Chris Perko

Staring blindly at a loading spinner for 15 seconds while an AI pipeline runs is a broken user experience. If you are building multi-step AI features, waiting for the entire backend loop to finish before sending a single byte to the client feels incredibly sluggish. You don't need complex WebSockets or heavy database polling to fix this. You can stream real-time execution telemetry milestones natively from a serverless backend directly into reactive Angular Signals. Here is how the architecture hangs together: The Backend Orchestration: Use Firebase Genkit’s onCallGenkit wrapper to expose a type-safe, streaming Callable Cloud Function. By leveraging the streamingCallback argument, you can explicitly push status keys down the wire (parsing_structure, analyzing_impact) before triggering your final structured Gemini 3.5 Flash model generation. The Client Interceptor: On the Angular side, Firebase’s standard httpsCallable handler returns a stream reference containing an asynchronous iterable. You can loop through these incoming chunks in true real-time using a clean for await...of block. Strict State Encapsulation: To keep data flow predictable, manage the internal states using private WritableSignals, but expose them to your components as true read-only states using .asReadonly(). This stops components from accidentally calling .set() or .update(). Declarative Templates: With your step states bound to a Signal, your template drops complex RxJS mapping or async pipe overhead entirely. Modern control flow blocks like @switch handle the visual progression beautifully. The major engineering win here is that passing these milestones natively in-memory via HTTP chunks completely bypasses the need to write temporary status strings to database rows, cutting cloud read/write storage overhead to zero. I wrote a complete, step-by-step implementation guide detailing the full backend and frontend code blocks. Link to the full post is in the reply below.

Did you miss our GDG Salt Lake City event in June? Don't worry, both talks are now on your YouTube channel! 👇