Appendix - Python & Systems

What We Learned

• Use typed Python, validated boundaries and locked environments for reproducible services.
• Use async concurrency for model calls; bound retries, queues, rate limits and deadlines.
• Match APIs to the interaction: short response, streaming or background job.
• Version code, data, models and configuration; use CI and eval gates to catch regressions.
• Operate remote GPU services with process supervision, logs and driver/runtime checks.

Key Acronyms, Concepts & Jargon

Term	Short meaning
API / SDK	Application Programming Interface / Software Development Kit: service contract / client tooling.
I/O / GIL	Input/Output / Global Interpreter Lock: external waiting / a constraint on Python thread execution.
asyncio	Python's cooperative concurrency system for overlapping I/O waits.
JSON / SSE	JavaScript Object Notation / Server-Sent Events: structured data / one-way HTTP streaming.
Idempotency	Repeated requests have the same intended effect as one request.
Backpressure / jitter	Slowing intake when overloaded / randomizing retry delays.
CI / PR	Continuous Integration / Pull Request: automated checks / a reviewable change proposal.
LFS / DVC	Large File Storage / Data Version Control: version large artifacts / datasets.
SSH	Secure Shell: encrypted remote terminal access.
SIGTERM / SIGKILL	Graceful termination request / immediate forced process termination.
GPU / CUDA	Graphics Processing Unit / Compute Unified Device Architecture: accelerator / NVIDIA's GPU computing platform.
Lockfile	Records resolved dependencies to reproduce an environment.

Back to section overview