feat: complete phase 1 -- core framework with chat loop, agents, and React UI

Backend:
- FastAPI WebSocket /ws endpoint with streaming via LangGraph astream
- LangGraph Supervisor connecting 3 mock agents (order_lookup, order_actions, fallback)
- YAML Agent Registry with Pydantic validation and immutable configs
- PostgresSaver checkpoint persistence via langgraph-checkpoint-postgres
- Session TTL with 30-min sliding window and interrupt extension
- LLM provider abstraction (Anthropic/OpenAI/Google)
- Token usage + cost tracking callback handler
- Input validation: message size cap, thread_id format, content length
- Security: no hardcoded defaults, startup API key validation, no input reflection

Frontend:
- React 19 + TypeScript + Vite chat UI
- WebSocket hook with reconnect + exponential backoff
- Streaming token display with agent attribution
- Interrupt approval/reject UI for write operations
- Collapsible tool call viewer

Testing:
- 87 unit tests, 87% coverage (exceeds 80% requirement)
- Ruff lint + format clean

Infrastructure:
- Docker Compose (PostgreSQL 16 + backend)
- pyproject.toml with full dependency management

backend/app/graph.py

@@ -0,0 +1,70 @@
+"""LangGraph Supervisor construction -- connects registry, agents, LLM, and persistence."""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+from langgraph.prebuilt import create_react_agent
+from langgraph_supervisor import create_supervisor
+
+from app.agents import get_tools_by_names
+
+if TYPE_CHECKING:
+    from langchain_core.language_models import BaseChatModel
+    from langgraph.checkpoint.postgres.aio import AsyncPostgresSaver
+    from langgraph.graph.state import CompiledStateGraph
+
+    from app.registry import AgentRegistry
+
+SUPERVISOR_PROMPT = (
+    "You are a customer support supervisor. "
+    "Route customer requests to the appropriate agent based on their description. "
+    "For order status and tracking queries, use the order_lookup agent. "
+    "For order modifications like cancellations, use the order_actions agent. "
+    "For anything else, use the fallback agent."
+)
+
+
+def build_agent_nodes(
+    registry: AgentRegistry,
+    llm: BaseChatModel,
+) -> list:
+    """Create LangGraph react agent nodes from registry configurations."""
+    agent_nodes = []
+    for agent_config in registry.list_agents():
+        tools = get_tools_by_names(agent_config.tools)
+
+        system_prompt = (
+            f"You are the {agent_config.name} agent. "
+            f"Personality: {agent_config.personality.tone}. "
+            f"{agent_config.description} "
+            f"Permission level: {agent_config.permission}."
+        )
+
+        agent_node = create_react_agent(
+            model=llm,
+            tools=tools,
+            name=agent_config.name,
+            prompt=system_prompt,
+        )
+        agent_nodes.append(agent_node)
+
+    return agent_nodes
+
+
+def build_graph(
+    registry: AgentRegistry,
+    llm: BaseChatModel,
+    checkpointer: AsyncPostgresSaver,
+) -> CompiledStateGraph:
+    """Build and compile the LangGraph supervisor graph."""
+    agent_nodes = build_agent_nodes(registry, llm)
+
+    workflow = create_supervisor(
+        agent_nodes,
+        model=llm,
+        prompt=SUPERVISOR_PROMPT,
+        output_mode="full_history",
+    )
+
+    return workflow.compile(checkpointer=checkpointer)