Allow agents to pass parameters to backend functions and receive responses

The problem today:

When a Base44 agent calls a backend function, the framework sends the HTTP request with an empty body (content-length: 0) - even though the agent correctly constructs the parameters (they're visible in the conversation view on the dashboard). The backend function receives no input and fails.

This means agents can only use functions that require zero parameters, which severely limits what they can do autonomously.

What we'd like to see:

1. Parameter forwarding - When the agent constructs parameters for a function call, actually include them in the HTTP request body so the function receives them

2. Response visibility - Let the agent see the function's return value so it can summarize results, make decisions, or chain multiple function calls based on previous output

Why this matters:

Today, agents are limited to entity CRUD and conversation. Any real backend logic (calling
external APIs, running LLM pipelines, processing data) requires workarounds like:

• Telling the user to go to a different page and do it themselves

• Building the entire workflow inside the agent's prompt (which can't call external services)

• Using entity writes as a "message queue" with a separate automation polling for new records

With parameter passing + response visibility, agents become truly autonomous assistants:

• Content generation: User sends a YouTube URL → agent calls getTranscript(url) → gets transcript → calls generateContent(transcriptId, type) → returns the finished article in chat

• Data lookup: Agent calls a function with filters → gets structured results → presents them conversationally

• Multi-step workflows: Agent chains function calls based on intermediate results, making decisions the way a human would