Use the public login or register routes. After authentication, FamiliarOS routes you into the authenticated shell at /app.
The shell keeps the project dashboard, editor, graph/search tools, Runboard, and the optional studio surfaces under one authenticated navigation frame.
+Download the latest build for your platform from the releases page:
+FamiliarOS-*-mac-arm64.dmgFamiliarOS-*-mac-x64.dmgFamiliarOS-*-win-x64-setup.exeFamiliarOS-*-linux-x86_64.AppImageCurrent builds may be unsigned, so macOS or Windows may show a security warning the first time. Launch the app and your default Familiar appears immediately.
), }, { - id: 'new-project', - title: 'Creating your first project', + id: 'create-familiar', + title: 'Create your Familiar', body: ( <> -Start in the dashboard and choose New Project. FamiliarOS uses the native corpus project service, not a legacy backend project path.
-A newly created project opens directly in the editor workspace, where the project file tree, authoring surface, and preview lanes are already connected.
+Open Settings from the tray icon or floating chat. Under General you can name your Familiar and set a character prompt that shapes its personality.
+The name appears in the tray tooltip and context menu. The character prompt is applied to future assistant replies, so you can make your Familiar terse, enthusiastic, technical, or whatever fits your workflow.
+ + ), + }, + { + id: 'first-chat', + title: 'Your first conversation', + body: ( + <> +Double-click the desktop Familiar to open the floating chat window. Type a message and press Enter to send. Use Ctrl+Enter to insert a newline.
+If you have configured an API key in Settings, your Familiar replies through your chosen provider. If not, it will prompt you to add one.
), }, { id: 'route-map', - title: 'Understanding the main app routes', + title: 'Main app surfaces', body: ( <> -The most important authenticated routes are:
+The FamiliarOS desktop app centers around a few surfaces:
/app for Projects/app/editor/:projectId for authoring/app/search and /app/corpus-graph for inspection/app/runboard for operational workflows/app/research-agent and /app/research-topic/new for AI research swarm sessions/app/skills-manager for enabling model-agnostic AI capabilities/app/openapi-routing for LLM provider routing, tier configuration, and environment vault/app/manim-sandbox for AI-driven Manim animation authoring and storyboard galleryFamiliarOS is designed as one editor-centered workbench. The dashboard, editor, search, Corpus Graph, Runboard, and optional studios are routed surfaces inside one shared shell rather than separate products with disconnected state.
-In practice, that means you should be able to move from a document to Runboard verification, into Search or Graph inspection, then back into the editor without losing project context or handoff provenance.
+Double-click the desktop Familiar to open the floating chat. Click the close button or press Escape to hide it. The chat stays always-on-top so it is available while you work in other apps.
+ + ), + }, + { + id: 'conversations', + title: 'Conversations and history', + body: ( + <> +Click the History button to see past conversations. Each thread is stored separately. You can create a new conversation, switch between threads, or delete old ones. Auto-generated titles come from your first message.
+Chat history is stored locally and can be searched. Relevant past excerpts are injected into the system prompt so your Familiar can reference earlier discussions.
+ + ), + }, + { + id: 'attachments', + title: 'Attachments', + body: ( + <> +Use the attachment button to select files from your computer. Attached files can be sent with your prompt. You can also store attachments directly into the Knowledge Store from the chat toolbar so your Familiar can retrieve them later.
+ + ), + }, + { + id: 'editor-mode', + title: 'Editor mode', + body: ( + <> +Toggle the Editor button to expand the chat into a larger history view. This is useful for reviewing long exchanges or copying earlier messages.
), }, ], }, { - id: 'editor-compile', - title: 'Editor and Compile', - description: 'Explorer, authoring controls, compile flow, preview, and Native PDF.', + id: 'memory', + title: 'Memory System', + description: 'How FamiliarOS remembers, retrieves, and lets you manage memory.', articles: [ { - id: 'editor-layout', - title: 'How the editor workspace is laid out', + id: 'memory-kinds', + title: 'Memory kinds', body: ( <> -The editor is a three-zone workspace:
+FamiliarOS stores four kinds of memory:
The explorer can be hidden or shown with Ctrl + Shift + E. The preview pane can be hidden or shown with Ctrl + P.
+Each memory has importance weighting and optional tags. The system retrieves relevant memories automatically during chat.
), }, { - id: 'editor-controls', - title: 'What the formatting row is for', + id: 'capture-patterns', + title: 'Capture patterns', body: ( <> -The formatting row is the main high-frequency authoring control area. It now owns the preview visibility control and supports both basic and advanced authoring actions rather than scattering those controls across the shell.
-Use it for common text structure, insertion, and workspace visibility tasks while keeping the document title, glyph tools, save, and compile actions in the shared editor chrome.
+You can ask your Familiar to remember things explicitly, for example: "Remember that I prefer dark themes" or "My name is Alex". The memory engine also captures patterns like remember ..., my name is ..., and I prefer ....
The compile action generates real served project artifacts, including bundle.pdf and bundle.synctex.gz. Preview and Native PDF both depend on that artifact contract.
Current editor content is forwarded in compile requests, so the system is no longer limited to stale saved state when generating preview output.
-A normal compile now opens the standard Preview tab first so the result appears in the expected authoring lane. Native PDF remains available as a separate tab for deeper inspection after the compile completes.
+Open Settings → Memory to inspect, search, edit, or delete stored memories. You are always in control of what your Familiar remembers.
), }, { - id: 'sync-surfaces', - title: 'Preview vs Native PDF', + id: 'chat-history-memory', + title: 'Chat history memory', body: ( <> -Preview is the fast embedded output lane for normal authoring flow. Native PDF is the deeper PDF-oriented lane for higher-fidelity inspection and SyncTeX-oriented workflows.
-Use the View menu to switch directly to either PDF Preview Tab or Native PDF Tab. Both are first-class panes in the editor rather than separate external tools.
- - ), - }, - { - id: 'artifact-backed-outputs', - title: 'Compile artifacts, retained outputs, and disposable cache', - body: ( - <> -FamiliarOS treats generated outputs in three different ways: authoritative source files, user-facing artifacts, and rebuildable cache residue. Generated does not automatically mean disposable.
-Compile bundles, SyncTeX outputs, exported publication artifacts, and saved visualizations should be treated as meaningful project outputs when they are still referenced by the workflow. Temporary cache directories and invalidation-only byproducts are the parts that are safe to rebuild or clear.
+Cross-conversation search injects relevant excerpts from past conversations into the active system prompt. This makes your Familiar feel continuous without requiring every detail to be stored as an explicit memory.
), }, ], }, { - id: 'history-review', - title: 'History, AI Memory, and Collaboration', - description: 'Local history, AI Memory, and multi-user review behavior.', + id: 'knowledge-store', + title: 'Knowledge Store', + description: 'Upload files and let your Familiar retrieve relevant context.', articles: [ { - id: 'local-history', - title: 'Local history and checkpoints', + id: 'upload-files', + title: 'Upload files', body: ( <> -Local history is the project-state continuity layer. Use it for checkpoint creation, comparison, and restore. It is meant for document history and recovery, not only for AI state.
-Restore flows are designed to preserve reversibility by keeping prior state available through checkpointing rather than destructively replacing your working context.
+Open Settings → Knowledge Store and upload text-like files such as Markdown, code, CSV, YAML, JSON, or plain text. FamiliarOS extracts the text, infers MIME types, and indexes the content for search.
), }, { - id: 'ai-memory', - title: 'AI Memory and checkpoint restore', + id: 'search-scoring', + title: 'Search and scoring', body: ( <> -The AI Memory surface stores structured checkpoint history for reasoning, review, and operator continuity. It supports preview-before-restore, checkpoint restore, comments, and lineage-aware state continuity.
-Use this when you need to preserve the decision trail around a document workflow, not only the raw file contents.
+The Knowledge Store scores results by token and phrase overlap, with bonuses for phrase matches, recency, and penalties for very large files. Relevant content is injected into the chat system prompt automatically.
), }, { - id: 'memory-lanes', - title: 'How local history, AI Memory, collaboration, and Memory Domain differ', + id: 'store-from-chat', + title: 'Store attachments from chat', body: ( <> -Local history is the authoritative lane for project and document checkpoints. AI Memory is the reasoning-continuity lane. Collaboration tracks shared live-session state such as presence, threads, and reconnect-aware save replay. Memory Domain Intelligence stores durable risk and readiness signals such as drift, provenance, and submission-readiness snapshots.
-Assistive summaries or convenience overlays can sit on top of those lanes, but they should remain optional and never silently replace the raw history or evidence-bearing state underneath.
+When you attach a file in the floating chat, a toolbar button lets you store it directly into the Knowledge Store. This keeps your library and your chat workflow connected.
), }, { - id: 'collaboration', - title: 'Real-time collaboration', + id: 'file-limits', + title: 'File limits', body: ( <> -FamiliarOS supports production-ready concurrent editing. The key user-facing behaviours are:
-Unsafe overlapping edit cases (asymmetric replacement conflicts) are intentionally left explicit instead of auto-merged, because preserving document integrity is more important than speculative merge automation.
- - ), - }, - { - id: 'queued-save-replay', - title: 'What happens if connectivity drops', - body: ( - <> -The collaboration layer supports bounded queued-save replay for reconnect scenarios. In practical terms, the system can preserve pending save intent and replay it after connectivity returns, instead of silently pretending a remote save succeeded.
+The current default limits are 100 files total and 5 MB per file. Invalid files — empty files, non-text content without a placeholder, or path-traversal names — are rejected.
), }, ], }, { - id: 'search-graph-runboard', - title: 'Search, Graph, and Runboard', - description: 'Inspection and operational surfaces beyond the editor.', + id: 'voice', + title: 'Voice & TTS', + description: 'Configure voices, providers, and offline speech.', articles: [ { - id: 'search', - title: 'What Search is for', + id: 'system-voice', + title: 'System voice', body: ( <> -Search is the main receiver surface for many cross-tool pivots. It is where you inspect document or entity matches, preserve handoff context, and decide whether to move into Corpus Graph, Runboard, or back into the editor.
+The default uses your operating system's speech engine through the Web Speech API. You can match a specific voice name and adjust speed between 0.5× and 2.0×.
), }, { - id: 'graph', - title: 'What Corpus Graph is for', + id: 'cloud-tts', + title: 'Cloud TTS providers', body: ( <> -Corpus Graph is the entity and relationship exploration surface. It now respects explicit-root handoff context when entered from other routed tools, which keeps graph inspection tied to the exact entity or provenance that sent you there.
+Connect OpenAI TTS, ElevenLabs, or an OpenAI-compatible endpoint such as OpenRouter or LiteLLM. Credentials are stored with platform encryption when available. Dynamic voice lists are fetched for ElevenLabs.
), }, { - id: 'runboard', - title: 'What Runboard is for', + id: 'piper', + title: 'Local Piper TTS', body: ( <> -Runboard is the operator surface for execution, diagnostics, verification, and remediation routing. It is where you inspect run state, queue health, verify-suite surfaces, and action-oriented follow-up rather than treating execution as a background black box.
+Piper runs as a local process for fully offline speech. Configure the path to the piper binary, select a voice model, and test playback from Settings. Audio is validated by magic bytes before playing.
), }, { - id: 'patches-kanban', - title: 'Patch Queue and AI Kanban', + id: 'quiet-hours', + title: 'Quiet hours', body: ( <> -Patch Queue is the review-oriented change lane. AI Kanban is the board-oriented organization lane for agentic workflow, triage, and remediation follow-through. They complement Runboard and the editor rather than replacing them.
- - ), - }, - { - id: 'paper2all', - title: 'Paper2All export and publication delivery', - body: ( - <> -Paper2All is the publication-delivery lane for packaging document outputs beyond a single editor compile. It keeps export-oriented handoff work inside the same operator context as compile, review, and readiness assessment.
-Treat it as the final-mile delivery surface once a manuscript is stable enough to leave the drafting loop and enter publication-oriented packaging.
+Set quiet hours in Settings to skip speech during focused time. The familiar will still process requests but will not speak out loud.
), }, ], }, { - id: 'studios', - title: 'Studios and Optional Surfaces', - description: 'The specialized routed surfaces beyond the core editor.', + id: 'mcp-tool-servers', + title: 'MCP Tool Servers', + description: 'Give your Familiar real tools through the Model Context Protocol.', articles: [ { - id: 'whiteboard', - title: 'Whiteboard', + id: 'tool-activation', + title: 'Tool activation', body: ( <> -Whiteboard is the product-facing name for the Excalidraw-backed lane. It now fails soft when no stable scene context is available, instead of trapping you behind a hard crash state.
-If no scene is available, use a project-aware route or enter through AI Diagram Studio when scene review is the real goal.
+Open Settings → MCP Tool Servers and select the tools you want to enable: filesystem, terminal, web fetch, sequential thinking, Git, GitHub, Docker, Playwright, SQLite, memory, and reasoning tools.
+Your Familiar can then use those tools in a ReAct-style loop with up to five iterations per request.
), }, { - id: 'mindmap', - title: 'MindMap', + id: 'permission-model', + title: 'Permission model', body: ( <> -MindMap is no longer a read-only novelty view. It is a routed projection surface with selected-node pivots into Search, Corpus Graph, Editor, and Runboard where stable metadata exists.
+Tools run inside the desktop app. You choose which tool families are active. The Curated MCP Toolkit provides a permission-aware reference surface so you can adopt new tools deliberately rather than enabling everything.
), }, { - id: 'pdf-intake', - title: 'PDF Intake', + id: 'safe-defaults', + title: 'Safe defaults', body: ( <> -PDF Intake is the PDF-to-Markdown intake and QA surface. It probes adapter readiness, benchmarks canonical PDF source manifests, inspects provenance-heavy results, and supports triage, remediation exports, search-first investigation, and Runboard/PDF compare handoff.
- - ), - }, - { - id: 'diagram-video', - title: 'AI Diagram Studio and Video Studio', - body: ( - <> -AI Diagram Studio is the product-facing name for the `next-ai-draw-io` lane. It is an AI-assisted diagram surface for planning, enhancing, validating, and routing diagram work through natural-language-driven workflows.
-Video Studio is the Remotion-backed orchestration and preflight surface. It prepares template, timing, render-profile, and handoff work, but it does not yet include a model-driven text-to-video or image-to-video generation backend.
-Both support explicit routed handoffs instead of ending at isolated summary states. That means they can hand you into Whiteboard, Runboard, or the editor while preserving provenance.
+By default, no tools are enabled. Activate only the tools you need. Terminal and filesystem tools can read or modify your local machine, so review each tool's permissions before enabling it.
), }, ], }, { - id: 'ai-settings', - title: 'AI, Settings, and Account', - description: 'Configuration-oriented user questions.', + id: 'agent-bridge', + title: 'External Agent Bridge', + description: 'Let coding agents control your Familiar safely.', articles: [ { - id: 'ai-orientation', - title: 'How AI fits into the product', + id: 'enable-mcp-server', + title: 'Enable the FamiliarOS MCP Server', body: ( <> -AI in FamiliarOS is part of the workspace, not a bolt-on chatbot. The important concepts are provider routing, reasoning continuity, source-grounded memory direction, optional assistive overlays, and user/operator control over the environment.
-Current AI-enabled surfaces include the AI assistant, AI Memory, AI Diagram Studio, AI Kanban, provider-routed generation tasks, and AI-aware execution or verification flows that preserve model context through the workspace.
-For deeper integration references, use docs/CROSS_CORPUS_CANONICAL_CONSOLIDATED.md and docs/references/integrations/OPTIONAL_EXTERNAL_REPO_USAGE.md.
Open Settings → Integrations → FamiliarOS MCP Server. Choose the command source, override the Node path if needed, pick which Familiar the agent controls, and run the health test.
), }, { - id: 'provider-neutral-routing', - title: 'Provider-neutral routing and BYOK at a glance', + id: 'copy-json', + title: 'Copy MCP JSON', body: ( <> -OpenAPI Routing is designed as a provider-neutral control surface. It lets you inspect tier, provider, model, and base-URL choices without treating any single model vendor as the product's default worldview.
-Bring Your Own Key is the same way: Scholar+ users can supply personal provider credentials for compatible routes, but those keys remain under first-party Scriptorium controls and are currently forwarded through a bounded request-scoped bridge rather than becoming a hidden server-side black box.
+Once the server is healthy, click Copy MCP JSON to get the mcpServers.familiaros entry. Paste it into your agent's MCP config for Claude Code, OpenCode, Cursor, or Codex CLI.
Settings is the environment-facing preferences surface. Use it for preferences, configuration choices, and integration-facing controls. It is not meant to replace Runboard or editor-level workflow controls.
- - ), - }, - { - id: 'account-menu', - title: 'What the account menu is for', - body: ( - <> -The top-right avatar area is now a real account menu. Use it for identity glance, account/session actions, and fast pivots such as settings or sign-out rather than treating it as a decorative placeholder.
- - ), - }, - { - id: 'pricing', - title: 'Where pricing and subscription details live', - body: ( - <> -The public pricing page is the live packaging reference. This Help Center intentionally avoids freezing plan/billing details into a static FAQ when those details may change more often than product surface behavior.
- - ), - }, - { - id: 'tutorial-mode', - title: 'Tutorial Mode and help flags', - body: ( - <> -Tutorial Mode adds instructional overlay flags to the annotated menu bar entries, account-menu actions, editor chrome controls, formatting-toolbar groups, and right-side workspace controls. When active, hovering over a supported UI element shows a small Tutorial badge with a plain-English description of what that element does.
-To toggle Tutorial Mode:
-scriptorium.tutorial-mode.v1).Tutorial Mode is designed for onboarding. It adds no runtime overhead when disabled — the tooltip wrapper renders children directly without any additional DOM nodes.
-The current canonical feature inventory that public docs and Writerside should mirror is maintained in docs/ops/SCRIPTORIUMAI_FEATURE_MANUAL.md.
To deactivate, open the account menu again and select Disable Tutorial Mode.
+The external-agent bridge is intentionally small. Agents can make your Familiar react, speak a safe bubble, or read and write memory. They cannot browse your filesystem, run terminal commands, or access tools that you have not explicitly enabled elsewhere.
), }, ], }, { - id: 'affiliate-marketplace', - title: 'Affiliate Marketplace', - description: 'Open-marketplace onboarding, disclosure, tracking, and assets.', + id: 'familiar-packs', + title: 'Familiar Packs', + description: 'Choose, import, and switch companion characters.', articles: [ { - id: 'affiliate-open-marketplace', - title: 'How the affiliate marketplace works', + id: 'default-familiar', + title: 'Default Familiar', body: ( <> -FamiliarOS is launching its affiliate lane as a broad marketplace. There is no separate agency or formal partner tier at launch. Creators, educators, reviewers, consultants, and workflow operators all enter through the same public marketplace path.
-Use Affiliate Program for onboarding and Affiliate Resources for assets, rules, and tracking guidance.
+FamiliarOS ships with a default companion. You can name it and adjust its character prompt in Settings. The default Familiar is used when no other routing rule applies.
), }, { - id: 'affiliate-tracking', - title: 'What the affiliate tracking model includes', + id: 'import-packs', + title: 'Import packs', body: ( <> -The target model is first-party plus server-side attribution with affiliate IDs, click IDs, optional sub-IDs, coupon-code fallback, and clean direct URLs. The goal is to keep attribution explainable later instead of treating affiliate credit as a black box.
+Import Familiar packs from a ZIP file or folder. Installed familiars appear in the gallery where you can preview animations and select a default.
), }, { - id: 'affiliate-compliance', - title: 'Disclosure and compliance expectations', + id: 'per-agent-routing', + title: 'Per-agent Familiar routing', body: ( <> -Affiliate content must clearly disclose the material connection. Channel rules, privacy-aware tracking, allowed placements, and fraud controls are part of the program specification, not optional fine print.
+When an external agent requests a specific Familiar, lease routing resolves to the requested installed pack or falls back to the default. You can override which Familiar each agent controls from the FamiliarOS MCP Server panel.
), }, ], }, { - id: 'citation-intelligence', - title: 'Citation Intelligence', - description: 'BibTeX auto-completion, citation graph memory, LaTeX error context, and integrity forecasting.', + id: 'troubleshooting', + title: 'Troubleshooting', + description: 'Common issues and how to fix them.', articles: [ { - id: 'bibtex-autocomplete', - title: 'BibTeX auto-completion from DOI and ArXiv', + id: 'macos-quarantine', + title: 'macOS says the app is damaged', body: ( <> -FamiliarOS can look up and generate a complete BibTeX entry from a DOI, ArXiv ID, or URL. Use the citation lookup route in the editor or Runboard to retrieve properly formatted entries without leaving the workspace.
-DOI resolution uses CrossRef content negotiation. ArXiv IDs resolve to the Atom feed and upgrade to a DOI-linked entry when available. Arbitrary URLs are dispatched through the best available resolver.
+Unsigned macOS builds may be quarantined. Remove the quarantine flag and reopen:
++ xattr -dr com.apple.quarantine /Applications/FamiliarOS.app + open /Applications/FamiliarOS.app +), }, { - id: 'citation-graph-memory', - title: 'Citation graph memory and event tracking', + id: 'windows-security', + title: 'Windows security warning', body: ( <> -
Every time cite-keys are added or removed from a LaTeX file, FamiliarOS can record a citation graph event — capturing which keys appeared, which disappeared, and the net delta for that commit or session. Over time this builds a per-project citation history.
-The citation summary surfaces your top-cited keys and most volatile files, giving you an empirical picture of how citation dependencies are evolving across the manuscript.
+Current Windows builds are not signed, so SmartScreen may warn you. Click More info and then Run anyway if you trust the download source.
), }, { - id: 'reference-sync', - title: 'Reference sync and bibliography continuity', + id: 'tts-fails', + title: 'TTS does not play', body: ( <> -Reference sync keeps your bibliography workflow connected to the writing surface instead of forcing manual export-import loops. It is intended for citation continuity: keeping libraries, cite-keys, and project-local bibliography state from drifting apart.
-Use it when you need bibliographic updates to stay traceable inside the same corpus and review workflow as the manuscript itself.
+Check that the selected provider has valid credentials and that the voice/model combination exists. For Piper, verify the binary path and that the model file is present. Check quiet-hours settings and system volume.
), }, { - id: 'latex-error-context', - title: 'LaTeX error context and fix suggestions', + id: 'mcp-not-responding', + title: 'MCP server not responding', body: ( <> -The compile diagnostics panel enriches raw LaTeX errors with plain-English explanations and actionable fix suggestions from a built-in knowledge base covering undefined references, missing packages, math mode mismatches, and other common failure categories.
-Each enriched error includes a broken-versus-fixed code example and machine-readable fix actions (insert / replace / recompile / install) so remediation can be applied directly from the panel.
- - ), - }, - { - id: 'citation-forecast', - title: 'Citation integrity forecasting', - body: ( - <> -The citation integrity forecast analyzes your citation graph event history to predict which cite-keys are at risk of breakage before your next submission. High-churn keys — those frequently added and removed across sessions — are flagged with a risk level (low / medium / high).
-Citation breakages (undefined citations, missing BibTeX entries, key renames, moved files) can also be recorded explicitly so the forecast model improves over time.
+Run the health test in the MCP Server panel. Confirm the Node path is correct and that the FamiliarOS desktop app is running. Check that no other process is holding the server's port.
), }, ], }, { - id: 'breathe-memory', - title: 'Memory Domain Intelligence', - description: 'Governance drift, integration boundary, provenance ledger, cost-adaptive analysis, offline queue, section churn, and submission readiness.', + id: 'privacy', + title: 'Privacy & Security', + description: 'How your data stays under your control.', articles: [ { - id: 'governance-drift', - title: 'Governance drift detection', + id: 'local-first', + title: 'Local-first model', body: ( <> -FamiliarOS can record a governance configuration baseline for your project and then detect drift as your corpus governance rules evolve. When a detect-drift check is run, the system compares the current configuration against the baseline and returns a drift score (minor / moderate / severe) along with the specific keys that changed.
-This is useful for catching unintentional rule erosion in long-horizon projects where policy changes are made incrementally and the total cumulative drift is not obvious from any single diff.
+Memory, Knowledge Store files, and chat history are stored locally in your user data directory. Cloud providers are only contacted when you configure your own API key and send a chat request.
), }, { - id: 'integration-boundary', - title: 'Integration boundary memory', + id: 'credential-storage', + title: 'Credential storage', body: ( <> -Every tool that FamiliarOS integrates with — PDF Intake, the PDF visual diff engine, SimpleMindMap, AI Diagram Studio, Remotion, Zotero, and others — can emit integration boundary events. These events record which tool was called, what operation was performed, the call direction, and when it happened.
-The integration boundary summary shows your top-used tools, direction counts (outbound / inbound / internal), and a full timestamped event log — giving you an auditable record of every external interaction a project has had.
+API keys and TTS credentials are encrypted with Electron safeStorage when the platform supports it. If encryption is unavailable, a plain local fallback is used and noted in Settings.
Every insight, suggestion, or AI-generated result in FamiliarOS can be linked to a provenance entry that records where it came from: which source references were used, what transformations were applied, what confidence level was assigned, and how the conclusion was reached.
-Use the Explain Insight function to retrieve the full provenance chain for any stored insight ID — so any AI-driven recommendation can be traced back to its evidence sources rather than treated as a black-box assertion.
- - ), - }, - { - id: 'cost-adaptive-analysis', - title: 'Cost-aware adaptive analysis depth', - body: ( - <> -For resource-intensive analysis operations, FamiliarOS tracks a per-project credit budget. The adaptive depth engine automatically adjusts analysis depth — full, standard, or shallow — based on remaining credits, repository size, and recent activity velocity.
-Operators can view the current budget, update the depth policy, and monitor adaptive depth recommendations from the Memory Domain panel. Credits replenish according to the configured policy cycle.
- - ), - }, - { - id: 'offline-queue', - title: 'Offline-first session memory capture', - body: ( - <> -When a session is offline or connectivity is degraded, FamiliarOS queues memory events locally rather than silently dropping them. When connectivity is restored, queued events can be reconciled in a single bulk operation.
-The offline queue is visible in the Memory Domain panel so you always know how many events are pending reconciliation and can trigger reconciliation manually when needed.
- - ), - }, - { - id: 'submission-readiness', - title: 'Submission readiness memory and trend tracking', - body: ( - <> -FamiliarOS computes a per-project submission readiness score across four weighted dimensions: compile stability (35%), section churn (25%), citation stability (25%), and task density (15%). Scores map to status levels: high risk, moderate risk, approaching ready, or ready.
-Readiness snapshots are recorded over time so you can track whether your manuscript is improving or declining in submission-readiness. The trend indicator — improving, stable, or declining — is based on the delta between your most recent and oldest recorded snapshots.
- - ), - }, - ], - }, - { - id: 'visualization', - title: 'Workspace Visualization System', - description: 'Layout presets, working-set tabs, graph heatmaps, glyph packs, governance cockpit, and external capability surfaces.', - articles: [ - { - id: 'layout-presets', - title: 'Layout presets and pane synchronization', - body: ( - <> -FamiliarOS ships four workspace layout presets — Writing, Triage, Graph, and Governance — each configuring a curated pane arrangement optimized for that workflow. Switch between presets from the layout bar without losing your current document context.
-Pane synchronization links the active selection across panels so that selecting an entity in Corpus Graph, for example, also highlights it in the Search or Editor surface.
- - ), - }, - { - id: 'working-set-tabs', - title: 'Working-set tabs and jump history', - body: ( - <> -The working-set tab bar tracks the documents and surfaces you have open in the current session. Tabs can be pinned (persisted across sessions), reordered, marked dirty, and closed individually or in bulk (close all unpinned).
-Jump back and forward navigation lets you retrace your path through the session without relying on browser history. The maximum working set enforces a reasonable open-tab cap to keep the shell performant.
- - ), - }, - { - id: 'graph-heatmaps', - title: 'Graph heatmaps and cluster expansion', - body: ( - <> -Graph mode in the Corpus Graph surface supports heatmap overlays that visualize entity density, activity recency, or citation weight. Cluster expansion lets you progressively reveal subgraph structure without rendering the full graph at once.
- - ), - }, - { - id: 'glyph-packs', - title: 'Glyph palette and corpus glyph packs', - body: ( - <> -The editor glyph system supports corpus-aware glyph packs — structured collections of frequently used symbols, notation shorthand, or domain-specific LaTeX fragments. Glyph bootstrap respects the live server cap. Packs can be switched or extended per project.
- - ), - }, - ], - }, - { - id: 'ai-orchestration', - title: 'AI Orchestration and Agent Tools', - description: 'Autonomous research-agent swarms, Skills Manager, OpenAPI routing configuration, and Manim animation sandbox.', - articles: [ - { - id: 'research-agent-overview', - title: 'What the autonomous Research Agent is and how it works', - body: ( - <> -The Research Agent surface at /app/research-agent manages multi-perspective research swarm sessions. Each session launches a coordinated batch of sub-agents that explore a topic from different academic angles, synthesise findings, and store results as corpus runs.
Sessions are created as live corpus runs via /api/corpus/runs using kind=research_agent plus a structured provenance payload. Pending sessions remain visible on the dashboard until a worker claims and executes them.
Session status badges reflect the live state returned by the server: active (swarm in progress), completed (results ready), failed (swarm error), and pending (queued but not yet started). The dashboard separates active and completed sessions into tabs and shows aggregate counters at the top.
- - ), - }, - { - id: 'new-research-topic', - title: 'Launching a new research swarm', - body: ( - <> -Use /app/research-topic/new to configure and launch a new research session. You must provide a topic title. Optionally supply background context or known URLs and choose a depth strategy:
Known URLs and research constraints are passed through the run provenance so the orchestration layer can recover the operator intent when the queued run is picked up.
- - ), - }, - { - id: 'skills-manager', - title: 'Skills Manager — enabling and disabling AI capabilities', - body: ( - <> -The Skills Manager at /app/skills-manager shows the full catalog of model-agnostic AI capabilities available in your installation. Skills are provider-neutral building blocks — they work across different LLM backends and can be enabled or disabled without restarting the system.
Use the search bar to filter by name and the category dropdown to narrow by domain (for example formatting, compilation, or exploration). The enabled count badge at the top reflects the current local override state.
-Toggles write an optimistic local override immediately so you see the change without waiting for a server round-trip. The override is persisted to localStorage under the key scriptorium.skills.enabled-overrides.v1 and reconciled with the live catalog when the backend is available.
The OpenAPI Routing surface at /app/openapi-routing controls which LLM provider and model is used for each analysis tier. FamiliarOS uses three tiers:
For each tier, choose from the supported provider/model options. The Sync from Server button fetches the current capabilities handshake and updates the key vault status badges — showing whether each provider currently has a live API key configured in the server environment.
-If the server policy locks base URL overrides, the base URL field will be read-only. Otherwise, you can supply a custom base URL to point a provider at a local or proxied endpoint. Tier/provider preferences remain local in this checkout; the active server bridge currently applies to compatible Scholar+ BYOK direct-chat requests and queued AI assistant tasks.
- - ), - }, - { - id: 'manim-sandbox', - title: 'Manim Sandbox — AI animation authoring and storyboard gallery', - body: ( - <> -The Manim Sandbox at /app/manim-sandbox is a two-panel surface: the left side is the interactive Manim authoring panel (code + prompt input for AI-assisted scene generation) and the right side is the storyboard gallery showing rendered visualization artifacts.
The gallery is populated from two sources in order of preference:
-listCorpusArtifacts filtered to type=manim, visualization, and animationlocalStorage under scriptorium.visualizations.local-renders.v1Selecting a gallery entry opens the Render Preview panel, which displays the video or image artifact and provides a direct download link. The View All Visualizations link at the bottom of the gallery navigates to the full Visualizations surface at /app/visualizations.
The AI Visualizer backend (LaTeX-OCR-Service render endpoint) must be running for new renders to be submitted. Existing local renders are always accessible regardless of server availability.
The AI Chat Panel lives inside the editor workspace as a collapsible right-hand pane. Open it by clicking the chat icon in the editor toolbar or pressing the shortcut key configured in your workspace settings.
-The panel has two operating modes:
-Supported models (selectable from the dropdown in the panel header):
-OPENAI_API_KEY)ANTHROPIC_API_KEY)GEMINI_API_KEY)QWEN_API_KEY if you are using your own quota)Responses stream token by token. Click the × button to cancel a running request. The abort is clean — partial tokens are discarded and the panel returns to its ready state.
- - ), - }, - { - id: 'byok-guide', - title: 'BYOK — Bring Your Own API Key (Scholar+)', - body: ( - <> -Scholar and higher plan holders can supply personal provider API keys through the Environment Vault section of OpenAPI Routing. This lets you use your own billing accounts for OpenAI, Anthropic, Google, Qwen, or OpenRouter instead of server-provisioned keys.
-How to use it:
-/app/openapi-routing and scroll to OpenAPI (BYOK) — Bring Your Own Key in the Environment Vault card.OPENAI_API_KEY for OpenAI, QWEN_API_KEY for Qwen/DashScope).⚠️ The current activation slice is request-scoped, not fully persisted server-side. Compatible direct chat and queued AI assistant tasks now forward OpenAI/OpenRouter/Qwen BYOK keys through explicit request headers, while broader provider adapters, deeper persistence, and non-LLM worker families remain future work.
-For Qwen, the default base URL is https://dashscope.aliyuncs.com/compatible-mode/v1. Local-model deployment is intentionally not part of the live product surface at this time.
Qwen 3.6 Plus is available as a managed provider route inside FamiliarOS's provider-neutral orchestration layer.
-QWEN_API_KEY in OpenAPI Routing if you want to use your own quota instead of the managed pool.To select Qwen in the editor, open the AI Chat Panel and choose Qwen 3.6 Plus from the model dropdown. For routing-level use, set Tier 3 to Qwen 3.6 Plus in OpenAPI Routing.
-Local-model deployment is not currently exposed on the live product. Operational planning for any future local-model activation is kept in repository documentation rather than presented as a live end-user feature today.
- - ), - }, - ], - }, - { - id: 'ops', - title: 'Ops and Self-Hosting References', - description: 'Where to go for technical and operational references.', - articles: [ - { - id: 'ops-docs', - title: 'Operational source of truth', - body: ( - <> -If you are self-hosting or operating FamiliarOS, the main references are the repository docs, especially:
-docs/ops/RUNBOOK.mddocs/ops/PRODUCTION_CUTOVER_AND_ROLLBACK_CHECKLIST.mddocs/ops/SCRIPTORIUMAI_LOCAL_MODEL_DEPLOYMENT_REFERENCE.mddocs/STACK_SETUP_CANONICAL_CONSOLIDATED.mddocs/CROSS_CORPUS_CANONICAL_CONSOLIDATED.mddocs/TEST_SUITE_STATUS.mdUse the Support page when you need deployment help, product clarification, or guidance that is not already covered in the public help content.
+FamiliarOS does not show prompts, code, logs, command output, URLs, paths, or secrets in the desktop Familiar's speech bubbles. Bubbles are reserved for short, safe, user-facing messages.
), }, @@ -741,140 +445,107 @@ const helpCategories: HelpCategory[] = [ }, ] -function ArticleItem({ article }: { article: HelpArticle }) { - const [open, setOpen] = useState(false) - return ( -Help Center
-- Use this page when you need route-level orientation, workflow answers, and the main operating rules around workbench continuity, memory lanes, provider routing, and artifact-backed outputs without reading the deeper repository docs. +
+ Guides for installing the desktop app, creating your Familiar, using the floating chat, managing memory and knowledge, configuring voice, enabling MCP tools, and connecting external agents.
-No help articles matched {search}.
- +Categories
+Still need help?
-- Use the support surface for questions that need operator follow-up, deployment help, or product clarification beyond this public help content. -
-{categoryIndex.get(openCategory)!.description}
+