pllmClient boundary
Understand which data and runtime state must stay with the client.
pllm.runtime exposes clients, transport code, privacy declarations, gateways,
and application factories. The Python API inventory
lists each public object. Some interfaces predate complete plan-based deployment;
an importable object is not necessarily supported for production use.
Plaintext requests terminate inside trusted client boundary. Client owns tokenization, state, sampling, and decoded output under methods that claim this placement. Gateway compatibility must be tested against exact application/SDK semantics; API resemblance does not establish full OpenAI compatibility.
The local gateway binds to loopback and is client-controlled. Provider roles do not receive plaintext prompts, tool schemas, arguments, or results. See status for runtime limits.
Python SDK example
import pllm
mode = pllm.PrivacyMode.PUBLIC
assert mode.value == "public"
assert mode.protocol == "masked_w4a4"API: pllm.PrivacyMode
This reads a public privacy declaration; it does not create a protected client session.