Code.

Capstone Project Code Snippets

Gen AI Capabilities include:

1. Agent

2. Function Calling

3. Grounding

4. Structured output / JSON

5. System Instruction tuning

Create local SQLite database and add synthetic data

Populate local SQLite database with a curated list of my favorite restaurants in downtown Palo Alto and details with my dining preferences

Function Calling: defining ‘list_tables’ to list all tables in database

Function Calling: defining ‘describe_table’ to look up table schema

Function Calling: defining ‘execute_query’ to execute an SQL statement, then return result

System Instruction: set rules and guidelines for model behavior

Model used: gemini-1.5-pro

Start Chat with Eat.Where.Now

Continue Chat with Eat.Where.Now

Inspect Chat History on calls that the model makes

The responses that client returns in Chat History

First check if model can access real-time weather information

Confirmed, It can’t!

Add Google Search Grounding capability to check for real-time weather condition “if sunny”

Initialize Gen AI Model

Model = gemini-2.0-flash

Function Calling: defining ‘get_lunchtime_weather’ to query model if it is sunny, set temperature=0.0 for deterministic output

Design Prompt: to query model to check if weather “is sunny” for specific date and time period (including late lunch schedule)

Structured Output in JSON: Check output in JSON structure if any reference to “is sunny” from real-time weather check

Function Calling: define function "find-restaurants-with-outdoor-seating” if weather is sunny

Main logic to determine if Eat.Where.Now. should recommend restaurants with out door seating if date and time period is sunny

Output: Yes, it is sunny!

Eat.Where.Now. recommends three restaurants that have outdoor seating!