Agent usability: docs, imports, REVIEW log level

[chiruu12]

Summary

• Fix Guard.init() docstring — it described auto-instrumentation that does not exist; now documents the process-wide singleton.
• Standardize import guidance across primary docs: apps/agents use from unplug import ...; server/MCP dependents use unplug.api.*; unplug.core.* is internal.
• Downgrade routine REVIEW pipeline outcome logs from WARNING to INFO (scanner errors / ML degradation warnings unchanged).
• ML degradation warnings in guard.py were already unified in Harden model store and guard ML init #72 — both load-failure paths cite pip install "unplug-ai[ml]" plus unplug-models download; no code change needed.

Changed files

Item	Files
1. Guard.init() docstring	sdk/src/unplug/guard.py
2. Import-style docs	README.md, sdk/README.md, sdk/docs/GETTING_STARTED.md, sdk/docs/PUBLIC_API.md, sdk/docs/AGENT_ACTIONS.md
3. REVIEW log level	sdk/src/unplug/pipelines/base.py, sdk/tests/unit/pipelines/test_pipelines.py
4. Degradation warnings	Already done (#72) — verified only

Test plan

• make check (ruff, mypy, pytest)
• make test-cov — 85.59% total coverage (80% gate)
• New TestPipelineLogging.test_review_outcome_logs_at_info asserts REVIEW logs at INFO via caplog

---

Note

Low Risk
Documentation and logging-only changes; no scan policy, auth, or enforcement behavior changes beyond log level for REVIEW.

Overview
Agent usability polish: correct Guard.init() docs, standardize import guidance, and quiet routine REVIEW pipeline logs.

Guard.init() no longer claims auto-instrumentation; the docstring now describes creating the process-wide singleton used with Guard.get().

Primary docs (README, GETTING_STARTED, PUBLIC_API, AGENT_ACTIONS) now spell out import paths: apps/agents use top-level from unplug import ...; server/MCP dependents use unplug.api.*; unplug.core.* is internal.

In BasePipeline, normal REVIEW outcomes log at INFO instead of WARNING (mid-risk scores no longer elevate REVIEW to warning). Scanner errors and other degradation warnings are unchanged.

Adds TestPipelineLogging.test_review_outcome_logs_at_info and a CHANGELOG entry under Fixed.

^{Reviewed by Cursor Bugbot for commit 4a7fc6d. Bugbot is set up for automated code reviews on this repo. Configure here.}

[greptile-apps]

Greptile Summary

This PR updates agent-facing docs and quiets routine review logging.

• Clarifies Guard.init() as a process-wide singleton.
• Standardizes public import guidance across README and SDK docs.
• Logs routine REVIEW pipeline outcomes at INFO.
• Adds test coverage for the new review log level.

Confidence Score: 5/5

The changed code looks mergeable after small cleanup to diagnostics and docs.

• Review events can disappear from warning-based monitoring unless operators update log routing.
• The main README gives mixed import guidance in the Quickstart.
• No blocking runtime, build, or security issues were found in the changed paths.

sdk/src/unplug/pipelines/base.py, README.md

Important Files Changed

Filename	Overview
sdk/src/unplug/pipelines/base.py	Routes Action.REVIEW pipeline outcome logs to INFO while leaving error logs and non-review risk-band logging unchanged.
sdk/tests/unit/pipelines/test_pipelines.py	Adds a regression test that triggers a review tool-call outcome and checks its log level.
sdk/src/unplug/guard.py	Updates the Guard.init() docstring to describe singleton creation.
README.md	Adds import-style guidance, with one adjacent Quickstart example still using the server-style import path.
sdk/README.md	Adds SDK import-style guidance for top-level app imports and unplug.api.* dependent imports.
sdk/docs/PUBLIC_API.md	Adds an audience-based import-path table for top-level and unplug.api.* usage.

_{Reviews (1): Last reviewed commit: "Clarify imports and quiet REVIEW logs" | Re-trigger Greptile}

[github-actions]

SDK Coverage •

File	Stmts	Miss	Cover	Missing
src/unplug
guard.py	343	28	91%	127, 194–195, 197, 213–214, 288, 301, 304, 309, 417, 449, 451, 453, 490, 509, 512, 517, 552, 599, 684–685, 687–690, 696, 715
src/unplug/pipelines
base.py	88	2	97%	100, 166
TOTAL	7164	1017	85%

Tests	Skipped	Failures	Errors	Time
1148	25 💤	0 ❌	0 🔥	40.043s ⏱️