Plan with an LLM
Built personalised meal-plan generation using Python, LangChain, and Gemini APIs, with LLM-driven planning and multi-step reasoning.
CASE STUDY / AGENTIC AI · MEAL PLANNING AND SHOPPING
AUSDOR: meal planning connected to shopping
A meal planning and shopping agent that turns personal preferences into meal plans and structured shopping lists through multi-step reasoning and tools.
01 / THE PROBLEM
What needed to work.
A meal plan is most useful when it can be translated into the next practical step. AUSDOR connects personalised meal planning to structured shopping lists, using an agentic workflow to carry context between the two.
02 / MY APPROACH
How I approached it.
Connect planning and tools
Implemented tool-integrated workflows that autonomously convert a generated meal plan into a structured shopping list.
Build the supporting stack
Used PostgreSQL, Docker, and AWS as part of the application’s backend and infrastructure toolchain.
03 / SYSTEM OVERVIEW
Connecting the pieces.
- 01Personal preferences
- 02LLM meal planning
- 03Multi-step reasoning & tools
- 04Structured shopping list
The focus was continuity between planning and action: using the meal plan as the basis for a structured shopping list, with multi-step reasoning and tools connecting the stages.
AUSDOR / AGENT WORKFLOW04
01Preferences
02Meal plan
03Shopping list
Personalised planning, connected to everyday life.
04 / OUTCOMES
What the system delivers.
Meal plans
LLM-driven planning tailored to personal preferences.
Shopping lists
An actionable list produced from the meal-planning workflow.
Independent project overview. The described workflow creates shopping lists; automated checkout and grocery purchases are not claimed.