Changelog

Product updates and improvements

• Switch desktop updates to Electron's standard whole-app updater: download a verified platform package from the single release feed, then restart to install the frontend, backend, and desktop shell together; 0.3.0 begins the new updater baseline
• Add Custom Script HTML Reports: enable reviewed custom JavaScript for a Standard report, keep managed Chart.js separate, edit scripts through controlled Agent tools, and preview on an isolated local origin
• Unify MCP Source control for the built-in and external Agents: Execute can test, connect, explore, and materialize; Admin can add, replace credentials, or remove; connected sources attempt one safe reconnect after restart

• Reference workspace objects directly in built-in Agent chat with @ mentions, including Plans, Reports, saved SQL tabs, and supported data tables or files
• Drag supported sidebar items into chat or use Add to chat to create the same verified object reference
• Add per-workspace Conversation timeout and History depth controls for long-running Agent work and retained chat context

• Switch built-in chat between Agent and Ask: Ask can read and run SQL, but will not write Plans, Reports, HTML, or SQL tab content
• After drafting a Plan, Agent runs a short grilling Q&A (one question at a time with a recommended answer) so metrics and scope land in the Plan before Execute
• HTML Report scoped edits: Chart.js value labels and named annotations; precise in-chunk string replace; insert or delete sections without a full re-generate

• Persist SQL sidebar: group tabs by project (with Unscoped), plus Archive soft-hide that keeps open editors and Plan Progress Steps
• External Agent (MCP): All-workspaces Execute as single-line dual-open—one selected writer workspace at a time; switching clears grants and pending reviews
• Execute-mode Admin approvals: richer pending cards, short expiry, and Approve that opens Connections to finish credential handoff when secrets are required

• Pro data sources: Additional folder, Cloud folder (S3/OSS/ADLS), and SQL Server / Azure SQL with staged remote queries and large-result confirmation
• Import local documents (PDF, DOCX, and more) into Project Reference Docs with managed images, source traceability, and portability; right-click Copy image in Visual view
• External Agent (MCP): Persistent connections, Observe / Execute / Admin modes, External Activity with undo, and All-workspaces discovery-only tools/list until a workspace is bound

• Project Reference Docs: dedicated sidebar subtree with ref_N.md files; Update Intro via agent proposal or Enter manually; [Ref:] rules in User Memory; Agent read_reference and patch_reference tools
• Reference tabs show orange labels in Tab bar and Compare view; portability export includes reference files
• Workspace and Full backup export can optionally include physical data (Source Run cache files); import wizard supports preview, conflict resolution, and large-file copy progress

• Annotation sidecar for Plan and Report: comments in `*.annotations.json`; Markdown persists id-only `<mark data-annotation-id>` tags
• New annotations and Re-anchor write clean sidecar metadata (no `anchor.span` / `span_text`); legacy sidecar spans ignored; removed Ctrl+S `flushSpansToSidecar`
• Legacy `data-comment` inline marks auto-migrate to sidecar with clean anchors on open

• Optional passphrase encryption for all `.lantide` export types (Project, Workspace, Application profile, Full backup)
• Encrypted bundles use container v2 (`LANTIDE` magic + plaintext header + AES-256-GCM); legacy ZIP bundles (`PK` header) remain fully supported
• Import flow: inspect bundle kind without a passphrase → unlock panel for encrypted files → full preview → import or rebind

• Inject a compact current timestamp into Agent context on every turn for reliable date-relative analysis
• Renaming a workspace now moves `chat_sessions` and `agent_sessions` data together with `workspaces/`; deleting a workspace also removes companion session directories
• Tab Source Run caches are preserved when renaming the active workspace (registry re-key, not cleared)

• Restructure USER_GUIDE sections 1–2 around Agent project analysis workflow
• Strengthen Agent analysis quality: machine-readable Execution Strategy, planning and executing guardrails, and improved D3 prompts
• Relax SQL complexity gate for multi-table JOINs; show non-blocking cache optimization hints on complex queries

• Add Agent Preferences onboarding after first API key setup to personalize reply language and style
• Add Help → Report issue to collect diagnostics and export a zip for bug reports
• Fix packaged installs that created a nested data folder; migrate existing data on startup