feat: add app logging

[ppatel9703]

Description

Producer side of app-owner log visibility: tags log lines with the project they belong to, writes them to a per-project file, and exposes a disk-based search over that file (and its rotated history) for a searchable "Logs" tab. Live tail delivery is not in this repo — that's Monolith#398, which streams this same file over the existing insight websocket. The frontend consuming both is semoss-ui#3302.

Changes Made

src/prerna/sablecc2/comm/PixelJobRunner.java
Tags the Log4j2 MDC with projectId for the duration of a pixel job, via insight.getContextProjectId() — the project this insight is currently scoped to (set by SetContext(...)). Originally read insight.getProjectId(), a different field entirely ("the project this insight is saved in"), which stays null for any transient REPL-style insight — Terminal's, Console's — that's only ever scoped via SetContext. That bug meant MDC tagging silently never fired for pixels run from Terminal, while looking like it worked everywhere else by coincidence (some other insight-loading paths populate the base field too).

log4j2.xml
Added [project=%X{projectId}] to the global Console/RollingRandomAccessFile patterns, purely so a human grepping semossLog.log can see which project a line belongs to. Unrelated to the per-project routing below.

src/prerna/logging/AppLogManager.java (already present, one fix in this PR)
Per-project Log4j2 RollingFileAppender factory — writes {projectVersionFolder}/logs/app.log (50MB × 10 rotations), filtered to only that project's MDC-tagged events. This PR's fix: now also attaches to the root logger (LogManager.ROOT_LOGGER_NAME), not just prerna/EngineLogger. Custom app reactors almost never live under prerna.* — a typical app's reactors are in their own package (e.g. reactors.vaapi.*), a sibling namespace to prerna, not a descendant, so they fall through to root instead of inheriting prerna's appenders. Without this, a custom reactor's own logger.error(...) calls never reached the per-project file no matter how much logging an app author added.
Known remaining gap: root's own level threshold (WARN in log4j2.xml) still drops INFO/DEBUG from custom packages before any appender sees them — fixing that means lowering Root's level platform-wide, a bigger, separate call.

src/prerna/logging/GetAppLogsReactor.java (deleted)
Was a polling reactor (offset/nextOffset client-side polling). Superseded entirely — live tail is now a websocket push from Monolith (AppLogStreamer), and historical reads are the new search reactor below. No reactor call is needed for either anymore.

src/prerna/logging/SearchAppLogsReactor.java (new)
Reads app.log plus its rotated siblings (app.log.1....10) directly from disk — no database. Filters by text substring and/or level, paginated (offset/limit, capped at 500), newest-first. Owner-gated, same rule the old reactor used (SecurityProjectUtils.getUserProjectPermission + AccessPermissionEnum.isOwner).
Deliberately not a DB-backed appender (an AuditLogsJDBCAppender-style table was the original plan) — app.log's own rotation already bounds storage per project, so a DB would've solved a retention problem that doesn't exist here, at the cost of a new schema/write path with no retention story of its own (the audit-log precedent doesn't have a purge job either).

How to Test

1. Open any app you own, run some pixels/reactors against it.
2. Confirm {projectVersionFolder}/logs/app.log exists and only contains this project's lines.
3. Add a logger.error(...)/logger.warn(...) call to a custom-package reactor (not under prerna.*) in that app, trigger it, confirm the line shows up in app.log (this is the root-logger fix — would not have worked before this PR).
4. Call SearchAppLogs(paramValues=[{"projectId": "<id>", "query": "...", "levels": "ERROR,WARN", "offset": "0", "limit": "50"}]); — confirm results are newest-first, correctly filtered, and paginate via hasMore/totalMatches.
5. Confirm a non-owner calling SearchAppLogs gets rejected.
6. Run a pixel from a Terminal/Console-style insight (one only ever SetContext'd, never setProjectId'd directly) — confirm its logs now carry [project=...] in the global log too (the PixelJobRunner fix).

Notes

• How the three repos fit together: this repo tags + writes + searches the per-project file. Monolith (AppLogStreamer) tails that same file and pushes it live over /insightSocket. semoss-ui has two consumers: a live-tail Console panel (talks to Monolith's socket directly) and a searchable Logs page (talks to SearchAppLogsReactor here via runQuery).
• Pod-locality: app.log is local to whichever pod's disk wrote it — no cross-pod visibility in a multi-replica deployment.
• Cluster sync risk: the file lives inside the project version folder, which rclone sync pushes to S3 with no exclude list (only fires from explicit actions — build/publish, upload, GitHub sync — never automatically). Two pods each syncing their own copy will clobber each other's history on whichever push lands last. Not fixed in this PR.
• Full architecture write-up (MDC background, full producer/delivery/consumer chain, known limitations): docs/app-logs/architecture.md in the workspace root (not part of this repo).

Review Updates

Addressed the following from review:

• Stripped em-dashes/typographic characters from comments and log strings in AppLogManager.java, SearchAppLogsReactor.java, PixelJobRunner.java, Insight.java — this repo's Maven compiler plugin enforces cp1252 source encoding, and non-ASCII punctuation risks build/encoding issues.
• SearchAppLogsReactor: added a user == null || user.getPrimaryLoginToken() == null guard before dereferencing the token, matching the existing convention elsewhere in the codebase (e.g. AgentRuntimeManager).
• Insight.java: fixed an unused FileSystems import (was imported but referenced fully-qualified) and corrected import ordering.

[snyk-io]

✅ Snyk checks have passed. No issues have been found so far.

Status	Scan Engine	Critical	High	Medium	Low	Total (0)
✅	Open Source Security	0	0	0	0	0 issues

💻 Catch issues earlier using the plugins for VS Code, JetBrains IDEs, Visual Studio, and Eclipse.