@apmantza/greedysearch-pi 1.8.7 → 1.8.9

package/CHANGELOG.md

@@ -2,8 +2,43 @@
 
 ## [Unreleased]
 
+## [1.8.9] — 2026-05-11
+
+### Changed
+
+- **Halved Gemini synthesis timeout** (`extractors/gemini.mjs`) — `waitForStreamComplete` timeout reduced from 90s to 45s. Gemini synthesis prompts are ~8-10k chars and typically respond in 15-30s. The extra 45s was pure dead time.
+- **Aligned Gemini extractor hard timeout** (`src/search/engines.mjs`) — reduced from 120s to 70s, matching the new 45s stream wait + ~25s nav/settle overhead.
+
+### Fixed
+
+- **Perplexity/Bing visible recovery now actually stores cookies** (`bin/search.mjs`) — Two issues fixed:
+  1. **Second visible retry**: The first visible retry resolves Cloudflare/Turnstile (navigating through the challenge which breaks the CDP session with "Inspected target navigated or closed"), but the search never ran. A second retry on the same tab now reuses the freshly-cached Turnstile cookies and executes the actual search.
+  2. **Keep Chrome alive on recovery success**: Previously Chrome was killed with `taskkill /F` after recovery, losing any pending cookie database writes. Now visible Chrome stays running when recovery succeeds (or needs human intervention), keeping the cookie session alive.
+- **Visible Chrome window minimized after recovery** (`bin/search.mjs`) — When visible Chrome is left open after recovery (for cookie persistence or user verification), the window is automatically minimized so it doesn't clutter the desktop.
+
+## [1.8.8] — 2026-05-09
+
+### Added
+
+- **`/set-greedy-locale` Pi command** (`index.ts`) — Set default locale for search results (e.g., `/set-greedy-locale de`, `/set-greedy-locale --clear`, `/set-greedy-locale --show`). Saves to `~/.config/greedysearch/config.json`.
+- **Browser lifecycle defense patterns** (`src/search/browser-lifecycle.mjs`, new) — Centralized lifecycle management adopted from open-websearch's robust cross-process browser patterns:
+  - **Structured JSON metadata** (`greedysearch-chrome-metadata.json`) replaces three scattered text files (PID, mode, activity) with a single file tracking `browserPid`, `debugPort`, `tempDir`, `clientPids[]`, `sessionMode`, `lastActivity`, `launchedAt`. Backward-compatible — legacy files still written.
+  - **Process command-line verification** — `verifyBrowserProcess()` checks not just PID alive but that the process command line contains the profile dir and debug port. Prevents PID collision false-positives where a different process reuses the same PID.
+  - **Cross-process launch lock** — `acquireLaunchLock()` uses exclusive-create (`wx` flag) to prevent concurrent `ensureChrome()` calls from racing to launch Chrome. Stale lock recovery after 15s.
+  - **Stale session cleanup** — `cleanupStaleSessions()` runs once per process on first `ensureChrome()`. Scans metadata for dead PIDs, verifies survivors via command line, force-kills orphans, reclaims ghost processes on port 9222.
+  - **Client PID tracking** — `registerClient`/`unregisterClient` track which processes share the Chrome instance.
+- **Mode-specific idle timeouts** (`src/search/chrome.mjs`) — Headless Chrome keeps the aggressive 5-minute idle timeout (`GREEDY_SEARCH_IDLE_TIMEOUT_MINUTES`) since it's cheap to restart. Visible Chrome (explicitly launched for captcha/cookie setup) gets a 60-minute grace period (`GREEDY_SEARCH_VISIBLE_IDLE_TIMEOUT_MINUTES`) to avoid wasting the user's captcha investment. Set either to 0 to disable for that mode.
+
+### Added
+
+- **System command path resolution** (`src/utils/system-cmds.mjs`, new) — `resolveSystemCmd()` resolves `powershell`, `netstat`, `taskkill`, `ps`, `lsof`, `ss`, `grep` to absolute paths for secure execution. `isPathSafe()` validates PATH environment variable composition. Satisfies SonarCloud security hotspot requirements for `execFileSync`/`execSync` PATH safety.
+
 ### Fixed
 
+- **SonarCloud security hotspots — 15 resolved** — Addressed all flagged items:
+  - **11 ReDoS-prone regex patterns**: Replaced greedy `.{0,50}` in fetcher's content quality check with lazy quantifier `.{0,50}?`; replaced alternation-heavy split regex in bing-copilot with `[^\S\n]*` horizontal whitespace; replaced `[\s\S]*` JSON extraction patterns in synthesis.mjs with `indexOf`/`lastIndexOf` brace matching; replaced `.+?\.` in selectors with `[^.]+`; replaced `\s+\S*$` trim patterns in sources.mjs, common.mjs, and content.mjs with `lastIndexOf` word-boundary detection; replaced markdown link regex in common.mjs with O(n) indexOf-based parser.
+  - **4 PATH-injection hotspots in browser-lifecycle.mjs and chrome.mjs**: Created `resolveSystemCmd()` utility returning absolute paths for `powershell.exe`, `netstat.exe`, `taskkill.exe` (Windows) and `/usr/bin/ps`, `/usr/bin/lsof`, `/usr/sbin/ss`, `/usr/bin/grep` (Unix). Replaced all bare command names in `execFileSync`/`execSync` calls.
+
 - **SonarCloud minor vulnerability false positives** — Confirmed both remaining issues are false positives (internal diagnostic logging in `bin/gschrome.mjs` and test debug output in `test/fetcher-cli.mjs`). Verified via full smoke test suite: all 33 unit tests pass, all 4 engines (Perplexity, Bing, Google, Gemini) return results at all depths (fast/standard/deep), CDP safety wrappers correctly enforce mode boundaries.
 
 - **SonarCloud security hotspots** (re-verified) — All previously fixed hotspots remain resolved: replaced `spawn("node", ...)` with `spawn(process.execPath, ...)`, replaced `Math.random()` with `crypto.randomInt()`, 19 remaining hotspots confirmed as false positives (hardcoded `execSync` commands, simple regex patterns).

package/README.md

@@ -71,9 +71,10 @@ greedy_search({ query: "Bing captcha setup", engine: "bing", visible: true });
 Inside Pi, prefer the extension commands (no package path needed):
 
 ```text
-/greedy-visible  # launch visible Chrome for captcha/login/cookie setup
-/greedy-status   # show GreedySearch Chrome status
-/greedy-kill     # stop GreedySearch Chrome
+/greedy-visible      # launch visible Chrome for captcha/login/cookie setup
+/greedy-status       # show GreedySearch Chrome status
+/greedy-kill         # stop GreedySearch Chrome
+/set-greedy-locale   # set default result language (de, fr, es, ja, etc.)
 ```
 
 Git install path:
@@ -109,7 +110,7 @@ Chrome is auto-cleaned after 5 min idle. Override with `GREEDY_SEARCH_IDLE_TIMEO
 ## Requirements
 
 - Chrome
-- Node.js 22+
+- Node.js 20.11.0+
 
 ## Known engine quirks