Autopilot live streaming

from pathlib import Path
import sys

# Resolve the repo root whether this cell is running from ./notebooks
# or from the repo root — mirrors the existing 01–36 notebook series.
ROOT = (
    Path.cwd().resolve().parent
    if Path.cwd().name == "notebooks"
    else Path.cwd().resolve()
)
if str(ROOT) not in sys.path:
    sys.path.insert(0, str(ROOT))

from examples.run_multi_tool_agent import build_llm_from_env

# Bedrock Llama 4 Scout is the default model this library ships with
# (see `shipit_agent/config.py`). No model name required — the helper
# reads `AWS_REGION` / credentials from env and returns an adapter that
# works with every Autopilot / Agent class.
llm = build_llm_from_env("bedrock")
print("Bedrock LLM ready:", type(llm).__name__)

from shipit_agent.autopilot import Autopilot, BudgetPolicy
from shipit_agent.deep import Goal

goal = Goal(
    objective="List the top 5 Python HTTP client libraries and their tradeoffs.",
    success_criteria=["At least 5 libraries listed", "Each has a one-line tradeoff"],
)
autopilot = Autopilot(
    llm=llm,
    goal=goal,
    budget=BudgetPolicy(max_seconds=300, max_iterations=6, max_tool_calls=15),
    heartbeat_every_seconds=20,
)

for ev in autopilot.stream(run_id="http-libs-v1"):
    kind = ev["kind"]
    if kind == "autopilot.run_started":
        print(f"🚀 run_started — goal: {ev['goal']['objective']}")
    elif kind == "autopilot.iteration":
        met = ev["criteria_met"]
        score = f"{sum(met)}/{len(met)}"
        print(f"✓ iter {ev['iteration']} — {score} criteria met")
    elif kind == "autopilot.heartbeat":
        print(f"♥ heartbeat iter={ev['iteration']}")
    elif kind == "autopilot.budget_exceeded":
        print(f"⛔ budget: {ev['reason']}")
    elif kind == "autopilot.result":
        print(f"🏁 {ev['status']} — {ev['iterations']} iters")

from shipit_agent.autopilot import Autopilot, BudgetPolicy
from shipit_agent.deep import Goal
from shipit_agent.live_renderer import render_stream

autopilot2 = Autopilot(
    llm=llm,
    goal=Goal(
        objective="Write a haiku about Bedrock Llama",
        success_criteria=["Three lines", "5-7-5 syllables"],
    ),
    budget=BudgetPolicy(max_iterations=4, max_seconds=120),
)
# `fmt="plain"` is notebook-friendly (no ANSI escapes).
final = render_stream(autopilot2.stream(run_id="haiku-v1"), fmt="plain")
print("\nfinal status:", final and final.get("status"))

from pathlib import Path
import io
from shipit_agent.live_renderer import render_stream

autopilot3 = Autopilot(
    llm=llm,
    goal=Goal(objective="2+2?", success_criteria=["Answer is 4"]),
    budget=BudgetPolicy(max_iterations=2, max_seconds=60),
)
buffer = io.StringIO()
render_stream(autopilot3.stream(run_id="math-v1"), fmt="jsonl", out=buffer)

for line in buffer.getvalue().strip().split("\n")[:6]:
    print(line[:120])

heartbeats = []


def record(ev):
    if ev.get("kind") == "autopilot.heartbeat":
        heartbeats.append(ev)


autopilot4 = Autopilot(
    llm=llm,
    goal=Goal(
        objective="Who invented Python?", success_criteria=["Mentions Guido van Rossum"]
    ),
    budget=BudgetPolicy(max_iterations=3, max_seconds=90),
    heartbeat_every_seconds=5,  # aggressive for demo
    on_heartbeat=record,
)
for _ in autopilot4.stream(run_id="guido-v1"):
    pass

print(f"Captured {len(heartbeats)} heartbeat events during the run.")