SDR Orchestration (LangGraph) Implementation Plan

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: Implement the core conversation graph using LangGraph to manage the SDR flow (Greeting -> Info Gathering -> Scheduling).

Architecture: Use a StateGraph with a shared SDRState. Micro-agents (nodes) will be implemented using Pydantic AI for deterministic tool calling. Persistence will use a MongoDB checkpointer.

Tech Stack: Python 3.12, LangGraph, Pydantic AI, MongoDB.

---

Task 1: Define Conversation State & Schema

• Step 1: Define SDRState in infrastructure/llm/agents/state.py Include fields for tenant_id, customer_id, chat_history, extracted_info (service, professional, time).

---

Task 2: Implement Base Graph Structure

• Step 1: Create the graph entry point in infrastructure/llm/agents/sdr_graph.py
• Step 2: Add basic nodes: greeting, collect_info, check_availability, finalize_booking
• Step 3: Define edges and conditional routing

---

Task 3: Node Implementation (Micro-Agents)

• Step 1: Implement collect_info node using Pydantic AI Define tools for listing services and professionals using the repositories created in the previous phase.

---

Task 4: Integration & Testing

• Step 1: Create a CLI test script to simulate a conversation
• Step 2: Verify state persistence in MongoDB between turns