Add real-time event streaming emitter for Browser.AI

[Copilot]

Implements a loosely-coupled, async-first event streaming system for real-time logging, agent execution tracking, and UI integration.

Core API (browser_ai/event_bus/emitter.py)

• emit() / emit_async() - fire events from anywhere
• subscribe(topic, handler) / unsubscribe() - topic-based pub/sub with wildcard * support
• EventLevel enum for filtering (DEBUG, INFO, WARNING, ERROR, CRITICAL)
• StreamingEventHandler - async iterator for WebSocket/SSE streaming
• FilteredEventHandler - filter by level or custom predicate

Logging Integration (browser_ai/event_bus/handlers/logging_handler.py)

• LoggingEventHandler bridges Python logging → event bus
• Prevents recursion by filtering event_bus module logs

Agent Service Integration

Events emitted at key lifecycle points:

• Agent started/completed/failed
• Step started/completed/failed
• LLM request started/completed, rate limits
• Planning started/completed
• User help requested

Usage

from browser_ai.event_bus import emit, emit_async, subscribe, StreamingEventHandler
from browser_ai.event_bus import AgentStartedEvent, AgentStepCompletedEvent

# Subscribe to events
subscribe("agent", lambda e: print(f"{e.topic}/{e.name}"))
subscribe("*", my_handler)  # wildcard

# Emit events
emit(AgentStartedEvent(task="Search for tutorials"))
await emit_async(AgentStepCompletedEvent(step_number=1, actions_taken=[...]))

# Stream for WebSocket/SSE
handler = StreamingEventHandler()
subscribe("*", handler)
async for event in handler.stream():
    await websocket.send(event.model_dump_json())

All events are Pydantic models with full type safety and JSON serialization support.

Original prompt

for the browsai agent library i want to have a real time event steaming emitter kind of logging . not only for logging its also for to streem the curen status of task llm outputs angent exection for ui and relevent subscriber in different level in should be a versatile and extendabel loosely coupedn one that can be called anyware async and pydentic type safe

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.