Context engineering
for LLMs

from mycontext.intelligence import smart_execute
import asyncio

# One call — auto-selects the right cognitive pattern,
# builds the context, and executes
response, meta = smart_execute(
    "Why did API response times triple after last deploy?",
    provider="openai",
)
print(meta["templates_used"])  # ['root_cause_analyzer']

# Or run multiple contexts concurrently — true async fan-out
async def parallel():
    from mycontext import Context, Guidance, Directive
    ctx1 = Context(guidance="Risk analyst", directive="Assess launch risk.")
    ctx2 = Context(guidance="Data analyst", directive="Review Q3 trends.")
    r1, r2 = await asyncio.gather(
        ctx1.aexecute(provider="openai"),
        ctx2.aexecute(provider="anthropic"),
    )
    return r1.response, r2.response

from mycontext.intelligence import QualityMetrics, ContextAmplificationIndex

# Score any context on 6 dimensions
metrics = QualityMetrics()
score = metrics.evaluate(ctx)
print(f"Quality: {score.overall:.2f}")  # 0.87

# Prove templates work with CAI
cai = ContextAmplificationIndex(provider="openai")
result = cai.measure(question, template_name="root_cause_analyzer")
print(f"CAI: {result.cai_overall:.2f}x")  # 1.42x — 42% better output