from datetime import UTC, datetime
+from enum import StrEnum
+
+logger = logging.getLogger(__name__)
+
+
+class TxType(StrEnum):
+    incoming = "incoming"
+    outgoing = "outgoing"
+
+
+class TxStatus(StrEnum):
+    pending = "pending"
+    settled = "settled"
+    failed = "failed"
+    expired = "expired"
+
+
+@dataclass
+class LedgerEntry:
+    """Single ledger row matching the ADR-018 schema."""
+
+    id: str
+    tx_type: TxType
+    status: TxStatus
+    payment_hash: str
+    amount_sats: int
+    memo: str
+    source: str
+    created_at: str
+    invoice: str = ""
+    preimage: str = ""
+    task_id: str = ""
+    agent_id: str = ""
+    settled_at: str = ""
+    fee_sats: int = 0
+
+
+# ── In-memory store ──────────────────────────────────────────────────
+_entries: list[LedgerEntry] = []
+
+
+def create_invoice_entry(
+    payment_hash: str,
+    amount_sats: int,
+    memo: str = "",
+    source: str = "tool_usage",
+    task_id: str = "",
+    agent_id: str = "",
+    invoice: str = "",
+) -> LedgerEntry:
+    """Record a new incoming invoice in the ledger."""
+    entry = LedgerEntry(
+        id=uuid.uuid4().hex[:16],
+        tx_type=TxType.incoming,
+        status=TxStatus.pending,
+        payment_hash=payment_hash,
+        amount_sats=amount_sats,
+        memo=memo,
+        source=source,
+        task_id=task_id,
+        agent_id=agent_id,
+        invoice=invoice,
+        created_at=datetime.now(UTC).isoformat(),
+    )
+    _entries.append(entry)
+    logger.debug("Ledger entry created: %s (%s sats)", entry.id, amount_sats)
+    return entry
+
+
+def mark_settled(payment_hash: str, preimage: str = "") -> LedgerEntry | None:
+    """Mark a pending entry as settled by payment hash."""
+    for entry in _entries:
+        if entry.payment_hash == payment_hash and entry.status == TxStatus.pending:
+            entry.status = TxStatus.settled
+            entry.preimage = preimage
+            entry.settled_at = datetime.now(UTC).isoformat()
+            logger.debug("Ledger settled: %s", payment_hash[:12])
+            return entry
+    return None
+
+
+def get_balance() -> dict:
+    """Compute the current balance from settled and pending entries."""
+    incoming_total = sum(
+        e.amount_sats
+        for e in _entries
+        if e.tx_type == TxType.incoming and e.status == TxStatus.settled
+    )
+    outgoing_total = sum(
+        e.amount_sats
+        for e in _entries
+        if e.tx_type == TxType.outgoing and e.status == TxStatus.settled
+    )
+    fees = sum(e.fee_sats for e in _entries if e.status == TxStatus.settled)
+    pending_in = sum(
+        e.amount_sats
+        for e in _entries
+        if e.tx_type == TxType.incoming and e.status == TxStatus.pending
+    )
+    pending_out = sum(
+        e.amount_sats
+        for e in _entries
+        if e.tx_type == TxType.outgoing and e.status == TxStatus.pending
+    )
+    net = incoming_total - outgoing_total - fees
+    return {
+        "incoming_total_sats": incoming_total,
+        "outgoing_total_sats": outgoing_total,
+        "fees_paid_sats": fees,
+        "net_sats": net,
+        "pending_incoming_sats": pending_in,
+        "pending_outgoing_sats": pending_out,
+        "available_sats": net - pending_out,
+    }
+
+
+def get_transactions(
+    tx_type: str | None = None,
+    status: str | None = None,
+    limit: int = 50,
+) -> list[LedgerEntry]:
+    """Return ledger entries, optionally filtered."""
+    result = _entries
+    if tx_type:
+        result = [e for e in result if e.tx_type.value == tx_type]
+    if status:
+        result = [e for e in result if e.status.value == status]
+    return list(reversed(result))[:limit]
+
+
+def clear() -> None:
+    """Reset the ledger (for testing)."""
+    _entries.clear()

src/timmy/backends.py

@@ -99,23 +99,27 @@ class GrokBackend:
 
     def _get_client(self):
         """Create OpenAI client configured for xAI endpoint."""
+        from config import settings
+
         import httpx
         from openai import OpenAI
 
         return OpenAI(
             api_key=self._api_key,
-            base_url="https://api.x.ai/v1",
+            base_url=settings.xai_base_url,
             timeout=httpx.Timeout(300.0),
         )
 
     async def _get_async_client(self):
         """Create async OpenAI client configured for xAI endpoint."""
+        from config import settings
+
         import httpx
         from openai import AsyncOpenAI
 
         return AsyncOpenAI(
             api_key=self._api_key,
-            base_url="https://api.x.ai/v1",
+            base_url=settings.xai_base_url,
             timeout=httpx.Timeout(300.0),
         )
 

src/timmy/conversation.py

@@ -174,15 +174,8 @@ class ConversationManager:
 
         return None
 
-    def should_use_tools(self, message: str, context: ConversationContext) -> bool:
-        """Determine if this message likely requires tools.
-
-        Returns True if tools are likely needed, False for simple chat.
-        """
-        message_lower = message.lower().strip()
-
-        # Tool keywords that suggest tool usage is needed
-        tool_keywords = [
+    _TOOL_KEYWORDS = frozenset(
+        {
             "search",
             "look up",
             "find",
@@ -203,10 +196,11 @@ class ConversationManager:
             "shell",
             "command",
             "install",
-        ]
+        }
+    )
 
-        # Chat-only keywords that definitely don't need tools
-        chat_only = [
+    _CHAT_ONLY_KEYWORDS = frozenset(
+        {
             "hello",
             "hi ",
             "hey",
@@ -221,30 +215,47 @@ class ConversationManager:
             "goodbye",
             "tell me about yourself",
             "what can you do",
-        ]
+        }
+    )
 
-        # Check for chat-only patterns first
-        for pattern in chat_only:
-            if pattern in message_lower:
-                return False
+    _SIMPLE_QUESTION_PREFIXES = ("what is", "who is", "how does", "why is", "when did", "where is")
+    _TIME_WORDS = ("today", "now", "current", "latest", "this week", "this month")
 
-        # Check for tool keywords
-        for keyword in tool_keywords:
-            if keyword in message_lower:
-                return True
+    def _is_chat_only(self, message_lower: str) -> bool:
+        """Return True if the message matches a chat-only pattern."""
+        return any(kw in message_lower for kw in self._CHAT_ONLY_KEYWORDS)
 
-        # Simple questions (starting with what, who, how, why, when, where)
-        # usually don't need tools unless about current/real-time info
-        simple_question_words = ["what is", "who is", "how does", "why is", "when did", "where is"]
-        for word in simple_question_words:
-            if message_lower.startswith(word):
-                # Check if it's asking about current/real-time info
-                time_words = ["today", "now", "current", "latest", "this week", "this month"]
-                if any(t in message_lower for t in time_words):
-                    return True
-                return False
+    def _has_tool_keyword(self, message_lower: str) -> bool:
+        """Return True if the message contains a tool-related keyword."""
+        return any(kw in message_lower for kw in self._TOOL_KEYWORDS)
+
+    def _is_simple_question(self, message_lower: str) -> bool | None:
+        """Check if message is a simple question.
+
+        Returns True if it needs tools (real-time info), False if it
+        doesn't, or None if the message isn't a simple question.
+        """
+        for prefix in self._SIMPLE_QUESTION_PREFIXES:
+            if message_lower.startswith(prefix):
+                return any(t in message_lower for t in self._TIME_WORDS)
+        return None
+
+    def should_use_tools(self, message: str, context: ConversationContext) -> bool:
+        """Determine if this message likely requires tools.
+
+        Returns True if tools are likely needed, False for simple chat.
+        """
+        message_lower = message.lower().strip()
+
+        if self._is_chat_only(message_lower):
+            return False
+        if self._has_tool_keyword(message_lower):
+            return True
+
+        simple = self._is_simple_question(message_lower)
+        if simple is not None:
+            return simple
 
-        # Default: don't use tools for unclear cases
         return False
 
 

src/timmy/mcp_tools.py

@@ -30,7 +30,7 @@ import shutil
 import sqlite3
 import uuid
 from contextlib import closing
-from datetime import datetime
+from datetime import UTC, datetime
 from pathlib import Path
 
 import httpx
@@ -196,7 +196,7 @@ def _bridge_to_work_order(title: str, body: str, category: str) -> None:
                     body,
                     category,
                     "timmy-thinking",
-                    datetime.utcnow().isoformat(),
+                    datetime.now(UTC).isoformat(),
                 ),
             )
             conn.commit()
@@ -204,15 +204,61 @@ def _bridge_to_work_order(title: str, body: str, category: str) -> None:
         logger.debug("Work order bridge failed: %s", exc)
 
 
+async def _ensure_issue_session():
+    """Get or create the cached MCP session, connecting if needed.
+
+    Returns the connected ``MCPTools`` instance.
+    """
+    from agno.tools.mcp import MCPTools
+
+    global _issue_session
+
+    if _issue_session is None:
+        _issue_session = MCPTools(
+            server_params=_gitea_server_params(),
+            timeout_seconds=settings.mcp_timeout,
+        )
+
+    if not getattr(_issue_session, "_connected", False):
+        await _issue_session.connect()
+        _issue_session._connected = True
+
+    return _issue_session
+
+
+def _build_issue_body(body: str) -> str:
+    """Append the auto-filing signature to the issue body."""
+    full_body = body
+    if full_body:
+        full_body += "\n\n"
+    full_body += "---\n*Auto-filed by Timmy's thinking engine*"
+    return full_body
+
+
+def _build_issue_args(title: str, full_body: str) -> dict:
+    """Build MCP tool arguments for ``issue_write`` with method=create."""
+    owner, repo = settings.gitea_repo.split("/", 1)
+    return {
+        "method": "create",
+        "owner": owner,
+        "repo": repo,
+        "title": title,
+        "body": full_body,
+    }
+
+
+def _category_from_labels(labels: str) -> str:
+    """Derive a work-order category from comma-separated label names."""
+    label_list = [tag.strip() for tag in labels.split(",") if tag.strip()] if labels else []
+    return "bug" if "bug" in label_list else "suggestion"
+
+
 async def create_gitea_issue_via_mcp(title: str, body: str = "", labels: str = "") -> str:
     """File a Gitea issue via the MCP server (standalone, no LLM loop).
 
     Used by the thinking engine's ``_maybe_file_issues()`` post-hook.
     Manages its own MCPTools session with lazy connect + graceful failure.
 
-    Uses ``tools.session.call_tool()`` for direct MCP invocation — the
-    ``MCPTools`` wrapper itself does not expose ``call_tool()``.
-
     Args:
         title: Issue title.
         body: Issue body (markdown).
@@ -225,46 +271,13 @@ async def create_gitea_issue_via_mcp(title: str, body: str = "", labels: str = "
         return "Gitea integration is not configured."
 
     try:
-        from agno.tools.mcp import MCPTools
+        session = await _ensure_issue_session()
+        full_body = _build_issue_body(body)
+        args = _build_issue_args(title, full_body)
 
-        global _issue_session
+        result = await session.session.call_tool("issue_write", arguments=args)
 
-        if _issue_session is None:
-            _issue_session = MCPTools(
-                server_params=_gitea_server_params(),
-                timeout_seconds=settings.mcp_timeout,
-            )
-
-        # Ensure connected
-        if not getattr(_issue_session, "_connected", False):
-            await _issue_session.connect()
-            _issue_session._connected = True
-
-        # Append auto-filing signature
-        full_body = body
-        if full_body:
-            full_body += "\n\n"
-        full_body += "---\n*Auto-filed by Timmy's thinking engine*"
-
-        # Parse owner/repo from settings
-        owner, repo = settings.gitea_repo.split("/", 1)
-
-        # Build tool arguments — gitea-mcp uses issue_write with method="create"
-        args = {
-            "method": "create",
-            "owner": owner,
-            "repo": repo,
-            "title": title,
-            "body": full_body,
-        }
-
-        # Call via the underlying MCP session (MCPTools doesn't expose call_tool)
-        result = await _issue_session.session.call_tool("issue_write", arguments=args)
-
-        # Bridge to local work order
-        label_list = [tag.strip() for tag in labels.split(",") if tag.strip()] if labels else []
-        category = "bug" if "bug" in label_list else "suggestion"
-        _bridge_to_work_order(title, body, category)
+        _bridge_to_work_order(title, body, _category_from_labels(labels))
 
         logger.info("Created Gitea issue via MCP: %s", title[:60])
         return f"Created issue: {title}\n{result}"

src/timmy/memory_system.py

@@ -46,6 +46,64 @@ DB_PATH = PROJECT_ROOT / "data" / "memory.db"
 # ───────────────────────────────────────────────────────────────────────────────
 
 
+_DEFAULT_HOT_MEMORY_TEMPLATE = """\
+# Timmy Hot Memory
+
+> Working RAM — always loaded, ~300 lines max, pruned monthly
+> Last updated: {date}
+
+---
+
+## Current Status
+
+**Agent State:** Operational
+**Mode:** Development
+**Active Tasks:** 0
+**Pending Decisions:** None
+
+---
+
+## Standing Rules
+
+1. **Sovereignty First** — No cloud dependencies
+2. **Local-Only Inference** — Ollama on localhost
+3. **Privacy by Design** — Telemetry disabled
+4. **Tool Minimalism** — Use tools only when necessary
+5. **Memory Discipline** — Write handoffs at session end
+
+---
+
+## Agent Roster
+
+| Agent | Role | Status |
+|-------|------|--------|
+| Timmy | Core | Active |
+
+---
+
+## User Profile
+
+**Name:** (not set)
+**Interests:** (to be learned)
+
+---
+
+## Key Decisions
+
+(none yet)
+
+---
+
+## Pending Actions
+
+- [ ] Learn user's name
+
+---
+
+*Prune date: {prune_date}*
+"""
+
+
 @contextmanager
 def get_connection() -> Generator[sqlite3.Connection, None, None]:
     """Get database connection to unified memory database."""
@@ -732,66 +790,12 @@ class HotMemory:
         logger.debug(
             "HotMemory._create_default() - creating default MEMORY.md for backward compatibility"
         )
-        default_content = """# Timmy Hot Memory
-
-> Working RAM — always loaded, ~300 lines max, pruned monthly
-> Last updated: {date}
-
----
-
-## Current Status
-
-**Agent State:** Operational
-**Mode:** Development
-**Active Tasks:** 0
-**Pending Decisions:** None
-
----
-
-## Standing Rules
-
-1. **Sovereignty First** — No cloud dependencies
-2. **Local-Only Inference** — Ollama on localhost
-3. **Privacy by Design** — Telemetry disabled
-4. **Tool Minimalism** — Use tools only when necessary
-5. **Memory Discipline** — Write handoffs at session end
-
----
-
-## Agent Roster
-
-| Agent | Role | Status |
-|-------|------|--------|
-| Timmy | Core | Active |
-
----
-
-## User Profile
-
-**Name:** (not set)
-**Interests:** (to be learned)
-
----
-
-## Key Decisions
-
-(none yet)
-
----
-
-## Pending Actions
-
-- [ ] Learn user's name
-
----
-
-*Prune date: {prune_date}*
-""".format(
-            date=datetime.now(UTC).strftime("%Y-%m-%d"),
-            prune_date=(datetime.now(UTC).replace(day=25)).strftime("%Y-%m-%d"),
+        now = datetime.now(UTC)
+        content = _DEFAULT_HOT_MEMORY_TEMPLATE.format(
+            date=now.strftime("%Y-%m-%d"),
+            prune_date=now.replace(day=25).strftime("%Y-%m-%d"),
         )
-
-        self.path.write_text(default_content)
+        self.path.write_text(content)
         logger.info("HotMemory: Created default MEMORY.md")
 
 

src/timmy/session_logger.py

@@ -392,31 +392,26 @@ def _build_insights(
     return insights or ["Conversations look healthy. Keep up the good work."]
 
 
-def self_reflect(limit: int = 30) -> str:
-    """Review recent conversations and reflect on Timmy's own behavior.
+def _format_recurring_topics(repeated: list[tuple[str, int]]) -> list[str]:
+    """Format the recurring-topics section of a reflection report."""
+    if repeated:
+        lines = ["### Recurring Topics"]
+        for word, count in repeated:
+            lines.append(f'- "{word}" ({count} mentions)')
+        lines.append("")
+        return lines
+    return ["### Recurring Topics\nNo strong patterns detected.\n"]
 
-    Scans past session entries for patterns: low-confidence responses,
-    errors, repeated topics, and conversation quality signals.  Returns
-    a structured reflection that Timmy can use to improve.
 
-    Args:
-        limit: How many recent entries to review (default 30).
-
-    Returns:
-        A formatted self-reflection report.
-    """
-    sl = get_session_logger()
-    sl.flush()
-    entries = sl.get_recent_entries(limit=limit)
-
-    if not entries:
-        return "No conversation history to reflect on yet."
-
-    _messages, errors, timmy_msgs, user_msgs = _categorize_entries(entries)
-    low_conf = _find_low_confidence(timmy_msgs)
-    repeated = _find_repeated_topics(user_msgs)
-
-    # Build reflection report
+def _assemble_report(
+    entries: list[dict],
+    errors: list[dict],
+    timmy_msgs: list[dict],
+    user_msgs: list[dict],
+    low_conf: list[dict],
+    repeated: list[tuple[str, int]],
+) -> str:
+    """Assemble the full self-reflection report from analyzed data."""
     sections: list[str] = ["## Self-Reflection Report\n"]
     sections.append(
         f"Reviewed {len(entries)} recent entries: "
@@ -446,16 +441,37 @@ def self_reflect(limit: int = 30) -> str:
         )
     )
 
-    if repeated:
-        sections.append("### Recurring Topics")
-        for word, count in repeated:
-            sections.append(f'- "{word}" ({count} mentions)')
-        sections.append("")
-    else:
-        sections.append("### Recurring Topics\nNo strong patterns detected.\n")
+    sections.extend(_format_recurring_topics(repeated))
 
     sections.append("### Insights")
     for insight in _build_insights(low_conf, errors, repeated):
         sections.append(f"- {insight}")
 
     return "\n".join(sections)
+
+
+def self_reflect(limit: int = 30) -> str:
+    """Review recent conversations and reflect on Timmy's own behavior.
+
+    Scans past session entries for patterns: low-confidence responses,
+    errors, repeated topics, and conversation quality signals.  Returns
+    a structured reflection that Timmy can use to improve.
+
+    Args:
+        limit: How many recent entries to review (default 30).
+
+    Returns:
+        A formatted self-reflection report.
+    """
+    sl = get_session_logger()
+    sl.flush()
+    entries = sl.get_recent_entries(limit=limit)
+
+    if not entries:
+        return "No conversation history to reflect on yet."
+
+    _messages, errors, timmy_msgs, user_msgs = _categorize_entries(entries)
+    low_conf = _find_low_confidence(timmy_msgs)
+    repeated = _find_repeated_topics(user_msgs)
+
+    return _assemble_report(entries, errors, timmy_msgs, user_msgs, low_conf, repeated)

src/timmy/tools_delegation/__init__.py

@@ -89,45 +89,31 @@ def list_swarm_agents() -> dict[str, Any]:
         }
 
 
-def delegate_to_kimi(task: str, working_directory: str = "") -> dict[str, Any]:
-    """Delegate a coding task to Kimi, the external coding agent.
-
-    Kimi has 262K context and is optimized for code tasks: writing,
-    debugging, refactoring, test writing. Timmy thinks and plans,
-    Kimi executes bulk code changes.
-
-    Args:
-        task: Clear, specific coding task description. Include file paths
-              and expected behavior. Good: "Fix the bug in src/timmy/session.py
-              where sessions don't persist." Bad: "Fix all bugs."
-        working_directory: Directory for Kimi to work in. Defaults to repo root.
-
-    Returns:
-        Dict with success status and Kimi's output or error.
-    """
+def _find_kimi_cli() -> str | None:
+    """Return the path to the kimi CLI binary, or None if not installed."""
     import shutil
-    import subprocess
+
+    return shutil.which("kimi")
+
+
+def _resolve_workdir(working_directory: str) -> str | dict[str, Any]:
+    """Return a validated working directory path, or an error dict."""
     from pathlib import Path
 
     from config import settings
 
-    kimi_path = shutil.which("kimi")
-    if not kimi_path:
-        return {
-            "success": False,
-            "error": "kimi CLI not found on PATH. Install with: pip install kimi-cli",
-        }
-
     workdir = working_directory or settings.repo_root
     if not Path(workdir).is_dir():
         return {
             "success": False,
             "error": f"Working directory does not exist: {workdir}",
         }
+    return workdir
 
-    cmd = [kimi_path, "--print", "-p", task]
 
-    logger.info("Delegating to Kimi: %s (cwd=%s)", task[:80], workdir)
+def _run_kimi(cmd: list[str], workdir: str) -> dict[str, Any]:
+    """Execute the kimi subprocess and return a result dict."""
+    import subprocess
 
     try:
         result = subprocess.run(
@@ -157,3 +143,34 @@ def delegate_to_kimi(task: str, working_directory: str = "") -> dict[str, Any]:
             "success": False,
             "error": f"Failed to run Kimi: {exc}",
         }
+
+
+def delegate_to_kimi(task: str, working_directory: str = "") -> dict[str, Any]:
+    """Delegate a coding task to Kimi, the external coding agent.
+
+    Kimi has 262K context and is optimized for code tasks: writing,
+    debugging, refactoring, test writing. Timmy thinks and plans,
+    Kimi executes bulk code changes.
+
+    Args:
+        task: Clear, specific coding task description. Include file paths
+              and expected behavior. Good: "Fix the bug in src/timmy/session.py
+              where sessions don't persist." Bad: "Fix all bugs."
+        working_directory: Directory for Kimi to work in. Defaults to repo root.
+
+    Returns:
+        Dict with success status and Kimi's output or error.
+    """
+    kimi_path = _find_kimi_cli()
+    if not kimi_path:
+        return {
+            "success": False,
+            "error": "kimi CLI not found on PATH. Install with: pip install kimi-cli",
+        }
+
+    workdir = _resolve_workdir(working_directory)
+    if isinstance(workdir, dict):
+        return workdir
+
+    logger.info("Delegating to Kimi: %s (cwd=%s)", task[:80], workdir)
+    return _run_kimi([kimi_path, "--print", "-p", task], workdir)

src/timmy/workshop_state.py

@@ -86,6 +86,40 @@ def _pip_snapshot(mood: str, confidence: float) -> dict:
     return pip_familiar.snapshot().to_dict()
 
 
+def _resolve_mood(state) -> str:
+    """Map cognitive mood/engagement to a presence mood string."""
+    if state.engagement == "idle" and state.mood == "settled":
+        return "calm"
+    return _MOOD_MAP.get(state.mood, "calm")
+
+
+def _resolve_confidence(state) -> float:
+    """Compute normalised confidence from cognitive tracker state."""
+    if state._confidence_count > 0:
+        raw = state._confidence_sum / state._confidence_count
+    else:
+        raw = 0.7
+    return round(max(0.0, min(1.0, raw)), 2)
+
+
+def _build_active_threads(state) -> list[dict]:
+    """Convert active commitments into presence thread dicts."""
+    return [
+        {"type": "thinking", "ref": c[:80], "status": "active"}
+        for c in state.active_commitments[:10]
+    ]
+
+
+def _build_environment() -> dict:
+    """Return the environment section using local wall-clock time."""
+    local_now = datetime.now()
+    return {
+        "time_of_day": _time_of_day(local_now.hour),
+        "local_time": local_now.strftime("%-I:%M %p"),
+        "day_of_week": local_now.strftime("%A"),
+    }
+
+
 def get_state_dict() -> dict:
     """Build presence state dict from current cognitive state.
 
@@ -98,37 +132,19 @@ def get_state_dict() -> dict:
     state = cognitive_tracker.get_state()
     now = datetime.now(UTC)
 
-    # Map cognitive mood to presence mood
-    mood = _MOOD_MAP.get(state.mood, "calm")
-    if state.engagement == "idle" and state.mood == "settled":
-        mood = "calm"
-
-    # Confidence from cognitive tracker
-    if state._confidence_count > 0:
-        confidence = state._confidence_sum / state._confidence_count
-    else:
-        confidence = 0.7
-
-    # Build active threads from commitments
-    threads = []
-    for commitment in state.active_commitments[:10]:
-        threads.append({"type": "thinking", "ref": commitment[:80], "status": "active"})
-
-    # Activity
+    mood = _resolve_mood(state)
+    confidence = _resolve_confidence(state)
     activity = _ACTIVITY_MAP.get(state.engagement, "idle")
 
-    # Environment
-    local_now = datetime.now()
-
     return {
         "version": 1,
         "liveness": now.strftime("%Y-%m-%dT%H:%M:%SZ"),
         "current_focus": state.focus_topic or "",
-        "active_threads": threads,
+        "active_threads": _build_active_threads(state),
         "recent_events": [],
         "concerns": [],
         "mood": mood,
-        "confidence": round(max(0.0, min(1.0, confidence)), 2),
+        "confidence": confidence,
         "energy": round(_current_energy(), 2),
         "identity": {
             "name": "Timmy",
@@ -143,11 +159,7 @@ def get_state_dict() -> dict:
             "visitor_present": False,
             "conversation_turns": state.conversation_depth,
         },
-        "environment": {
-            "time_of_day": _time_of_day(local_now.hour),
-            "local_time": local_now.strftime("%-I:%M %p"),
-            "day_of_week": local_now.strftime("%A"),
-        },
+        "environment": _build_environment(),
         "familiar": _pip_snapshot(mood, confidence),
         "meta": {
             "schema_version": 1,

static/css/mission-control.css

@@ -2493,3 +2493,57 @@
 .db-cell { max-width: 300px; overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
 .db-cell:hover { white-space: normal; word-break: break-all; }
 .db-truncated { font-size: 0.7rem; color: var(--amber); padding: 0.3rem 0; }
+
+/* ── Tower ────────────────────────────────────────────────────────────── */
+.tower-container { max-width: 1400px; margin: 0 auto; }
+.tower-header { margin-bottom: 1rem; }
+.tower-title { font-size: 1.6rem; font-weight: 700; color: var(--green); letter-spacing: 0.15em; }
+.tower-subtitle { font-size: 0.85rem; color: var(--text-dim); }
+
+.tower-conn-badge { font-size: 0.7rem; font-weight: 600; padding: 2px 8px; border-radius: 3px; letter-spacing: 0.08em; }
+.tower-conn-live { color: var(--green); border: 1px solid var(--green); }
+.tower-conn-offline { color: var(--red); border: 1px solid var(--red); }
+.tower-conn-connecting { color: var(--amber); border: 1px solid var(--amber); }
+
+.tower-phase-card { min-height: 300px; }
+.tower-phase-thinking { border-left: 3px solid var(--purple); }
+.tower-phase-predicting { border-left: 3px solid var(--orange); }
+.tower-phase-advising { border-left: 3px solid var(--green); }
+.tower-scroll { max-height: 50vh; overflow-y: auto; }
+.tower-empty { text-align: center; color: var(--text-dim); padding: 16px; font-size: 0.85rem; }
+
+.tower-stat-grid { display: grid; grid-template-columns: repeat(4, 1fr); gap: 0.5rem; text-align: center; }
+.tower-stat-label { display: block; font-size: 0.65rem; color: var(--text-dim); letter-spacing: 0.1em; }
+.tower-stat-value { display: block; font-size: 1.1rem; font-weight: 700; color: var(--text-bright); }
+
+.tower-event { padding: 8px; margin-bottom: 6px; border-left: 3px solid var(--border); border-radius: 3px; background: var(--bg-card); }
+.tower-etype-task_posted { border-left-color: var(--purple); }
+.tower-etype-bid_submitted { border-left-color: var(--orange); }
+.tower-etype-task_completed { border-left-color: var(--green); }
+.tower-etype-task_failed { border-left-color: var(--red); }
+.tower-etype-agent_joined { border-left-color: var(--purple); }
+.tower-etype-tool_executed { border-left-color: var(--amber); }
+.tower-ev-head { display: flex; justify-content: space-between; align-items: center; margin-bottom: 4px; }
+.tower-ev-badge { font-size: 0.65rem; font-weight: 600; color: var(--text-bright); letter-spacing: 0.08em; }
+.tower-ev-dots { font-size: 0.6rem; color: var(--amber); }
+.tower-ev-desc { font-size: 0.8rem; color: var(--text); }
+.tower-ev-time { font-size: 0.65rem; color: var(--text-dim); margin-top: 2px; }
+
+.tower-pred { padding: 8px; margin-bottom: 6px; border-radius: 3px; background: var(--bg-card); border-left: 3px solid var(--orange); }
+.tower-pred-done { border-left-color: var(--green); }
+.tower-pred-pending { border-left-color: var(--amber); }
+.tower-pred-head { display: flex; justify-content: space-between; align-items: center; }
+.tower-pred-task { font-size: 0.75rem; font-weight: 600; color: var(--text-bright); font-family: monospace; }
+.tower-pred-acc { font-size: 0.75rem; font-weight: 700; }
+.tower-pred-detail { font-size: 0.75rem; color: var(--text-dim); margin-top: 4px; }
+
+.tower-advisory { padding: 8px; margin-bottom: 6px; border-radius: 3px; background: var(--bg-card); border-left: 3px solid var(--border); }
+.tower-adv-high { border-left-color: var(--red); }
+.tower-adv-medium { border-left-color: var(--orange); }
+.tower-adv-low { border-left-color: var(--green); }
+.tower-adv-head { display: flex; justify-content: space-between; font-size: 0.65rem; margin-bottom: 4px; }
+.tower-adv-cat { font-weight: 600; color: var(--text-dim); letter-spacing: 0.08em; }
+.tower-adv-prio { font-weight: 700; color: var(--amber); }
+.tower-adv-title { font-size: 0.85rem; font-weight: 600; color: var(--text-bright); }
+.tower-adv-detail { font-size: 0.8rem; color: var(--text); margin-top: 2px; }
+.tower-adv-action { font-size: 0.75rem; color: var(--green); margin-top: 4px; font-style: italic; }

tests/dashboard/middleware/test_csrf_no_side_effects.py

@@ -0,0 +1,100 @@
+"""Tests that CSRF rejection does NOT execute the endpoint handler.
+
+Regression test for #626: the middleware was calling call_next() before
+checking @csrf_exempt, causing side effects even on CSRF-rejected requests.
+"""
+
+import pytest
+from fastapi import FastAPI
+from fastapi.testclient import TestClient
+
+from dashboard.middleware.csrf import CSRFMiddleware, csrf_exempt
+
+
+class TestCSRFNoSideEffects:
+    """Verify endpoints are NOT executed when CSRF validation fails."""
+
+    @pytest.fixture(autouse=True)
+    def enable_csrf(self):
+        """Re-enable CSRF for these tests."""
+        from config import settings
+
+        original = settings.timmy_disable_csrf
+        settings.timmy_disable_csrf = False
+        yield
+        settings.timmy_disable_csrf = original
+
+    def test_protected_endpoint_not_executed_on_csrf_failure(self):
+        """A protected endpoint must NOT run when CSRF token is missing.
+
+        Before the fix, the middleware called call_next() to resolve the
+        endpoint, executing its side effects before returning 403.
+        """
+        app = FastAPI()
+        app.add_middleware(CSRFMiddleware)
+
+        side_effect_log = []
+
+        @app.post("/transfer")
+        def transfer_money():
+            side_effect_log.append("money_transferred")
+            return {"message": "transferred"}
+
+        client = TestClient(app)
+        response = client.post("/transfer")
+
+        assert response.status_code == 403
+        assert side_effect_log == [], (
+            "Endpoint was executed despite CSRF failure — side effects occurred!"
+        )
+
+    def test_csrf_exempt_endpoint_still_executes(self):
+        """A @csrf_exempt endpoint should still execute without a CSRF token."""
+        app = FastAPI()
+        app.add_middleware(CSRFMiddleware)
+
+        side_effect_log = []
+
+        @app.post("/webhook-handler")
+        @csrf_exempt
+        def webhook_handler():
+            side_effect_log.append("webhook_processed")
+            return {"message": "processed"}
+
+        client = TestClient(app)
+        response = client.post("/webhook-handler")
+
+        assert response.status_code == 200
+        assert side_effect_log == ["webhook_processed"]
+
+    def test_exempt_and_protected_no_cross_contamination(self):
+        """Mixed exempt/protected: only exempt endpoints execute without tokens."""
+        app = FastAPI()
+        app.add_middleware(CSRFMiddleware)
+
+        execution_log = []
+
+        @app.post("/safe-webhook")
+        @csrf_exempt
+        def safe_webhook():
+            execution_log.append("safe")
+            return {"message": "safe"}
+
+        @app.post("/dangerous-action")
+        def dangerous_action():
+            execution_log.append("dangerous")
+            return {"message": "danger"}
+
+        client = TestClient(app)
+
+        # Exempt endpoint runs
+        resp1 = client.post("/safe-webhook")
+        assert resp1.status_code == 200
+
+        # Protected endpoint blocked WITHOUT executing
+        resp2 = client.post("/dangerous-action")
+        assert resp2.status_code == 403
+
+        assert execution_log == ["safe"], (
+            f"Expected only 'safe' execution, got: {execution_log}"
+        )

tests/dashboard/test_tower.py

@@ -0,0 +1,187 @@
+"""Tests for Tower dashboard route (/tower)."""
+
+from unittest.mock import MagicMock, patch
+
+
+def _mock_spark_engine():
+    """Return a mock spark_engine with realistic return values."""
+    engine = MagicMock()
+
+    engine.status.return_value = {
+        "enabled": True,
+        "events_captured": 5,
+        "memories_stored": 3,
+        "predictions": {"total": 2, "avg_accuracy": 0.85},
+        "event_types": {
+            "task_posted": 2,
+            "bid_submitted": 1,
+            "task_assigned": 1,
+            "task_completed": 1,
+            "task_failed": 0,
+            "agent_joined": 0,
+            "tool_executed": 0,
+            "creative_step": 0,
+        },
+    }
+
+    event = MagicMock()
+    event.event_type = "task_completed"
+    event.description = "Task finished"
+    event.importance = 0.8
+    event.created_at = "2026-01-01T00:00:00"
+    event.agent_id = "agent-1234-abcd"
+    event.task_id = "task-5678-efgh"
+    event.data = '{"result": "ok"}'
+    engine.get_timeline.return_value = [event]
+
+    pred = MagicMock()
+    pred.task_id = "task-5678-efgh"
+    pred.accuracy = 0.9
+    pred.evaluated_at = "2026-01-01T01:00:00"
+    pred.created_at = "2026-01-01T00:30:00"
+    pred.predicted_value = '{"outcome": "success"}'
+    engine.get_predictions.return_value = [pred]
+
+    advisory = MagicMock()
+    advisory.category = "performance"
+    advisory.priority = "high"
+    advisory.title = "Slow tasks"
+    advisory.detail = "Tasks taking longer than expected"
+    advisory.suggested_action = "Scale up workers"
+    engine.get_advisories.return_value = [advisory]
+
+    return engine
+
+
+class TestTowerUI:
+    """Tests for GET /tower endpoint."""
+
+    @patch("dashboard.routes.tower.spark_engine", new_callable=_mock_spark_engine)
+    def test_tower_returns_200(self, mock_engine, client):
+        response = client.get("/tower")
+        assert response.status_code == 200
+
+    @patch("dashboard.routes.tower.spark_engine", new_callable=_mock_spark_engine)
+    def test_tower_returns_html(self, mock_engine, client):
+        response = client.get("/tower")
+        assert "text/html" in response.headers["content-type"]
+
+    @patch("dashboard.routes.tower.spark_engine", new_callable=_mock_spark_engine)
+    def test_tower_contains_dashboard_content(self, mock_engine, client):
+        response = client.get("/tower")
+        body = response.text
+        assert "tower" in body.lower() or "spark" in body.lower()
+
+
+class TestSparkSnapshot:
+    """Tests for _spark_snapshot helper."""
+
+    @patch("dashboard.routes.tower.spark_engine", new_callable=_mock_spark_engine)
+    def test_snapshot_structure(self, mock_engine):
+        from dashboard.routes.tower import _spark_snapshot
+
+        snap = _spark_snapshot()
+        assert snap["type"] == "spark_state"
+        assert "status" in snap
+        assert "events" in snap
+        assert "predictions" in snap
+        assert "advisories" in snap
+
+    @patch("dashboard.routes.tower.spark_engine", new_callable=_mock_spark_engine)
+    def test_snapshot_events_parsed(self, mock_engine):
+        from dashboard.routes.tower import _spark_snapshot
+
+        snap = _spark_snapshot()
+        ev = snap["events"][0]
+        assert ev["event_type"] == "task_completed"
+        assert ev["importance"] == 0.8
+        assert ev["agent_id"] == "agent-12"
+        assert ev["task_id"] == "task-567"
+        assert ev["data"] == {"result": "ok"}
+
+    @patch("dashboard.routes.tower.spark_engine", new_callable=_mock_spark_engine)
+    def test_snapshot_predictions_parsed(self, mock_engine):
+        from dashboard.routes.tower import _spark_snapshot
+
+        snap = _spark_snapshot()
+        pred = snap["predictions"][0]
+        assert pred["task_id"] == "task-567"
+        assert pred["accuracy"] == 0.9
+        assert pred["evaluated"] is True
+        assert pred["predicted"] == {"outcome": "success"}
+
+    @patch("dashboard.routes.tower.spark_engine", new_callable=_mock_spark_engine)
+    def test_snapshot_advisories_parsed(self, mock_engine):
+        from dashboard.routes.tower import _spark_snapshot
+
+        snap = _spark_snapshot()
+        adv = snap["advisories"][0]
+        assert adv["category"] == "performance"
+        assert adv["priority"] == "high"
+        assert adv["title"] == "Slow tasks"
+        assert adv["suggested_action"] == "Scale up workers"
+
+    @patch("dashboard.routes.tower.spark_engine")
+    def test_snapshot_handles_empty_state(self, mock_engine):
+        mock_engine.status.return_value = {"enabled": False}
+        mock_engine.get_timeline.return_value = []
+        mock_engine.get_predictions.return_value = []
+        mock_engine.get_advisories.return_value = []
+
+        from dashboard.routes.tower import _spark_snapshot
+
+        snap = _spark_snapshot()
+        assert snap["events"] == []
+        assert snap["predictions"] == []
+        assert snap["advisories"] == []
+
+    @patch("dashboard.routes.tower.spark_engine")
+    def test_snapshot_handles_invalid_json_data(self, mock_engine):
+        mock_engine.status.return_value = {"enabled": True}
+
+        event = MagicMock()
+        event.event_type = "test"
+        event.description = "bad data"
+        event.importance = 0.5
+        event.created_at = "2026-01-01T00:00:00"
+        event.agent_id = None
+        event.task_id = None
+        event.data = "not-json{"
+        mock_engine.get_timeline.return_value = [event]
+
+        pred = MagicMock()
+        pred.task_id = None
+        pred.accuracy = None
+        pred.evaluated_at = None
+        pred.created_at = "2026-01-01T00:00:00"
+        pred.predicted_value = None
+        mock_engine.get_predictions.return_value = [pred]
+
+        mock_engine.get_advisories.return_value = []
+
+        from dashboard.routes.tower import _spark_snapshot
+
+        snap = _spark_snapshot()
+        ev = snap["events"][0]
+        assert ev["data"] == {}
+        assert "agent_id" not in ev
+        assert "task_id" not in ev
+
+        pred = snap["predictions"][0]
+        assert pred["task_id"] == "?"
+        assert pred["predicted"] == {}
+
+
+class TestTowerWebSocket:
+    """Tests for WS /tower/ws endpoint."""
+
+    @patch("dashboard.routes.tower.spark_engine", new_callable=_mock_spark_engine)
+    @patch("dashboard.routes.tower._PUSH_INTERVAL", 0)
+    def test_ws_sends_initial_snapshot(self, mock_engine, client):
+        import json
+
+        with client.websocket_connect("/tower/ws") as ws:
+            data = json.loads(ws.receive_text())
+            assert data["type"] == "spark_state"
+            assert "status" in data
+            assert "events" in data

tests/infrastructure/test_error_capture.py

@@ -5,11 +5,13 @@ from datetime import UTC, datetime, timedelta
 from unittest.mock import patch
 
 from infrastructure.error_capture import (
+    _build_report_description,
     _create_bug_report,
     _dedup_cache,
     _extract_traceback_info,
     _get_git_context,
     _is_duplicate,
+    _log_bug_report_created,
     _log_error_event,
     _notify_bug_report,
     _record_to_session,
@@ -231,6 +233,68 @@ class TestLogErrorEvent:
             _log_error_event(e, "test", "abc123", "file.py", 42, {"branch": "main"})
 
 
+class TestBuildReportDescription:
+    """Test _build_report_description helper."""
+
+    def test_includes_error_info(self):
+        try:
+            raise RuntimeError("desc test")
+        except RuntimeError as e:
+            desc = _build_report_description(
+                e,
+                "test_src",
+                None,
+                "hash1",
+                "tb...",
+                "file.py",
+                10,
+                {"branch": "main"},
+            )
+        assert "RuntimeError" in desc
+        assert "test_src" in desc
+        assert "file.py:10" in desc
+        assert "hash1" in desc
+
+    def test_includes_context_when_provided(self):
+        try:
+            raise RuntimeError("ctx desc")
+        except RuntimeError as e:
+            desc = _build_report_description(
+                e,
+                "src",
+                {"path": "/api"},
+                "h",
+                "tb",
+                "f.py",
+                1,
+                {},
+            )
+        assert "path=/api" in desc
+
+    def test_omits_context_when_none(self):
+        try:
+            raise RuntimeError("no ctx")
+        except RuntimeError as e:
+            desc = _build_report_description(
+                e,
+                "src",
+                None,
+                "h",
+                "tb",
+                "f.py",
+                1,
+                {},
+            )
+        assert "**Context:**" not in desc
+
+
+class TestLogBugReportCreated:
+    """Test _log_bug_report_created helper."""
+
+    def test_does_not_crash_on_missing_deps(self):
+        _log_bug_report_created("test", "task-1", "hash1", "title")
+
+
 class TestCreateBugReport:
     """Test _create_bug_report helper."""
 

tests/infrastructure/test_router_history.py

@@ -0,0 +1,149 @@
+"""Tests for provider health history store and API endpoint."""
+
+import time
+from datetime import UTC, datetime, timedelta
+from unittest.mock import MagicMock
+
+import pytest
+from src.infrastructure.router.history import HealthHistoryStore
+
+
+@pytest.fixture
+def store():
+    """In-memory history store for testing."""
+    s = HealthHistoryStore(db_path=":memory:")
+    yield s
+    s.close()
+
+
+@pytest.fixture
+def sample_providers():
+    return [
+        {
+            "name": "anthropic",
+            "status": "healthy",
+            "error_rate": 0.01,
+            "avg_latency_ms": 250.5,
+            "circuit_state": "closed",
+            "total_requests": 100,
+        },
+        {
+            "name": "local",
+            "status": "degraded",
+            "error_rate": 0.15,
+            "avg_latency_ms": 80.0,
+            "circuit_state": "closed",
+            "total_requests": 50,
+        },
+    ]
+
+
+def test_record_and_retrieve(store, sample_providers):
+    store.record_snapshot(sample_providers)
+    history = store.get_history(hours=1)
+    assert len(history) == 1
+    assert len(history[0]["providers"]) == 2
+    assert history[0]["providers"][0]["name"] == "anthropic"
+    assert history[0]["providers"][1]["name"] == "local"
+    assert "timestamp" in history[0]
+
+
+def test_multiple_snapshots(store, sample_providers):
+    store.record_snapshot(sample_providers)
+    time.sleep(0.01)
+    store.record_snapshot(sample_providers)
+    history = store.get_history(hours=1)
+    assert len(history) == 2
+
+
+def test_hours_filtering(store, sample_providers):
+    old_ts = (datetime.now(UTC) - timedelta(hours=48)).isoformat()
+    store._conn.execute(
+        """INSERT INTO snapshots
+           (timestamp, provider_name, status, error_rate,
+            avg_latency_ms, circuit_state, total_requests)
+           VALUES (?, ?, ?, ?, ?, ?, ?)""",
+        (old_ts, "anthropic", "healthy", 0.0, 100.0, "closed", 10),
+    )
+    store._conn.commit()
+    store.record_snapshot(sample_providers)
+
+    history = store.get_history(hours=24)
+    assert len(history) == 1
+
+    history = store.get_history(hours=72)
+    assert len(history) == 2
+
+
+def test_prune(store, sample_providers):
+    old_ts = (datetime.now(UTC) - timedelta(hours=200)).isoformat()
+    store._conn.execute(
+        """INSERT INTO snapshots
+           (timestamp, provider_name, status, error_rate,
+            avg_latency_ms, circuit_state, total_requests)
+           VALUES (?, ?, ?, ?, ?, ?, ?)""",
+        (old_ts, "anthropic", "healthy", 0.0, 100.0, "closed", 10),
+    )
+    store._conn.commit()
+    store.record_snapshot(sample_providers)
+
+    deleted = store.prune(keep_hours=168)
+    assert deleted == 1
+    history = store.get_history(hours=999)
+    assert len(history) == 1
+
+
+def test_empty_history(store):
+    assert store.get_history(hours=24) == []
+
+
+def test_capture_snapshot_from_router(store):
+    mock_metrics = MagicMock()
+    mock_metrics.error_rate = 0.05
+    mock_metrics.avg_latency_ms = 200.0
+    mock_metrics.total_requests = 42
+
+    mock_provider = MagicMock()
+    mock_provider.name = "test-provider"
+    mock_provider.status.value = "healthy"
+    mock_provider.metrics = mock_metrics
+    mock_provider.circuit_state.value = "closed"
+
+    mock_router = MagicMock()
+    mock_router.providers = [mock_provider]
+
+    store._capture_snapshot(mock_router)
+    history = store.get_history(hours=1)
+    assert len(history) == 1
+    p = history[0]["providers"][0]
+    assert p["name"] == "test-provider"
+    assert p["status"] == "healthy"
+    assert p["error_rate"] == 0.05
+    assert p["total_requests"] == 42
+
+
+def test_history_api_endpoint(store, sample_providers):
+    """GET /api/v1/router/history returns snapshot data."""
+    store.record_snapshot(sample_providers)
+
+    from fastapi import FastAPI
+    from fastapi.testclient import TestClient
+    from src.infrastructure.router.api import get_cascade_router
+    from src.infrastructure.router.api import router as api_router
+    from src.infrastructure.router.history import get_history_store
+
+    app = FastAPI()
+    app.include_router(api_router)
+
+    app.dependency_overrides[get_history_store] = lambda: store
+    app.dependency_overrides[get_cascade_router] = lambda: MagicMock()
+
+    client = TestClient(app)
+    resp = client.get("/api/v1/router/history?hours=1")
+    assert resp.status_code == 200
+    data = resp.json()
+    assert len(data) == 1
+    assert len(data[0]["providers"]) == 2
+    assert data[0]["providers"][0]["name"] == "anthropic"
+
+    app.dependency_overrides.clear()

tests/timmy/test_cli.py

@@ -56,7 +56,7 @@ def test_think_sends_topic_to_agent():
 
 
 def test_think_ignores_model_size_option():
-    """think --model-size is accepted but not forwarded to create_timmy."""
+    """think --model-size 70b must not forward model_size to create_timmy."""
     mock_timmy = MagicMock()
 
     with patch("timmy.cli.create_timmy", return_value=mock_timmy) as mock_create:

tests/timmy/test_mcp_tools.py

@@ -6,6 +6,9 @@ import pytest
 
 from timmy.mcp_tools import (
     _bridge_to_work_order,
+    _build_issue_args,
+    _build_issue_body,
+    _category_from_labels,
     _generate_avatar_image,
     _parse_command,
     close_mcp_sessions,
@@ -132,6 +135,49 @@ def test_filesystem_mcp_returns_tools():
     assert "/home/user/project" in params_kwargs["args"]
 
 
+# ---------------------------------------------------------------------------
+# _build_issue_body / _build_issue_args / _category_from_labels
+# ---------------------------------------------------------------------------
+
+
+def test_build_issue_body_appends_signature():
+    """_build_issue_body appends the auto-filing signature."""
+    result = _build_issue_body("Some description")
+    assert result.startswith("Some description\n\n")
+    assert "Auto-filed by Timmy" in result
+
+
+def test_build_issue_body_empty():
+    """_build_issue_body handles empty body."""
+    result = _build_issue_body("")
+    assert result.startswith("---\n")
+
+
+def test_build_issue_args():
+    """_build_issue_args returns correct MCP arguments."""
+    with patch("timmy.mcp_tools.settings") as mock_settings:
+        mock_settings.gitea_repo = "owner/repo"
+        result = _build_issue_args("Title", "Body")
+    assert result == {
+        "method": "create",
+        "owner": "owner",
+        "repo": "repo",
+        "title": "Title",
+        "body": "Body",
+    }
+
+
+def test_category_from_labels_bug():
+    """_category_from_labels returns 'bug' when labels contain bug."""
+    assert _category_from_labels("bug, enhancement") == "bug"
+
+
+def test_category_from_labels_default():
+    """_category_from_labels returns 'suggestion' by default."""
+    assert _category_from_labels("enhancement") == "suggestion"
+    assert _category_from_labels("") == "suggestion"
+
+
 # ---------------------------------------------------------------------------
 # create_gitea_issue_via_mcp
 # ---------------------------------------------------------------------------

tests/unit/test_lightning.py

@@ -0,0 +1,109 @@
+"""Unit tests for the lightning package (factory + ledger)."""
+
+from __future__ import annotations
+
+import pytest
+
+from lightning.factory import Invoice, MockBackend, get_backend
+from lightning.ledger import (
+    TxStatus,
+    TxType,
+    clear,
+    create_invoice_entry,
+    get_balance,
+    get_transactions,
+    mark_settled,
+)
+
+
+@pytest.fixture(autouse=True)
+def _clean_ledger():
+    """Reset the in-memory ledger between tests."""
+    clear()
+    yield
+    clear()
+
+
+# ── Factory tests ────────────────────────────────────────────────────
+
+
+class TestMockBackend:
+    def test_create_invoice_returns_invoice(self):
+        backend = MockBackend()
+        inv = backend.create_invoice(100, "test memo")
+        assert isinstance(inv, Invoice)
+        assert inv.amount_sats == 100
+        assert inv.memo == "test memo"
+        assert len(inv.payment_hash) == 64  # SHA-256 hex
+        assert inv.payment_request.startswith("lnbc")
+
+    def test_invoices_have_unique_hashes(self):
+        backend = MockBackend()
+        a = backend.create_invoice(10)
+        b = backend.create_invoice(10)
+        assert a.payment_hash != b.payment_hash
+
+
+class TestGetBackend:
+    def test_returns_mock_backend(self):
+        backend = get_backend()
+        assert isinstance(backend, MockBackend)
+
+
+# ── Ledger tests ─────────────────────────────────────────────────────
+
+
+class TestLedger:
+    def test_create_invoice_entry(self):
+        entry = create_invoice_entry(
+            payment_hash="abc123",
+            amount_sats=500,
+            memo="test",
+            source="unit_test",
+        )
+        assert entry.tx_type == TxType.incoming
+        assert entry.status == TxStatus.pending
+        assert entry.amount_sats == 500
+
+    def test_mark_settled(self):
+        create_invoice_entry(payment_hash="hash1", amount_sats=100)
+        result = mark_settled("hash1", preimage="secret")
+        assert result is not None
+        assert result.status == TxStatus.settled
+        assert result.preimage == "secret"
+        assert result.settled_at != ""
+
+    def test_mark_settled_unknown_hash(self):
+        assert mark_settled("nonexistent") is None
+
+    def test_get_balance_empty(self):
+        bal = get_balance()
+        assert bal["net_sats"] == 0
+        assert bal["available_sats"] == 0
+
+    def test_get_balance_with_settled(self):
+        create_invoice_entry(payment_hash="h1", amount_sats=1000)
+        mark_settled("h1")
+        bal = get_balance()
+        assert bal["incoming_total_sats"] == 1000
+        assert bal["net_sats"] == 1000
+        assert bal["available_sats"] == 1000
+
+    def test_get_balance_pending_not_counted(self):
+        create_invoice_entry(payment_hash="h2", amount_sats=500)
+        bal = get_balance()
+        assert bal["incoming_total_sats"] == 0
+        assert bal["pending_incoming_sats"] == 500
+
+    def test_get_transactions_returns_entries(self):
+        create_invoice_entry(payment_hash="t1", amount_sats=10)
+        create_invoice_entry(payment_hash="t2", amount_sats=20)
+        txs = get_transactions()
+        assert len(txs) == 2
+
+    def test_get_transactions_filter_by_status(self):
+        create_invoice_entry(payment_hash="f1", amount_sats=10)
+        create_invoice_entry(payment_hash="f2", amount_sats=20)
+        mark_settled("f1")
+        assert len(get_transactions(status="settled")) == 1
+        assert len(get_transactions(status="pending")) == 1