CRYENX LABS / ENGINEERING NOTES
Under the hood.
Engineering notes from the systems we build and run: agents, retrieval, evals, guardrails and data pipelines. What we measured, what broke, and what it cost.
Cryenx Labs
desk_agent.py
python
r
= client.system_one(
state=case,
questions={
"etas_agree"
: Noul(...),
"safe_to_run"
: Noul(...),
},
)
# 362 ms, $0.00011
evals
555 calls
ETA check
90.0%
Invoice
100%
Baseline
56.7%
guardrail passed
agent run
freight desk
read carrier email
jev
classify
362 ms
propose re-tender $2,523
jev
safe_to_run
0.91
✓
executed