Settings reference
All Vault Operator settings live under Settings > Vault Operator. The settings tab has six groups: Providers, Agents, Customize, Vault, Advanced, and Help. Each group has sub-tabs. This page walks every sub-tab in the order they appear.
UI paths in this page use the format Settings > Vault Operator > {Group} > {Sub-tab} in sentence case, matching the labels in src/i18n/locales/en.ts.
Providers group
Providers
Settings > Vault Operator > Providers > Providers
Configure AI providers. Each provider exposes its own model list and is mapped to three tiers (Budget, Main, Frontier) that the agent picks from based on the current task.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Provider list | All configured providers with type, display name, sign-in status, and tier mapping | Empty | src/types/settings.ts (providerConfigs[]) |
| Active provider | The provider used for chat by default | First added | settings.activeProviderId |
| Add provider | Opens the provider detail modal | n/a | ProviderDetailModal.ts |
| Refresh (in modal) | Pulls the provider's model list and auto-classifies models into tiers | n/a | ProviderDetailModal.ts:985-993 |
| Tier mapping | Manual override for Budget / Main / Frontier slots | Auto-classified | model-registry.ts |
| Test connection | Verifies the provider's credentials and endpoint with a minimal request | n/a | testModelConnection.ts |
The provider modal supports twelve provider types: Anthropic, OpenAI, Gemini, Ollama, LM Studio, OpenRouter, Azure, Custom (OpenAI-compatible), GitHub Copilot, Kilo Gateway, AWS Bedrock, and ChatGPT (OAuth). Ollama and LM Studio prefill their Base URL with the local default port. ChatGPT (OAuth) bills against your existing Plus or Pro subscription instead of a per-token key.
Tiers and overrides
The Main tier drives chat by default. The agent escalates to Frontier on hard synthesis steps via the consult_flagship tool (budget: 3 calls per task, 3000 tokens per call). The chat-header model picker lets you pin a specific provider and model for a single task without changing the active provider.
Plaintext API key warning banner (v3.0.0)
If the OS keychain is unavailable (for example a Linux install without libsecret-1-0), the Providers tab shows a persistent banner titled "API keys stored as plaintext" above the provider list. In this state, API keys, OAuth tokens, and MCP secrets are written as plain strings to data.json and are visible to any process that can read the vault. A button labeled "I understand, dismiss this warning" sets settings.safeStoragePlaintextFallbackAcknowledged and fires a one-time confirmation toast. The banner itself stays visible after dismissal so the degraded state remains clear. Source: src/ui/settings/ProvidersTab.ts.
Per-model reasoning and thinking
Each model row in the provider modal exposes reasoning controls when the underlying model supports them. Pin a specific model in the chat-header picker to use these. Auto mode uses the model's default.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Extended thinking | Enables Claude budget-token thinking on Sonnet 4.6, Opus 4.6 and older, Haiku, and 3.x models | Off | BUILT_IN_MODELS in settings.ts |
| Thinking budget (tokens) | Token budget reserved for the model's internal reasoning before visible output | 10000 (Sonnet/Opus), 5000 (Haiku) | BUILT_IN_MODELS |
| Reasoning effort | Effort level for adaptive Claude (Opus 4.7+, Fable, Mythos) and GPT-5 / o-series. Claude: Low, Medium, High, XHigh, Max. OpenAI: Minimal, Low, Medium, High | Model default | model-registry.ts |
| Max output tokens | Output budget. Auto clamps to the model ceiling and remaining context room | Auto | resolveOutputBudget in model-registry.ts |
Caching reality
Anthropic uses explicit cache_control blocks. Bedrock Claude uses explicit bedrock-cachepoint. OpenAI gpt-4o, 4.1, o1, o3, and o4 use implicit prefix caching. Gemini has no prefix caching (TTL context caching is deferred). DeepSeek is not a registered provider type.
Models (legacy)
Settings > Vault Operator > Providers > Models
Legacy view kept for back-compat. New work happens in the Providers sub-tab via provider modals. Use this only if you need to inspect or remove a model entry that predates the provider-only refactor.
Embeddings
Settings > Vault Operator > Providers > Embeddings
Configure the semantic index for meaning-based vault search. The Embeddings sub-tab has four sections: Embedding models, Semantic index, Index configuration, Graph expansion.
Embedding models
| Setting | What it does | Default | Source |
|---|---|---|---|
| Embedding model list | Configured embedding models with provider, model id, and API key | Empty | settings.embeddingModels |
| Add embedding model | Opens the embedding model modal | n/a | EmbeddingsTab.ts |
The first-run wizard suggests OpenAI text-embedding-3-small or Google text-embedding-004. Other choices are fine if you bring your own provider.
Semantic index
| Setting | What it does | Default | Source |
|---|---|---|---|
| Enable semantic index | Master toggle. Build index is blocked until this is on | Off | settings.enableSemanticIndex |
| Build index | Indexes the vault | n/a | EmbeddingsTab.ts:230 |
| Force rebuild | Deletes the index and re-indexes from scratch. Cancel keeps progress | n/a | EmbeddingsTab.ts:294 |
| Auto-index trigger | When to re-index automatically: never, on startup, on agent switch | never | settings.semanticAutoIndex |
| Auto-reindex on change | Re-index when files change | false | settings.semanticAutoIndexOnChange |
Build index is gated
The Build index button shows "Enable semantic index first." until you turn the master toggle on. Re-indexing on change is off by default. You stay in manual mode unless you opt in.
Index configuration
| Setting | What it does | Default | Source |
|---|---|---|---|
| Chunk size | Chunk size for embedding. Small 800, Medium 1200, Standard 2000, Large 3000 | Standard (2000) | EmbeddingsTab.ts:394-400, en.ts:117-120 |
| Local reranking | Re-rank semantic search results with a local cross-encoder model | On | settings.enableReranking |
Reranker model
The reranker uses Xenova/ms-marco-MiniLM-L-6-v2 and is delivered as an optional asset. If the asset is not installed, the agent falls back silently to the vector score. Install under Settings > Vault Operator > Providers > Embeddings > Local reranking. The JavaScript half (Reranker library, ~0.6 MB) is installed separately under Advanced > Optional assets; both are needed.
Implicit connections
| Setting | What it does | Default | Source |
|---|---|---|---|
| Implicit connections | Compute cosine-similarity edges between notes for discovery and suggestion-banner | On | settings.enableImplicitConnections (EmbeddingsTab.ts:623) |
| Similarity threshold | Minimum cosine similarity to count as an implicit connection (0.5 loose, 0.9 strict) | 0.7 | settings.implicitThreshold (EmbeddingsTab.ts:634) |
| Suggestion banner | Show implicit-connection suggestions in the sidebar | On | settings.enableSuggestionBanner (EmbeddingsTab.ts:646) |
Knowledge properties
Vault conventions used by the knowledge ingest workflow and vault health check. Set these once to match your vault's schema.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Category property | Frontmatter key that holds the note's type or category | type | settings.categoryProperty |
| Backlinks property | Frontmatter key that holds reciprocal backlink wikilinks, used by the vault health repair pass | related | settings.backlinksProperty |
| Summary property | Frontmatter key for the short note summary | description | settings.summaryProperty |
| Source naming convention | Filename pattern for source notes created by ingest | Author-Year_Title | settings.sourceNamingConvention |
| MOC properties | Frontmatter keys that participate in Maps of Content | moc | settings.mocPropertyNames |
Defaults follow the OKF vocabulary
The defaults (type, description, moc, related) follow the OKF frontmatter vocabulary used by the built-in templates. Adapt them in the same panel to match your vault's language and naming; persisted settings always win over the defaults.
Web search
Settings > Vault Operator > Providers > Web search
Enable tools for accessing the internet.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Enable web tools | Allow the agent to use web_fetch and web_search | Off | settings.webTools.enabled |
| Search provider | Which search API to use: None (web_fetch only), Brave, or Tavily | None | settings.webTools.provider |
| API key | Key for the selected search provider; Brave and Tavily each keep their own key | Empty | settings.webTools.braveApiKey, settings.webTools.tavilyApiKey |
Agents group
Agents
Settings > Vault Operator > Agents > Agents
Configure agents. One built-in agent ships: Default agent. You can add custom agents with their own system prompt, tool sets, and per-agent model overrides.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Default agent | The single built-in agent. All built-in tool groups are available | Built-in | builtinModes.ts:58 |
| Custom agents | User-defined agents with custom tool sets and system prompts | Empty | settings.customModes |
| Per-agent model | Override which model an agent uses | Active provider's Main tier | settings.modeModelKeys |
| Per-agent tool overrides | Restrict tool groups for an agent (modeToolOverrides) | None | settings.modeToolOverrides |
| Forced workflow | Per-agent forced workflow, set from the workflows menu in chat: the mapped workflow is applied to every message unless the message starts with a slash command | None | settings.forcedWorkflow |
There is only one built-in agent
The earlier Ask + Agent split was removed in v2.11. For read-only behaviour, either restrict a custom agent's tool groups to read and vault, or set Auto-approve to "ask every time" for the write groups. The mid-conversation mode switcher was removed from the chat header in v2.11.
Auto-approve
Settings > Vault Operator > Agents > Auto-approve
Control what the agent can do without asking. See the safety and control guide for details.
Since ADR-153 the toggles map to effect classes (EFFECT_POLICY in src/core/tools/toolEffects.ts), not to tool groups. A master switch (default off) gates the categories below; reads always run and are not a toggle; two effects can never be auto-approved at all.
| Toggle | What it auto-approves | Default |
|---|---|---|
| (none) Reads | read_file, search_files, semantic_search, listings -- always run, master-independent | Always on |
| Note edits | write_file, edit_file, append_to_file, update_frontmatter, ingest | Off |
| Vault changes | create_folder, move_file, delete_file, extract_zip, restore_checkpoint, canvas/office creators | Off |
| Web | web_fetch, web_search, anti_echo_search | Off |
| MCP | use_mcp_tool, invoke_mcp_server | Off |
| Subtasks | new_task | Off |
| Skills | invoke_skill (trusted built-in / Pro only; imported skills still prompt) | Off |
| Recipes | execute_recipe | Off |
| Plugin API reads / writes | call_plugin_api, split by call shape into two toggles | reads on, writes off |
| Sandbox | evaluate_expression, run_skill_script, dynamic custom_* tools (confirm required) | Off |
Never auto-approvable, regardless of these toggles or any preset: config (update_settings, configure_model, manage_mcp_server) and self-modify (update_soul, mark_for_memory, manage_source). The agent cannot enable its own permissions.
A drift test (autoApprovalConfigDrift.test.ts) enforces that every stored toggle gates a real effect and every effect resolves to a real toggle, so this table cannot silently rot again.
Permissive combination warning
Turning on Web together with a write category lights up a "Permissive" indicator in the Auto-approve tab. That combination lets the agent fetch internet content and act on it without asking.
Loop
Settings > Vault Operator > Agents > Loop
Control how the agent loop runs.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Consecutive error limit | How many consecutive tool errors before the agent stops | 3 | settings.advancedApi.consecutiveMistakeLimit |
| Rate limit | Minimum milliseconds between API calls | 0 | settings.advancedApi.rateLimitMs |
| Max iterations | Maximum tool calls per conversation turn | 25 | settings.advancedApi.maxIterations |
| Context condensing | Summarize older messages when context gets long | On | settings.advancedApi.condensingEnabled |
| Condensing threshold | Percentage of context window before condensing triggers | 80 | settings.advancedApi.condensingThreshold |
| Microcompaction | Compact older tool results in place when their token cost exceeds a threshold | On | settings.advancedApi.microcompactionEnabled |
| Rolling-summary threshold | Percentage of the condensing threshold at which a rolling summary is folded in | 50 | settings.advancedApi.rollingSummaryThreshold |
| Power steering | Re-inject key instructions every N assistant turns (0 disables) | 0 | settings.advancedApi.powerSteeringFrequency |
| Subtask depth | Maximum nesting depth for sub-agents | 2 | settings.advancedApi.maxSubtaskDepth |
| Subtask token budget | Token budget per new_task spawn message | 8000 | settings.advancedApi.subtaskTokenBudget |
| Cost-warn threshold | EUR cost threshold per task that triggers a warning (0 disables) | 0 | settings.advancedApi.costWarnThresholdEur |
| Default main-tier model | Which tier the chat loop uses by default | mid (Main) | settings.defaultMainModelTier |
| Task routing (Helper model) | Model used for context condensing, fast-path planning, plan_presentation, and recipe promotion | Falls back to active provider's Budget tier | settings.helperModelKey |
Memory
Settings > Vault Operator > Agents > Memory
Configure how the agent remembers across conversations.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Chat history | Save conversation history for future reference | On | settings.enableChatHistory |
| Enable memory | Master toggle: allow the agent to build long-term memory from conversations | On | settings.memory.enabled |
| Auto-extract session summaries | Automatically extract memory when a conversation ends. The star button and "remember this" prompts work regardless of this setting | On | settings.memory.autoExtractSessions |
| Minimum messages before extraction | Minimum total messages (user plus assistant) a conversation needs before auto-extraction triggers | 6 | settings.memory.extractionThreshold |
Memory model picker removed in FEAT-24-08
The separate "Memory model" dropdown is gone. The Task routing helper model runs memory extraction.
Customize group
Rules
Settings > Vault Operator > Customize > Rules
Persistent instructions that guide the agent in every conversation.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Rule list | All active rules injected into the system prompt | Empty | settings.rules |
| Add rule | Create a new rule (plain text or Markdown) | n/a | RulesTab.ts |
| Import | Import rules from a file | n/a | RulesTab.ts |
Workflows
Settings > Vault Operator > Customize > Workflows
Slash-command triggered instruction sequences. Type / in chat to invoke.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Workflow list | All workflows with name, trigger, and body | Built-in defaults | settings.workflows |
| Add workflow | Create a new workflow | n/a | WorkflowsTab.ts |
Skills
Settings > Vault Operator > Customize > Skills
Persistent instruction sets matched by keywords. Like mini-manuals the agent follows.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Skill list | All skills with name, trigger pattern, and body | Built-in bundled skills | settings.skills and bundled-skills/ |
| Skill registry | Browse and install skills on demand (including skill-creator) | n/a | SkillRegistryModal.ts |
| Add skill | Create a new skill | n/a | SkillsTab.ts |
Prompts
Settings > Vault Operator > Customize > Prompts
Reusable message templates with optional variables.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Prompt list | All prompts | Empty | settings.customPrompts |
| Add prompt | Create a new prompt | n/a | PromptsTab.ts |
Connectors
Settings > Vault Operator > Customize > Connectors
Connect external tool servers and expose Vault Operator as a server. The Connectors sub-tab has three sections: Local connector, Remote access, External tool servers.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Local connector | Vault Operator as MCP server for desktop clients (Claude Desktop, ChatGPT, Perplexity) | Off | settings.enableMcpServer |
| Remote access | Cloudflare-tunnelled long-polling endpoint with token-in-URL auth | Off | settings.enableRemoteRelay |
| External tool server list | MCP servers the agent can call tools on | Empty | settings.mcpServers |
| Add server | Configure a new MCP server connection. Transport types: SSE, streamable-http | n/a | ManageMcpServerTool.ts:7,51, McpTab.ts:372 |
| Allow local network addresses (per server) | Permit this server to connect to localhost or RFC 1918 private network addresses. Off by default. With this off, saving rejects loopback or private-network URLs with a Notice | Off | mcpServers.<name>.allowLocalUrls, src/ui/settings/McpTab.ts |
| Test server | Verify connectivity to a configured server | n/a | McpTab.ts |
Transport limitation
Vault Operator runs inside Electron (Obsidian's runtime), so only SSE and streamable-http transports are supported. Stdio-based MCP servers do not work. To bridge a stdio-only server (for example Playwright MCP), run it locally with an HTTP wrapper such as npx @playwright/mcp@latest --port 3001.
Recipes
Settings > Vault Operator > Customize > Recipes
Vetted calls to external CLI programs (for example Pandoc export) the agent may run through the execute_recipe tool with fixed, validated parameters. The tab also explains how recipes differ from rules (standing instructions in every conversation) and workflows (step-by-step instructions inserted per message with a /command).
| Setting | What it does | Default | Source |
|---|---|---|---|
| Enable recipes | Master switch for the execute_recipe tool; your choice persists across reloads | On | settings.recipes.enabled |
| Per-recipe toggles | Enable or disable each built-in recipe individually | All on | settings.recipes.recipeToggles |
Vault group
Vault
Settings > Vault Operator > Vault > Vault
Vault-level settings, including the agent folder, default output folder, and checkpoint behaviour.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Agent folder | Vault-relative folder where Vault Operator keeps plugin skills, vault-dna snapshot, externalised tmp results, cache, and the local knowledge database | .vault-operator | DEFAULT_AGENT_FOLDER (agentFolder.ts:38) |
| Pick folder | Fuzzy-picker to choose an existing folder. Type a new path to create on next use | n/a | VaultTab.ts |
| Default output folder | Where the agent writes new notes (including ingest source notes) | Inbox/ | settings.defaultOutputFolder |
| Show health badge | Stethoscope icon in the sidebar changes colour when findings exist | On | AgentSidebarView.ts:287-298 |
| Silence with-context orphans | Hide orphan findings whose only signal is a property-only edge | On | settings.vaultHealth.silenceWithContextOrphans |
| Task extraction | Detect and extract tasks from agent responses into the task folder | On | settings.taskExtraction.enabled |
| Task folder | Vault folder where extracted task notes are created | Tasks | settings.taskExtraction.taskFolder |
| Enable checkpoints | Create snapshots before file modifications | On | settings.enableCheckpoints |
| Snapshot timeout (s) | Maximum seconds to wait for a checkpoint snapshot to complete | 30 | settings.checkpointTimeoutSeconds |
| Auto-cleanup | Automatically remove old checkpoints | On | settings.checkpointAutoCleanup |
Agent folder layout
The agent folder contains data/ (skills, logs, telemetry, knowledge.db), cache/ (backups, checkpoints, externalised tmp), and assets/ (optional assets like the reranker model). Existing files are not auto-migrated when you change the path. The legacy name .obsidian-agent is still accepted for back-compat (upgraded in v2.13).
Backup
Settings > Vault Operator > Advanced > Data & diagnostics
Export and import your Vault Operator configuration. Useful when moving to a new device, sharing settings with a team, or restoring after a bad change.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Export categories | Checkboxes for each settings category (providers, rules, skills, workflows, prompts, agents, soul, memory) | All on | BackupTab.ts |
| Export | Bundle the selected categories into a JSON file | n/a | BackupTab.ts |
| Select file (Import) | Pick a previously exported JSON file | n/a | BackupTab.ts |
| Import categories | Pick which categories from the file to import | All on | BackupTab.ts |
| Confirm import | Apply the imported settings. Existing settings in the selected categories are overwritten | n/a | BackupTab.ts |
Import legacy soul.md | Read memory/soul.md and add each bullet under Identity / Values / Anti-Patterns / Communication into the soul store. Idempotent | n/a | One-off migration from older plugin versions |
API keys travel with the export
A full export includes provider API keys. Treat the JSON file like a password vault: never commit it, never share it publicly. Uncheck Providers before sharing if you want to keep keys private.
Advanced group
Interface
Settings > Vault Operator > Advanced > Interface
Appearance, input behaviour, and first-run setup. Inline editor AI actions live in a separate sub-tab (see Inline AI below).
| Setting | What it does | Default | Source |
|---|---|---|---|
| Auto-add active file | Include the currently open note as context | On | settings.autoAddActiveFileContext |
| Send with enter | Enter sends the message. Off means Ctrl/Cmd+Enter sends | On | settings.sendWithEnter |
| Include current time | Add the exact time of day to the system prompt. The calendar date is always included; the exact time changes every call and defeats prompt caching | Off | settings.includeCurrentTimeInContext |
| Chat linking | Link chat sessions to notes for traceability (frontmatter stamping plus semantic titling) | On | settings.chatLinking.enabled |
| Excluded paths | Folders that never receive a chat reference. Accepts a folder path or a /regex/ entry. Excluded notes stay fully editable, they are only left unstamped. | empty | settings.chatLinking.excludedPaths |
| Restart setup | Re-runs the first-run wizard. Under the Setup section | n/a | InterfaceTab.ts:42 |
Plugin API
Settings > Vault Operator > Advanced > Plugin API
The call_plugin_api tool surface. The read/write catalog and the auto-promotion consent toggle live next to the auto-approve toggles under Agents > Auto-approve; this tab keeps the existence gate and the revocation list.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Plugin API | Allow the agent to call JavaScript APIs on other plugins (Dataview, Omnisearch, etc.) | On | settings.pluginApi.enabled |
| User safe-marked methods | Runtime-promoted methods that auto-approve as reads; remove entries here to revoke | Empty | settings.pluginApi.safeMethodOverrides |
Data & diagnostics: Operation log
Settings > Vault Operator > Advanced > Data & diagnostics
Daily audit trail of every tool call. Each tool invocation is appended to a JSONL log file with timestamp, tool name, arguments, result status, and approval decision.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Date selector | Pick which day's log to load | Today | LogTab.ts |
| Load | Render the selected day's log as a table | n/a | LogTab.ts |
| Download | Save the raw JSONL log for the selected day | n/a | LogTab.ts |
| Clear all | Delete every log file from disk (asks for confirmation) | n/a | LogTab.ts |
Where logs live
Logs are stored at <vault>/.vault-operator/data/logs/<YYYY-MM-DD>.jsonl (one file per day). Retention is 30 days. Logs do not contain conversation content, only tool calls.
Audit log write failures banner (v3.0.0)
When the OperationLogger fails to persist a log entry, the Log tab shows a persistent banner titled "Audit log write failures" at the top of the tab. It reports the count of failed writes since plugin start (operations continued, but the audit trail has a gap) and prints the last error message, truncated to 200 characters. A "Clear notice" button resets the logger's failure state and hides the banner until the next failure. Source: src/ui/settings/LogTab.ts.
Data & diagnostics: Debug
Settings > Vault Operator > Advanced > Data & diagnostics
Internal diagnostics. Shares the Data & diagnostics tab with backup and the operation log.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Debug mode | Enable verbose logging to the developer console | Off | settings.debugMode |
Inspecting the running state
Use the inspect_self tool from chat ("inspect your tools" or "show me your current settings") to see live introspection of the running plugin. It returns a Markdown summary of the actual runtime state.
Optional assets
Settings > Vault Operator > Advanced > Optional assets
One-time downloads stored under .vault-operator/assets/. Install only what you need.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Reranker model | Xenova/ms-marco-MiniLM-L-6-v2 cross-encoder for semantic re-ranking | Not installed | OptionalAssetManager |
| Self-development source | One-time download (~5 MB) of the plugin's TypeScript source. Required for the manage_source tool, so the agent can answer "how does feature X work?" questions and propose patches. Downloaded from the plugin's GitHub release, verified by SHA256 | Not installed | OptionalAssetManager |
| Office assets | Bundled fonts and theme assets used by create_pptx, create_docx, create_xlsx | Not installed | OptionalAssetManager |
Inline chat
Settings > Vault Operator > Agents > Inline chat
Chat with the agent directly in the editor. It runs the same agent loop as the sidebar chat and inherits all its settings; only the inline-specific triggers and quick-action behaviour are configured here. Open it via the inline-AI command (bind a hotkey in Obsidian's hotkey settings, for example Cmd+K). See the inline chat guide for the workflow.
| Setting | What it does | Default | Source |
|---|---|---|---|
| Inline editor AI actions enabled | Master toggle for the inline menu, hotkey, and command-palette entry | On | settings.inlineActions.enabled |
| Auto-open floating menu on selection | When on, the inline AI affordance appears automatically after you finish selecting text. When off, only the hotkey or command palette opens it | Off | settings.inlineActions.floatingMenuEnabled |
| Use vault knowledge in lookup | Augment the lookup action with semantic-search hits from your vault | On | settings.inlineActions.vaultRagInLookup |
| Vault knowledge confidence threshold | Cosine similarity slider, range 0 to 1. Lookup falls back to LLM-only when no vault hit meets this threshold | 0.7 | settings.inlineActions.vaultRagConfidenceThreshold |
| Inline chat display | How the inline chat renders: block widget in the editor (source and live preview) or a popover overlay that also works in reading view | Block widget | settings.inlineActions.inlineChatDisplay |
All inlineActions fields are optional in data.json; missing fields resolve to the defaults above at load time, and the confidence threshold is clamped to the range 0 to 1 (see src/core/inline/inlineSettings.ts).
Plugin reload may be needed
Some changes (action registration in particular) only take effect after reloading the plugin. The tab includes a footer note to this effect.
Auto-approve still gates write tools
The Auto-approve groups still gate the write tools the inline panel invokes. Rewrite and Translate route the proposed change through the EditReviewModal (Änderungen prüfen) instead of the auto-approve pipeline, so you confirm the edit even when Edit auto-approve is on. See the safety and control guide for how the approval groups interact with inline actions.
Language
Settings > Vault Operator > Advanced > Language
Set the agent's response language. The setting follows Obsidian's language by default. UI strings (settings labels, modals, errors) follow the Obsidian language separately.
Help group
Settings > Vault Operator > Help
The Help group is not a content tab. It opens the public documentation in a new window. Use the in-app Restart setup button under Interface to re-run the first-run wizard.