Configuration Reference
Every setting in SANE, explained. Access them at Settings → SANE inside Obsidian.
AI Provider
Choose which service handles language generation and embedding computation. These can be different providers.
aiProvider
default: openai
The provider used for LLM calls (tags, keywords, links, summary). Options:
openai, google, grok, azure, local.embeddingProvider
default: openai
The provider used for semantic embeddings. Can differ from
aiProvider. Options: openai, google, local. Falls back to a hash-based similarity when no key is set.azureEndpoint
default: ""
Required when
aiProvider is azure. Full URL of your Azure OpenAI deployment endpoint.localEndpoint
default: http://localhost:11434/v1
Base URL for a local LLM server (Ollama, vLLM, llama.cpp). Used when
aiProvider is local.embeddingLocalEndpoint
default: http://localhost:11434/v1
Base URL for a local embedding server. Can be the same as
localEndpoint or a separate server.Models
SANE picks sensible defaults per provider. Override them here if you want a specific model version.
llmModel
default: gpt-4o-mini (OpenAI)
The model used for generating tags, keywords, links, and summaries. Examples:
gpt-4o, gemini-1.5-flash, grok-beta, llama3.embeddingModel
default: text-embedding-3-small (OpenAI)
The model used to create semantic embeddings for similarity search. Examples:
text-embedding-3-large, embedding-001 (Google), nomic-embed-text (Ollama).Processing
Control when and how often SANE processes notes. See Trigger modes for a full comparison.
processingTrigger
default: delayed
When to process a note after it changes.
immediate — on every save; delayed — after a quiet period; scheduled — daily at a set hour; manual — only via command.delayMinutes
default: 10
Minutes of inactivity before a note is processed. Only applies when
processingTrigger is delayed.scheduleHour
default: 2 (2 AM)
Hour (0–23) when the nightly batch run fires. Only applies when
processingTrigger is scheduled.relevantNotesCount
default: 3
How many of the most semantically similar notes to surface as context when generating tags, links, and summaries.
Security & Scope
targetFolder
default: "" (all folders)
Restrict processing to this folder path relative to your vault root. Example:
Notes/Work. Leave empty to process all notes.requireBackupWarning
default: true
Show a confirmation dialog before running bulk operations. Recommended to leave on until you're comfortable with SANE's output.
Cost Management
SANE tracks every API call and pauses automatically when your budget is exceeded.
dailyBudget
default: 1.00 (USD)
Maximum dollars to spend on API calls per calendar day. SANE will stop processing and notify you when this is reached. Set to
0 to disable the cap.costTracking
default: true
Whether to record and display cost data. Disabling removes the budget cap. Run SANE: Show cost summary to see today's and this month's totals.
Feature Toggles
Enable or disable each type of frontmatter field SANE generates independently.
enableTags
default: true
Generate
sane_tags — 3 to 7 lowercase, hyphenated topic tags.enableKeywords
default: true
Generate
sane_keywords — 3 to 8 key phrases extracted from the note's content.enableLinks
default: true
Generate
sane_links — wiki-style links to the most relevant notes in your vault.enableSummary
default: true
Generate
sane_summary — a 1–2 sentence summary, capped at roughly 50 words.enableCreationTimestamp
default: true
Write
sane_created_at the first time SANE processes a note.enableModificationTimestamp
default: true
Update
sane_modified_at each time SANE reprocesses a note.Advanced
maxTokens
default: 2000
Token limit for LLM responses. Lower values reduce cost but may truncate longer summaries or tag lists.
temperature
default: 0.3
LLM sampling temperature. Lower values (closer to 0) produce more consistent, deterministic output. Raise towards 1 for more varied suggestions.
debugMode
default: false
Enables extra logging in the developer console and unlocks the SANE: Test AI response command. Use when troubleshooting a provider connection.