Feat/ai samples function calling

[sedanah-m]

No description provided.

[wiz-9635d3485b]

Wiz Scan Summary

Scanner	Findings
Vulnerabilities	-
Sensitive Data	-
Secrets	-
IaC Misconfigurations	-
SAST Findings	-
Software Management Findings	-
Total	-

View scan details in Wiz

To detect these findings earlier in the dev lifecycle, try the Wiz Code extension for VS Code, JetBrains, or Visual Studio.

[gemini-code-assist]

Code Review

This pull request introduces implementation examples for Firebase AI features, specifically adding function-calling capabilities and structured output generation (JSON and Enum schemas) with corresponding UI views. The review feedback highlights two critical issues in the function-calling service: first, the tools parameter in getAiModel must be passed as an array rather than a single object to avoid compilation or runtime errors; second, when the model returns multiple function calls, all responses must be collected and sent back together to prevent the Gemini API from throwing a runtime error.

[gemini-code-assist]

The tools configuration parameter expects an array of tools (e.g., [fetchWeatherTool]) rather than a single tool object. Passing a single object will cause a TypeScript compilation error or runtime failure. Additionally, the model property is redundant here since it is already specified as the first argument to getAiModel.

const model = getAiModel('gemini-3.6-flash', {
    // Provide the function declaration to the model.
    tools: [fetchWeatherTool]
});

[gemini-code-assist]

When the model returns multiple function calls in a single turn, the Gemini API requires that you provide a response for each of the requested function calls in the next message. If you only send a single response back (the last one that was overwritten in the loop), the API will throw a runtime error. Collecting all responses into an array and sending them together resolves this issue and ensures robust multi-turn function calling.

const functionCalls = result.response.functionCalls() ?? [];
const functionResponses = [];

// When the model responds with one or more function calls, invoke the function(s).
for (const call of functionCalls) {
    if (call.name === "fetchWeather") {
        // Forward the structured input data prepared by the model
        // to the hypothetical external API.
        const functionResult = await fetchWeather(call.args as { location: { city: string; state: string }; date: string });
        functionResponses.push({
            functionResponse: {
                name: call.name,
                response: functionResult,
            },
        });
    }
}

/**
 * Step 5: Send the response from the function back to the model
 * so that the model can use it to generate its final response.
 */
if (functionResponses.length > 0) {
    result = await chat.sendMessage(functionResponses);
}

[sedanah-m]

Branch got messy; creating a new one for a cleaner start.