Personal CookFlow AI Agent V1.0(Nov 2025)

Summary

Objective

Create an AI agent system that reduces meal‑planning stress by guiding users through a structured, one‑day batch‑cooking ritual tailored to their household, preferences, and schedule.

Solution:

CookFlow uses a coordinated team of specialized agents:

• Root Agent: orchestrates the full planning and cooking flow.
• User Preferences Agent: captures household size, dietary needs, and pantry staples.
• Recipe Finder Agent: sources batch‑friendly recipes that meet constraints.
• Grocery Planner Agent: generates categorized, quantity‑aware shopping lists.
• Batch Cooking Agent: sequences tasks for efficient, low‑stress cooking.
• Meal Distribution Agent: maps meals into a weekly calendar with portioning guidance.

All agents share a unified session context to maintain consistency across the workflow.

Technical Implementation:

• Built in Python: modular agent architecture, clean JSON contracts, and extensible workflows make it easy to adapt.
• Deployed on GCP Cloud Run: serverless scaling ensures agents spin up only when needed, keeping costs lean and performance sharp.
• Orchestrated agents: Root, Preferences, Finder, Planner, Batch Cooking, and Distribution all run as independent services, coordinated seamlessly.
• Shared session context: every agent reads from the same source of truth, so preferences, pantry, and provenance stay consistent end‑to‑end.

Evaluation:

• 15 real user conversations + 4 survey responses.
• Users highlighted clear grocery lists, organized cooking workflows, and recipe variety.
• Identified improvements: faster response times, better recipe retrieval, stronger memory of servings, and mobile formatting fixes.