Optimize agent runtime and SLA demo flow (#15)

* feat: Enhance SLA Breach Risk functionality and UI integration
- Increased max_length for agent prompt to 5000
- Added fields parameter to list and search tickets for selective data retrieval
- Updated timeout for usecase demo agent to 300 seconds
- Introduced SLA Breach Risk demo with detailed prompt and ticket analysis
- Added E2E tests for SLA Breach Risk demo page

* feat: add incident_id field to ticket model and related components

- Added incident_id to the ticket mapping in app.py.
- Updated csv_data.py to include incident_id when converting CSV rows to tickets.
- Modified operations.py to define incident_id as a CSV ticket field.
- Enhanced the Ticket model in tickets.py to include incident_id.
- Updated usecase_demo.py to accommodate changes in ticket structure.
- Modified CSVTicketTable.jsx to display incident_id in the ticket table.
- Updated TicketList.jsx to filter and display incident_id in the ticket list.
- Enhanced TicketsWithoutAnAssignee.jsx to include incident_id in ticket operations.
- Updated UsecaseDemoPage.jsx to pass matchingTickets to the render function.
- Enhanced demoDefinitions.js to improve prompts for use case demos.
- Added SLA Breach Overview result view in resultViews.jsx to visualize SLA status of tickets.

Signed-off-by: Andre Bossard <anbossar@microsoft.com>

* refactor: clean up import statements across multiple components

Signed-off-by: Andre Bossard <anbossar@microsoft.com>

* refactor: standardize import statement formatting in resultViews.jsx

Signed-off-by: Andre Bossard <anbossar@microsoft.com>

* feat: add SLA breach reporting functionality and related API endpoints

Signed-off-by: Andre Bossard <anbossar@microsoft.com>

* feat: implement SLA breach report retrieval for unassigned tickets

Signed-off-by: Andre Bossard <anbossar@microsoft.com>

---------

Signed-off-by: Andre Bossard <anbossar@microsoft.com>

Author: abossard

backend/agents.py

@@ -36,7 +36,6 @@
 
 # CSV ticket service
 from csv_data import Ticket, get_csv_ticket_service
-from usecase_demo import UsecaseDemoRunCreate, usecase_demo_run_service
 
 # FastMCP client for direct ticket MCP calls (no AI)
 from fastmcp import Client as MCPClient
@@ -51,6 +50,7 @@
     op_update_task,
     task_service,
 )
+from usecase_demo import UsecaseDemoRunCreate, usecase_demo_run_service
 
 # Ticket MCP server URL (same as in agents.py)
 TICKET_MCP_SERVER_URL = "https://yodrrscbpxqnslgugwow.supabase.co/functions/v1/mcp/a7f2b8c4-d3e9-4f1a-b5c6-e8d9f0123456"
@@ -375,6 +375,7 @@ def _map_mcp_ticket_to_frontend(mcp_ticket: dict) -> dict:
 
     return {
         "id": str(mcp_ticket.get("id", "")),
+        "incident_id": mcp_ticket.get("incident_id"),
         "title": mcp_ticket.get("summary", ""),
         "description": mcp_ticket.get("description", ""),
         "status": status,
@@ -622,6 +623,28 @@ async def get_csv_ticket_stats():
     })
 
 
+@app.route("/api/csv-tickets/sla-breach", methods=["GET"])
+async def get_csv_tickets_sla_breach():
+    """
+    Return unassigned tickets grouped by SLA breach status (breached → at_risk),
+    sorted by age_hours descending within each group.
+
+    Query params:
+    - unassigned_only: true/false (default: true)
+    - include_ok: true/false (default: false) — include non-breached tickets too
+    """
+    from tickets import get_sla_breach_report
+
+    unassigned_only = request.args.get("unassigned_only", "true").lower() != "false"
+    include_ok = request.args.get("include_ok", "false").lower() == "true"
+
+    tickets = _csv_ticket_service.list_tickets(
+        has_assignee=False if unassigned_only else None,
+    )
+    report = get_sla_breach_report(tickets, reference_time=None, include_ok=include_ok)
+    return jsonify(report.model_dump(mode="json"))
+
+
 @app.route("/api/health", methods=["GET"])
 async def health_check():
     """Health check endpoint."""

backend/app.py

@@ -290,6 +290,7 @@ def csv_row_to_ticket(row: CSVTicketRow) -> Ticket:
 
     return Ticket(
         id=ticket_id,
+        incident_id=row.incident_id or row.entry_id or None,
         summary=row.summary or "No summary",
         description=row.notes or row.summary or "No description",
         status=map_status(row.status or row.status_ppl),

backend/csv_data.py

@@ -12,7 +12,13 @@
 from api_decorators import operation
 from csv_data import get_csv_ticket_service
 from tasks import Task, TaskCreate, TaskFilter, TaskService, TaskStats, TaskUpdate
-from tickets import Ticket, TicketStatus
+from tickets import (
+    SlaBreachReport,
+    Ticket,
+    TicketSlaInfo,
+    TicketStatus,
+    get_sla_breach_report,
+)
 
 # Service instances shared across interfaces
 _task_service = TaskService()
@@ -21,6 +27,7 @@
 
 
 CSV_TICKET_FIELDS = [
+    {"name": "incident_id", "label": "Incident ID", "type": "string"},
     {"name": "id", "label": "ID", "type": "uuid"},
     {"name": "summary", "label": "Summary", "type": "string"},
     {"name": "status", "label": "Status", "type": "enum"},
@@ -271,6 +278,44 @@ async def op_csv_ticket_fields() -> list[dict[str, str]]:
     return CSV_TICKET_FIELDS
 
 
+@operation(
+    name="csv_sla_breach_tickets",
+    description=(
+        "Return tickets at SLA breach risk from the CSV dataset. "
+        "By default only unassigned tickets (assigned to a group but no individual) are included. "
+        "Results contain pre-computed age_hours, sla_threshold_hours, and breach_status. "
+        "Grouped: 'breached' first, then 'at_risk'. Within each group sorted by age_hours descending. "
+        "The reference timestamp is the maximum created_at date found in the selected tickets "
+        "(not the current system time), making results deterministic for historical datasets. "
+        "SLA thresholds: critical=4h, high=24h, medium=72h, low=120h."
+    ),
+    http_method="GET",
+)
+async def op_csv_sla_breach_tickets(
+    unassigned_only: bool = True,
+    include_ok: bool = False,
+) -> SlaBreachReport:
+    """
+    Pre-compute SLA breach status for CSV tickets.
+
+    Args:
+        unassigned_only: When True (default), only return tickets that are assigned
+            to a group but have no individual assignee — the primary use case for
+            proactive SLA monitoring.
+        include_ok: When True, also include tickets that are within their SLA window.
+            Default False keeps the result focused on actionable items.
+
+    Returns:
+        SlaBreachReport with reference_timestamp, counts, and a sorted list of
+        TicketSlaInfo objects ready for display or further AI commentary.
+    """
+    _ensure_csv_loaded()
+    tickets = _csv_service.list_tickets(
+        has_assignee=False if unassigned_only else None,
+    )
+    return get_sla_breach_report(tickets, reference_time=None, include_ok=include_ok)
+
+
 # Export shared services for callers (REST app, CLI tools, etc.)
 task_service = _task_service
 csv_ticket_service = _csv_service
@@ -289,5 +334,6 @@ async def op_csv_ticket_fields() -> list[dict[str, str]]:
     "op_csv_search_tickets",
     "op_csv_ticket_stats",
     "op_csv_ticket_fields",
+    "op_csv_sla_breach_tickets",
     "CSV_TICKET_FIELDS",
 ]

backend/operations.py

@@ -17,7 +17,6 @@
 
 from pydantic import BaseModel, Field
 
-
 # ============================================================================
 # ENUMS - Status and Priority types
 # ============================================================================
@@ -58,7 +57,7 @@ class WorkLogType(str, Enum):
 
 
 # ============================================================================
-# PRIORITY SLA DEADLINES (in minutes)
+# PRIORITY SLA DEADLINES (in minutes) — kept for backwards compatibility
 # ============================================================================
 
 PRIORITY_SLA_MINUTES: dict[TicketPriority, int] = {
@@ -69,6 +68,27 @@ class WorkLogType(str, Enum):
 }
 
 
+# ============================================================================
+# SLA BREACH THRESHOLDS (in hours) — used for breach status calculations
+# Aligns with ITSM standard expectations and the frontend dashboard
+# ============================================================================
+
+SLA_THRESHOLD_HOURS: dict[TicketPriority, float] = {
+    TicketPriority.CRITICAL: 4.0,
+    TicketPriority.HIGH: 24.0,
+    TicketPriority.MEDIUM: 72.0,
+    TicketPriority.LOW: 120.0,
+}
+
+
+class SlaBreachStatus(str, Enum):
+    """SLA breach status for a ticket."""
+    BREACHED = "breached"   # age > threshold
+    AT_RISK = "at_risk"     # age > 75% of threshold
+    OK = "ok"               # age <= 75% of threshold
+    UNKNOWN = "unknown"     # cannot determine
+
+
 # ============================================================================
 # WORKLOG MODEL
 # ============================================================================
@@ -159,6 +179,7 @@ class Ticket(BaseModel):
     """
     # Core identifiers
     id: UUID = Field(..., description="Unique ticket identifier")
+    incident_id: Optional[str] = Field(None, description="Original incident ID (INC...)")
 
     # Summary and description
     summary: str = Field(..., max_length=500, description="Short issue summary")
@@ -377,6 +398,141 @@ def build_reminder_candidate(
     )
 
 
+# ============================================================================
+# SLA BREACH MODELS — for breach-status reporting
+# ============================================================================
+
+class TicketSlaInfo(BaseModel):
+    """SLA breach information for a single ticket."""
+    ticket_id: str = Field(..., description="Incident ID or UUID of the ticket")
+    priority: str = Field(..., description="Ticket priority (critical/high/medium/low)")
+    urgency: Optional[str] = Field(None, description="Urgency level")
+    assigned_group: Optional[str] = Field(None, description="Responsible support team")
+    reported_date: str = Field(..., description="Ticket creation date (ISO format)")
+    age_hours: float = Field(..., description="Hours elapsed since creation (1 decimal)")
+    sla_threshold_hours: float = Field(..., description="SLA threshold in hours for this priority")
+    breach_status: SlaBreachStatus = Field(..., description="Current SLA breach status")
+
+
+class SlaBreachReport(BaseModel):
+    """Aggregated SLA breach report."""
+    reference_timestamp: str = Field(..., description="Reference time used for age calculation")
+    total_breached: int = Field(..., ge=0)
+    total_at_risk: int = Field(..., ge=0)
+    tickets: list[TicketSlaInfo] = Field(default_factory=list)
+
+
+# ============================================================================
+# SLA BREACH CALCULATIONS — Pure functions
+# ============================================================================
+
+def get_sla_threshold_hours(priority: TicketPriority) -> float:
+    """Return the SLA threshold in hours for a given priority."""
+    return SLA_THRESHOLD_HOURS.get(priority, SLA_THRESHOLD_HOURS[TicketPriority.LOW])
+
+
+def calculate_ticket_sla_info(
+    ticket: "Ticket",
+    reference_time: Optional[datetime] = None,
+) -> TicketSlaInfo:
+    """
+    Compute SLA breach information for a single ticket.
+
+    Args:
+        ticket: The ticket to evaluate.
+        reference_time: Anchor for age calculation. When None, uses the current time.
+            Callers managing historical datasets should pass the max date in the
+            dataset so results are deterministic.
+
+    Returns:
+        TicketSlaInfo with pre-computed age_hours and breach_status.
+    """
+    if reference_time is None:
+        reference_time = datetime.now(ticket.created_at.tzinfo)
+
+    delta = reference_time - ticket.created_at
+    age_hours = round(delta.total_seconds() / 3600, 1)
+
+    threshold = get_sla_threshold_hours(ticket.priority)
+
+    if age_hours > threshold:
+        status = SlaBreachStatus.BREACHED
+    elif age_hours > threshold * 0.75:
+        status = SlaBreachStatus.AT_RISK
+    else:
+        status = SlaBreachStatus.OK
+
+    ticket_id = ticket.incident_id or str(ticket.id)
+
+    return TicketSlaInfo(
+        ticket_id=ticket_id,
+        priority=ticket.priority.value,
+        urgency=ticket.urgency,
+        assigned_group=ticket.assigned_group,
+        reported_date=ticket.created_at.isoformat(),
+        age_hours=age_hours,
+        sla_threshold_hours=threshold,
+        breach_status=status,
+    )
+
+
+def get_sla_breach_report(
+    tickets: "list[Ticket]",
+    reference_time: Optional[datetime] = None,
+    include_ok: bool = False,
+) -> SlaBreachReport:
+    """
+    Build a sorted, grouped SLA breach report from a ticket list.
+
+    Grouping order: breached first, then at_risk, then ok (if include_ok).
+    Within each group, sorted by age_hours descending.
+
+    Args:
+        tickets: Tickets to evaluate.
+        reference_time: Anchor timestamp. Uses max created_at in the list when None.
+        include_ok: Whether to include non-breached, non-at-risk tickets.
+
+    Returns:
+        SlaBreachReport with grouped/sorted TicketSlaInfo entries.
+    """
+    if not tickets:
+        ref_str = (reference_time or datetime.now()).isoformat()
+        return SlaBreachReport(reference_timestamp=ref_str, total_breached=0, total_at_risk=0)
+
+    if reference_time is None:
+        reference_time = max(t.created_at for t in tickets)
+
+    infos = [calculate_ticket_sla_info(t, reference_time) for t in tickets]
+
+    group_order = {
+        SlaBreachStatus.BREACHED: 0,
+        SlaBreachStatus.AT_RISK: 1,
+        SlaBreachStatus.OK: 2,
+        SlaBreachStatus.UNKNOWN: 3,
+    }
+
+    filtered = [
+        i for i in infos
+        if i.breach_status in (SlaBreachStatus.BREACHED, SlaBreachStatus.AT_RISK)
+        or (include_ok and i.breach_status == SlaBreachStatus.OK)
+    ]
+
+    sorted_infos = sorted(
+        filtered,
+        key=lambda i: (group_order[i.breach_status], -i.age_hours),
+    )
+
+    total_breached = sum(1 for i in sorted_infos if i.breach_status == SlaBreachStatus.BREACHED)
+    total_at_risk = sum(1 for i in sorted_infos if i.breach_status == SlaBreachStatus.AT_RISK)
+
+    return SlaBreachReport(
+        reference_timestamp=reference_time.isoformat(),
+        total_breached=total_breached,
+        total_at_risk=total_at_risk,
+        tickets=sorted_infos,
+    )
+
+
 # ============================================================================
 # EXPORTS
 # ============================================================================
@@ -387,8 +543,10 @@ def build_reminder_candidate(
     "TicketPriority",
     "ModificationStatus",
     "WorkLogType",
+    "SlaBreachStatus",
     # Constants
     "PRIORITY_SLA_MINUTES",
+    "SLA_THRESHOLD_HOURS",
     # Models
     "Ticket",
     "TicketWithDetails",
@@ -403,6 +561,9 @@ def build_reminder_candidate(
     "ModificationCreate",
     "ModificationReview",
     "OverlayMetadata",
+    # SLA breach models
+    "TicketSlaInfo",
+    "SlaBreachReport",
     # Reminder models
     "ReminderCandidate",
     "ReminderRequest",
@@ -414,4 +575,8 @@ def build_reminder_candidate(
     "is_assigned_without_assignee",
     "count_reminders_in_worklogs",
     "build_reminder_candidate",
+    # SLA breach calculations
+    "get_sla_threshold_hours",
+    "calculate_ticket_sla_info",
+    "get_sla_breach_report",
 ]

backend/tickets.py

@@ -23,7 +23,7 @@
 from agents import AgentRequest, agent_service
 
 USECASE_DEMO_AGENT_TIMEOUT_SECONDS = float(
-    os.getenv("USECASE_DEMO_AGENT_TIMEOUT_SECONDS", "120")
+    os.getenv("USECASE_DEMO_AGENT_TIMEOUT_SECONDS", "300")
 )
 
 
@@ -191,14 +191,13 @@ async def _execute_run(self, run_id: str) -> None:
         # Enforce a predictable output block for table rendering.
         structured_prompt = (
             f"{run.prompt}\n\n"
-            "Zusatzformat:\n"
-            "- Gib zuerst eine kurze Zusammenfassung.\n"
-            "- Füge danach einen JSON-Codeblock mit `rows` hinzu.\n"
-            "- JSON-Schema:\n"
-            "  {\"rows\": [{\"menu_point\": \"...\", \"project_name\": \"...\", "
-            "\"summary\": \"...\", \"agent_prompt\": \"...\", \"ticket_ids\": \"...\", "
-            "\"csv_evidence\": \"...\"}]}\n"
-            "- Falls keine sinnvollen Zeilen existieren, gib `{\"rows\": []}` zurück."
+            "Antwortformat:\n"
+            "- Führe die Anfrage mit möglichst wenigen Tool-Aufrufen aus.\n"
+            "- Nutze kompakte fields und sinnvolle limits.\n"
+            "- Fordere notes/resolution nur bei explizitem Bedarf an.\n"
+            "- Gib einen JSON-Codeblock mit {\"rows\": [...]} zurück.\n"
+            "- Falls keine sinnvollen Zeilen existieren, gib {\"rows\": []} zurück.\n"
+            "- Optional danach: kurze Zusammenfassung in 2-4 Stichpunkten."
         )
 
         try: