Ai baby workbench (#13)

* refactor: update configuration from Azure OpenAI to OpenAI and enhance agent service initialization

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

* refactor: update AgentChat component for OpenAI integration and enhance markdown support

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

---------

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

Author: abossard

.env.example

@@ -1,8 +1,10 @@
-# Azure OpenAI Configuration
-# Get your API key from: https://portal.azure.com
+# OpenAI Configuration
+# Get your API key from: https://platform.openai.com/api-keys
 
-# Your Azure OpenAI API key (Bearer token)
-AZURE_API_KEY=your-api-key-here
+OPENAI_API_KEY=your-openai-api-key-here
+OPENAI_MODEL=gpt-4o-mini
+# Optional override
+# OPENAI_BASE_URL=https://api.openai.com/v1
 
 # Optional: Frontend build path override
 # FRONTEND_DIST=/path/to/custom/frontend/dist

backend/agents.py

@@ -1,10 +1,10 @@
 """
-Agent Management Module with Azure OpenAI and LangGraph
+Agent Management Module with OpenAI and LangGraph
 
 Provides LangGraph-based agents for task automation:
 - Type-safe data models with Pydantic
 - Self-documenting schemas for REST and MCP
-- Azure OpenAI integration via langchain-openai
+- OpenAI integration via langchain-openai
 - ReAct agent pattern with automatic tool discovery
 
 Following "Grokking Simplicity" and "A Philosophy of Software Design":
@@ -39,18 +39,28 @@
 
 load_dotenv()
 
+from uuid import UUID
+
 # Ensure operations register before we request LangChain tools
 import operations  # noqa: F401
+
 # Local - Import operations registry for automatic tool discovery
 from api_decorators import get_langchain_tools
+
+# Local CSV service
+from csv_data import get_csv_ticket_service
+
 # Third-party - FastMCP client for external MCP servers
 from fastmcp import Client as MCPClient
 from langchain_core.tools import StructuredTool
+
 # Third-party - LangChain and LangGraph
-from langchain_openai import AzureChatOpenAI
+from langchain_openai import ChatOpenAI
 from langgraph.prebuilt import create_react_agent
+
 # Third-party - Pydantic for validation
 from pydantic import BaseModel, Field, create_model, field_validator
+from tickets import TicketStatus
 
 # ============================================================================
 # DATA MODELS - Pydantic for validation and schema generation
@@ -133,14 +143,12 @@ class AgentResponse(BaseModel):
 
 
 # ============================================================================
-# CONFIGURATION - Azure OpenAI settings (hardcoded except API key)
+# CONFIGURATION - OpenAI settings
 # ============================================================================
 
-# Azure OpenAI configuration - only API key from environment
-AZURE_OPENAI_ENDPOINT = "https://can-i-haz-houze-resource.cognitiveservices.azure.com"
-AZURE_OPENAI_API_KEY = os.getenv("AZURE_API_KEY", "")
-AZURE_OPENAI_DEPLOYMENT = os.getenv("AZURE_OPENAI_DEPLOYMENT", "gpt-5-mini")
-AZURE_OPENAI_API_VERSION = os.getenv("AZURE_OPENAI_API_VERSION", "2025-04-01-preview")
+OPENAI_API_KEY = os.getenv("OPENAI_API_KEY", "")
+OPENAI_MODEL = os.getenv("OPENAI_MODEL", "gpt-4o-mini")
+OPENAI_BASE_URL = os.getenv("OPENAI_BASE_URL", "")  # optional override
 
 # External MCP server URL for ticket management (hardcoded)
 TICKET_MCP_SERVER_URL = "https://yodrrscbpxqnslgugwow.supabase.co/functions/v1/mcp/a7f2b8c4-d3e9-4f1a-b5c6-e8d9f0123456"
@@ -241,7 +249,7 @@ class AgentService:
     Agent service handling all LLM agent operations.
     
     Consolidated operations:
-    - Agent initialization with Azure OpenAI
+    - Agent initialization with OpenAI
     - Tool discovery from @operation registry
     - External MCP server tool integration (persistent connection)
     - ReAct agent execution with task tools
@@ -253,71 +261,39 @@ class AgentService:
 
     def __init__(self):
         """
-        Initialize the agent service with Azure OpenAI.
+        Initialize the agent service with OpenAI.
         
         Validates that required environment variables are set and creates
         the LLM client for agent execution.
         
         Raises:
-            ValueError: If Azure OpenAI configuration is incomplete
+            ValueError: If OpenAI configuration is incomplete
         """
         # Validate configuration
-        if not AZURE_OPENAI_API_KEY:
+        if not OPENAI_API_KEY:
             raise ValueError(
-                "Azure OpenAI API key not set. "
-                "Please set AZURE_API_KEY environment variable."
+                "OpenAI API key not set. "
+                "Please set OPENAI_API_KEY environment variable."
             )
 
-        # Initialize AzureChatOpenAI with hardcoded endpoint
-        self.llm = AzureChatOpenAI(
-            azure_endpoint=AZURE_OPENAI_ENDPOINT,
-            api_key=AZURE_OPENAI_API_KEY,  # type: ignore
-            azure_deployment=AZURE_OPENAI_DEPLOYMENT,
-            api_version=AZURE_OPENAI_API_VERSION,
+        # Initialize ChatOpenAI
+        self.llm = ChatOpenAI(
+            model=OPENAI_MODEL,
+            api_key=OPENAI_API_KEY,
+            base_url=OPENAI_BASE_URL or None,
+            temperature=0.0,
         )
 
-        # Load LangGraph tools from operation registry
-        # These are the actual Python functions decorated with @tool
-        self.tools = get_langchain_tools()
-        
-        # Ticket MCP client state (lazy initialization)
+        # CSV tools only (do not expose operations or external MCP)
+        self.tools = self._build_csv_tools()
+
+        # Ticket MCP client state (unused)
         self._ticket_mcp_client: Optional[MCPClient] = None
         self._ticket_mcp_tools_loaded = False
 
     async def _ensure_ticket_mcp_connection(self):
-        """
-        Ensure ticket MCP client is connected and tools are loaded.
-        
-        Opens a persistent connection to the external ticket MCP server and
-        converts its tools to LangChain format. Called lazily on first
-        agent run.
-        """
-        if self._ticket_mcp_tools_loaded:
-            return
-        
-        try:
-            # Create and connect ticket MCP client (keep connection open)
-            client = MCPClient(TICKET_MCP_SERVER_URL)
-            await client.__aenter__()
-            self._ticket_mcp_client = client
-            
-            # Fetch and convert ticket MCP tools
-            mcp_tools = await client.list_tools()
-            print(f"\n{'='*60}")
-            print(f"🎫 TICKET MCP SERVER CONNECTED")
-            print(f"{'='*60}")
-            print(f"   URL: {TICKET_MCP_SERVER_URL}")
-            print(f"   Tools available: {len(mcp_tools)}")
-            for tool in mcp_tools:
-                lc_tool = _mcp_tool_to_langchain(client, tool)
-                self.tools.append(lc_tool)
-                print(f"   ✓ {tool.name}: {(tool.description or '')[:60]}...")
-            print(f"{'='*60}\n")
-            self._ticket_mcp_tools_loaded = True
-            
-        except Exception as e:
-            print(f"WARNING: Failed to load ticket MCP tools from {TICKET_MCP_SERVER_URL}: {e}")
-            # Continue without ticket MCP tools - local tools still work
+        """No-op: external MCP tools not exposed."""
+        return
 
     async def close(self):
         """Close the ticket MCP client connection."""
@@ -327,7 +303,78 @@ async def close(self):
             except Exception:
                 pass
             self._ticket_mcp_client = None
-    
+
+    def _build_csv_tools(self) -> list[StructuredTool]:
+        """Build LangChain tools backed by CSVTicketService."""
+        import json
+        service = get_csv_ticket_service()
+
+        def _csv_list_tickets(status: str | None = None, assigned_group: str | None = None, has_assignee: bool | None = None) -> str:
+            try:
+                status_enum = TicketStatus(status.lower()) if status else None
+            except Exception:
+                status_enum = None
+            tickets = service.list_tickets(status=status_enum, assigned_group=assigned_group, has_assignee=has_assignee)
+            return json.dumps([t.model_dump() for t in tickets[:200]], default=str)
+
+        def _csv_get_ticket(ticket_id: str) -> str:
+            try:
+                tid = UUID(ticket_id)
+            except Exception:
+                return json.dumps({"error": "invalid ticket id"})
+            ticket = service.get_ticket(tid)
+            if not ticket:
+                return json.dumps({"error": "not found"})
+            return json.dumps(ticket.model_dump(), default=str)
+
+        def _csv_search_tickets(query: str, limit: int = 50) -> str:
+            q = query.lower()
+            tickets = service.list_tickets()
+            matched = []
+            for t in tickets:
+                text = " ".join([
+                    t.summary or "",
+                    t.description or "",
+                    t.notes or "",
+                    t.resolution or "",
+                    t.requester_name or "",
+                    t.assigned_group or "",
+                    t.city or "",
+                ]).lower()
+                if q in text:
+                    matched.append(t.model_dump())
+                    if len(matched) >= limit:
+                        break
+            return json.dumps(matched, default=str)
+
+        def _csv_ticket_fields() -> str:
+            # Use Ticket model fields as schema
+            from tickets import Ticket
+            return json.dumps(list(Ticket.model_fields.keys()))
+
+        return [
+            StructuredTool.from_function(
+                func=_csv_list_tickets,
+                name="csv_list_tickets",
+                description="List tickets from CSV with optional filters: status (new, assigned, in_progress, pending, resolved, closed, cancelled), assigned_group, has_assignee (true/false). Returns JSON array.",
+            ),
+            StructuredTool.from_function(
+                func=_csv_get_ticket,
+                name="csv_get_ticket",
+                description="Get a ticket by UUID (id). Returns JSON object including notes/resolution.",
+            ),
+            StructuredTool.from_function(
+                func=_csv_search_tickets,
+                name="csv_search_tickets",
+                description="Search tickets by text across summary, description, notes, resolution, requester, group, city. Returns JSON array.",
+            ),
+            StructuredTool.from_function(
+                func=_csv_ticket_fields,
+                name="csv_ticket_fields",
+                description="List available ticket fields (schema) as JSON array of field names.",
+            ),
+        ]
+
     async def run_agent(self, request: AgentRequest) -> AgentResponse:
         """
         Run a ReAct agent with the given request using LangGraph.
@@ -352,35 +399,40 @@ async def run_agent(self, request: AgentRequest) -> AgentResponse:
         Raises:
             ValueError: If agent execution fails
         """
-        # Ensure ticket MCP tools are loaded (lazy initialization)
-        await self._ensure_ticket_mcp_connection()
-        
         try:
             # Create ReAct agent with LangGraph tools
             # The tools are the actual Python functions with @tool decorator
             agent = create_react_agent(self.llm, self.tools)
 
             # System message to guide the agent's behavior
-            system_msg = (
-                "You are a support ticket management assistant. "
-                "Your primary role is to help users manage, search, and evaluate support tickets. "
-                "You MUST use the available tools to perform all actions - NEVER simulate or pretend. "
-                "\n\n"
-                "TICKET CAPABILITIES:\n"
-                "- Search and list tickets by status, priority, city, service\n"
-                "- Get detailed ticket information and work logs\n"
-                "- Analyze ticket statistics and trends\n"
-                "- Request modifications to tickets (status, priority, assignee, etc.)\n"
-                "- Review and approve/reject modification requests\n"
-                "\n"
-                "TASK CAPABILITIES:\n"
-                "- Create, update, delete, and list tasks\n"
-                "- Get task statistics\n"
-                "\n"
-                "When users ask about tickets, use the ticket tools (list_tickets, get_ticket, search_tickets, etc.). "
-                "When users ask about tasks, use task tools (create_task, list_tasks, etc.). "
-                "Always confirm actions based on actual tool results."
-            )
+            tool_lines = []
+            for t in self.tools:
+                name = t.name if hasattr(t, 'name') else str(t)
+                desc = (t.description if hasattr(t, 'description') else "") or ""
+                tool_lines.append(f"- `{name}`: {desc}".strip())
+            tools_md = "\n".join(tool_lines) if tool_lines else "- (none)"
+
+            system_msg = f"""
+Du bist ein freundlicher CSV-Ticket-Assistent. Sprich **Deutsch**.
+
+Antwortstil:
+- Starte immer mit einer kurzen Begrüßung.
+- Liste sofort die verfügbaren Tools (Markdown-Bullets).
+- Nutze **Markdown** mit klaren Überschriften (##), Bullet-Listen und Tabellen, wenn sinnvoll.
+- Für JSON-Daten nutze fenced Code-Blöcke:
+  ```json
+  {{"example": "value"}}
+  ```
+- Halte Antworten knapp und gut strukturiert.
+
+Verfügbare Tools:
+{tools_md}
+
+Verhalten:
+- Verwende ausschließlich die csv_* Tools für Ticket-Informationen. Keine Daten erfinden.
+- Falls Daten fehlen, sage das explizit.
+- Fasse Ergebnisse klar zusammen; für Listen sind kompakte Tabellen ideal.
+"""
 
             # Execute agent with user prompt
             print(f"\n{'='*60}")

backend/app.py

@@ -29,7 +29,7 @@
 
 # Import unified operation system
 
-# Agent service for Azure OpenAI LangGraph agents
+# Agent service for OpenAI LangGraph agents
 from agents import AgentRequest, AgentResponse, agent_service
 from api_decorators import operation
 
@@ -157,12 +157,12 @@ async def rest_get_stats():
 
 
 # ============================================================================
-# AGENT ENDPOINT - Azure OpenAI LangGraph Agent
+# AGENT ENDPOINT - OpenAI LangGraph Agent
 # ============================================================================
 
 @app.route("/api/agents/run", methods=["POST"])
 async def rest_run_agent():
-    """REST wrapper: run AI agent with Azure OpenAI.
+    """REST wrapper: run AI agent with OpenAI.
     
     The agent has access to task tools and ticket MCP tools.
     """

backend/test_agents.py

@@ -3,7 +3,7 @@
 Test script for LangGraph agent integration.
 
 This script tests:
-1. Agent service initialization (without Azure OpenAI)
+1. Agent service initialization (without OpenAI)
 2. Operation to LangChain tool conversion
 3. MCP tool schema generation
 
@@ -66,7 +66,7 @@ def test_agent_import():
         print()
     except ImportError as e:
         print(f"  ⚠ Could not import agents module: {e}")
-        print("  This is expected if Azure OpenAI dependencies are not configured")
+        print("  This is expected if OpenAI dependencies are not configured")
         print()
 
 def test_task_service():
@@ -103,7 +103,7 @@ def main():
         print()
         print("Next steps:")
         print("1. Copy .env.example to .env")
-        print("2. Configure Azure OpenAI credentials in .env")
+        print("2. Configure OPENAI_API_KEY in .env")
         print("3. Start the server: python app.py")
         print("4. Test agent: POST /api/agents/run with {\"prompt\": \"List all tasks\"}")
         print()

frontend/package.json

@@ -23,7 +23,9 @@
     "@nivo/stream": "^0.99.0",
     "react": "^18.2.0",
     "react-dom": "^18.2.0",
-    "react-router-dom": "^7.9.6"
+    "react-markdown": "^10.1.0",
+    "react-router-dom": "^7.9.6",
+    "remark-gfm": "^4.0.1"
   },
   "devDependencies": {
     "@playwright/test": "^1.42.1",