allenhutchison
diff --git a/‎README.md‎
Lines changed: 1 addition & 1 deletion b/‎README.md‎
Lines changed: 1 addition & 1 deletion
diff --git a/‎docs/guide/faq.md‎
Lines changed: 2 additions & 2 deletions b/‎docs/guide/faq.md‎
Lines changed: 2 additions & 2 deletions
diff --git a/‎docs/reference/settings.md‎
Lines changed: 29 additions & 6 deletions b/‎docs/reference/settings.md‎
Lines changed: 29 additions & 6 deletions
diff --git a/‎src/main.ts‎
Lines changed: 4 additions & 0 deletions b/‎src/main.ts‎
Lines changed: 4 additions & 0 deletions
diff --git a/‎src/ui/settings-api.ts‎ ‎src/ui/settings-agent-config.ts‎src/ui/settings-api.ts renamed to src/ui/settings-agent-config.ts
Lines changed: 101 additions & 43 deletions b/‎src/ui/settings-api.ts‎ ‎src/ui/settings-agent-config.ts‎src/ui/settings-api.ts renamed to src/ui/settings-agent-config.ts
Lines changed: 101 additions & 43 deletions
@@ -30,7 +30,7 @@ Gemini Scribe is an Obsidian plugin that integrates Google's Gemini AI models, p
 ## Features
 
 - **Agent Mode with Tool Calling:** An AI agent that can actively work with your vault! It can search for files, read content, create new notes, edit existing ones, move and rename files, create folders, and even conduct deep research with proper citations. Features persistent sessions, granular permission controls, session-specific model configuration, and a diff review view that lets you inspect and edit proposed file changes before they're written.
-- **Semantic Vault Search:** [Experimental] Search your vault by meaning, not just keywords. Uses Google's File Search API to index your notes in the background. The AI can find relevant content even when you don't remember exact words. Supports PDFs and attachments, with pause/resume controls and detailed status tracking.
+- **Semantic Vault Search:** Search your vault by meaning, not just keywords. Uses Google's File Search API to index your notes in the background. The AI can find relevant content even when you don't remember exact words. Supports PDFs and attachments, with pause/resume controls and detailed status tracking.
 - **Context-Aware Agent:** Add specific notes as persistent context for your agent sessions. The agent can access and reference these context files throughout your conversation, providing highly relevant and personalized responses.
 - **Smart Summarization:** Quickly generate concise, one-sentence summaries of your notes and automatically store them in the document's frontmatter, using a dedicated Gemini model optimized for summarization.
 - **Selection-Based AI Features:** Work with selected text in powerful ways:
 
@@ -171,11 +171,11 @@ Make sure you're on v4.0 or later — earlier versions had a bug where context f
 
 This command requires: (1) a markdown file actively open in the editor, and (2) context sending to be enabled. If no file is open, you'll see "Failed to get file content for summary." ([#134](https://github.com/allenhutchison/obsidian-gemini/issues/134))
 
-## Semantic Vault Search (Experimental)
+## Semantic Vault Search
 
 ### What does the Vault Index feature do? Is my data private?
 
-The vault index uses Google's File Search API to enable semantic (meaning-based) search of your vault. Files are stored in an index private to your GCP project, tied to your API key. Your data is not shared or used for model training. The feature is experimental and located under Advanced Settings. ([#297](https://github.com/allenhutchison/obsidian-gemini/discussions/297))
+The vault index uses Google's File Search API to enable semantic (meaning-based) search of your vault. Files are stored in an index private to your GCP project, tied to your API key. Your data is not shared or used for model training. The feature is configured under the **Vault Search Index** section in the plugin settings. ([#297](https://github.com/allenhutchison/obsidian-gemini/discussions/297))
 
 ## Plugin Conflicts
 
 
@@ -2,16 +2,31 @@
 
 This document provides a comprehensive reference for all Obsidian Gemini Scribe settings.
 
+The settings tab is organised into a permanently-open **General** section at the top — covering provider, API key, models, and the plugin state folder — followed by collapsible sections (▶ collapsed, ▼ expanded). Click any header to toggle it; expand/collapse state is remembered between sessions in the `expandedSettingsSections` setting. All collapsible sections start collapsed.
+
+The order of sections is:
+
+1. **General** (always open) — provider, API key, models, plugin state folder, Show Advanced Settings toggle.
+2. **User Experience** — your name, frontmatter key, streaming, diff view, session history toggle, tool execution logging.
+3. **Automation** — scheduled tasks, scheduler catch-up, and lifecycle hooks combined.
+4. **Vault Search Index** — semantic search over your vault using Google File Search.
+
+Advanced sections — Tool Permissions, MCP Servers, Agent Config, Debug — are tagged with an **ADVANCED** pill and only appear after toggling **Show Advanced Settings** at the bottom of General. **Agent Config** bundles four related sub-areas (Custom Prompts, API Configuration, Context Management, Tool Loop Detection) under one collapsible since they all tune how the agent talks to the model.
+
 ## Table of Contents
 
-- [Basic Settings](#basic-settings)
-- [Model Configuration](#model-configuration)
-- [Custom Prompts](#custom-prompts)
-- [UI Settings](#ui-settings)
-- [Context Management](#context-management)
-- [Developer Settings](#developer-settings)
+The reference below groups settings by topic for lookup, which doesn't always map 1:1 to the UI section names. The annotation in parentheses tells you which UI section a topic appears under.
+
+- [Basic Settings](#basic-settings) (UI: _General_ — provider, API key, models, plugin state folder)
+- [Model Configuration](#model-configuration) (UI: _General_ — chat/summary/completion/image model selection)
+- [Custom Prompts](#custom-prompts) (UI: _Agent Config_ — advanced)
+- [UI Settings](#ui-settings) (UI: _User Experience_ — streaming, diff view, identity, frontmatter key, session history)
+- [Context Management](#context-management) (UI: _Agent Config_ — advanced)
+- [Developer Settings](#developer-settings) (UI: split across _Agent Config_, _Tool Permissions_, _MCP Servers_, _Debug_)
 - [Session-Level Settings](#session-level-settings)
 
+UI sections without a dedicated topic in this reference: _Automation_ (covered in the [Scheduled Tasks](/guide/scheduled-tasks) and [Lifecycle Hooks](/guide/lifecycle-hooks) guides) and _Vault Search Index_ (covered in [Semantic Search](/guide/semantic-search)).
+
 ## Basic Settings
 
 ### Provider
@@ -173,6 +188,14 @@ See the [Custom Prompts Guide](/guide/custom-prompts) for detailed instructions.
 - **Why opt-in**: Vault events fire continuously; an unintentionally-broad hook can drain API quota quickly. The default is off so users opt in deliberately.
 - **See also**: [Lifecycle Hooks](/guide/lifecycle-hooks)
 
+### Expanded Settings Sections
+
+- **Setting**: `expandedSettingsSections`
+- **Type**: `string[]`
+- **Default**: `[]`
+- **Description**: Internal list of section ids that are currently expanded in the settings tab. Updated automatically when you toggle a section. Known ids: `ui`, `automation`, `rag`, `tool-permissions`, `mcp-servers`, `agent-config`, `debug`. (General is always open; it has no id and ignores this setting.)
+- **Note**: Edit `data.json` directly to pre-expand sections (for example, on a new install) or restore a custom layout after migrating vaults.
+
 ## Context Management
 
 Context management automatically monitors and controls conversation size to prevent exceeding model token limits.
 
@@ -99,6 +99,8 @@ export interface ObsidianGeminiSettings {
 	autoRunCatchUp: boolean;
 	// Lifecycle hooks (opt-in: AI runs triggered by vault events)
 	hooksEnabled: boolean;
+	// IDs of collapsible settings sections currently expanded; persists across reloads.
+	expandedSettingsSections: string[];
 	// Cached remote model list (managed by ModelListProvider)
 	remoteModelCache?: { models: GeminiModel[]; timestamp: number };
 }
@@ -156,6 +158,8 @@ const DEFAULT_SETTINGS: ObsidianGeminiSettings = {
 	autoRunCatchUp: false,
 	// Lifecycle hooks default off (opt-in)
 	hooksEnabled: false,
+	// All settings sections start collapsed
+	expandedSettingsSections: [],
 };
 
 const MIGRATION_SECRET_NAME = 'gemini-scribe-api-key';
 
@@ -1,5 +1,6 @@
 import type ObsidianGemini from '../main';
 import { Setting, Notice, debounce } from 'obsidian';
+import { createCollapsibleSection } from './settings-helpers';
 import { getErrorMessage } from '../utils/error-utils';
 import type { SettingsSectionContext } from './settings';
 
@@ -8,15 +9,24 @@ let temperatureRunId = 0;
 let topPDebounceTimer: NodeJS.Timeout | null = null;
 let topPRunId = 0;
 
-export async function renderApiSettings(
+/**
+ * "Agent Config" advanced section — combines Custom Prompts, API Configuration,
+ * Context Management, and Tool Loop Detection into a single collapsible with
+ * labeled sub-groups, since they all tune how the agent talks to the model.
+ */
+export async function renderAgentConfigSettings(
 	containerEl: HTMLElement,
 	plugin: ObsidianGemini,
-	_context: SettingsSectionContext
+	context: SettingsSectionContext
 ): Promise<void> {
+	const sectionEl = createCollapsibleSection(plugin, containerEl, 'Agent Config', 'agent-config', {
+		description:
+			'Tune how the agent talks to the model: custom prompts, retry/generation parameters, conversation summarization, and loop guards.',
+		advanced: true,
+	});
+
 	// Debounce saveSettings() for text inputs so typing doesn't trigger the plugin
 	// lifecycle on every keystroke. Settings are mutated immediately; only the save is delayed.
-	// The callback is async + wrapped in try/catch so rejections from saveSettings() don't
-	// become unhandled promise rejections.
 	const debouncedSave = debounce(
 		async () => {
 			try {
@@ -30,40 +40,39 @@ export async function renderApiSettings(
 		true
 	);
 
-	// File Logging
-	new Setting(containerEl)
-		.setName('Log to file')
+	// --- Custom Prompts ---
+	new Setting(sectionEl).setName('Custom Prompts').setHeading();
+
+	new Setting(sectionEl)
+		.setName('Allow system prompt override')
 		.setDesc(
-			'Write log entries to a file in the plugin state folder. ' +
-				'Errors and warnings are always logged; debug entries require Debug Mode. ' +
-				'Log files are automatically rotated at 1 MB.'
+			'WARNING: Allows custom prompts to completely replace the system prompt. This may break expected functionality.'
 		)
 		.addToggle((toggle) =>
-			toggle.setValue(plugin.settings.fileLogging).onChange(async (value) => {
-				plugin.settings.fileLogging = value;
+			toggle.setValue(plugin.settings.allowSystemPromptOverride ?? false).onChange(async (value) => {
+				plugin.settings.allowSystemPromptOverride = value;
 				await plugin.saveSettings();
 			})
 		);
 
-	// Custom Prompts Advanced Settings
-	new Setting(containerEl).setName('Custom Prompts').setHeading();
+	// --- API Configuration ---
+	new Setting(sectionEl).setName('API Configuration').setHeading();
 
-	new Setting(containerEl)
-		.setName('Allow system prompt override')
+	new Setting(sectionEl)
+		.setName('Log to file')
 		.setDesc(
-			'WARNING: Allows custom prompts to completely replace the system prompt. This may break expected functionality.'
+			'Write log entries to a file in the plugin state folder. ' +
+				'Errors and warnings are always logged; debug entries require Debug Mode. ' +
+				'Log files are automatically rotated at 1 MB.'
 		)
 		.addToggle((toggle) =>
-			toggle.setValue(plugin.settings.allowSystemPromptOverride ?? false).onChange(async (value) => {
-				plugin.settings.allowSystemPromptOverride = value;
+			toggle.setValue(plugin.settings.fileLogging).onChange(async (value) => {
+				plugin.settings.fileLogging = value;
 				await plugin.saveSettings();
 			})
 		);
 
-	// API Configuration
-	new Setting(containerEl).setName('API Configuration').setHeading();
-
-	new Setting(containerEl)
+	new Setting(sectionEl)
 		.setName('Maximum Retries')
 		.setDesc('Maximum number of retries when a model request fails.')
 		.addText((text) =>
@@ -79,7 +88,7 @@ export async function renderApiSettings(
 				})
 		);
 
-	new Setting(containerEl)
+	new Setting(sectionEl)
 		.setName('Initial Backoff Delay (ms)')
 		.setDesc('Initial delay in milliseconds before the first retry. Subsequent retries will use exponential backoff.')
 		.addText((text) =>
@@ -95,11 +104,75 @@ export async function renderApiSettings(
 				})
 		);
 
-	// Create temperature setting with dynamic ranges
-	await createTemperatureSetting(containerEl, plugin);
+	await createTemperatureSetting(sectionEl, plugin);
+	await createTopPSetting(sectionEl, plugin);
+
+	// --- Context Management ---
+	new Setting(sectionEl).setName('Context Management').setHeading();
+
+	const thresholdSetting = new Setting(sectionEl)
+		.setName('Context compaction threshold')
+		.setDesc(
+			`Automatically summarize older conversation turns when token usage exceeds this percentage of the model context window. Current: ${plugin.settings.contextCompactionThreshold}%`
+		);
+
+	thresholdSetting.addSlider((slider) =>
+		slider
+			.setLimits(5, 50, 5)
+			.setValue(plugin.settings.contextCompactionThreshold)
+			.setDynamicTooltip()
+			.onChange(async (value) => {
+				plugin.settings.contextCompactionThreshold = value;
+				thresholdSetting.setDesc(
+					`Automatically summarize older conversation turns when token usage exceeds this percentage of the model context window. Current: ${value}%`
+				);
+				await plugin.saveSettings();
+			})
+	);
+
+	// --- Tool Loop Detection ---
+	new Setting(sectionEl).setName('Tool Loop Detection').setHeading();
+
+	new Setting(sectionEl)
+		.setName('Enable loop detection')
+		.setDesc('Prevent the AI from repeatedly calling the same tool with identical parameters.')
+		.addToggle((toggle) =>
+			toggle.setValue(plugin.settings.loopDetectionEnabled).onChange(async (value) => {
+				plugin.settings.loopDetectionEnabled = value;
+				await plugin.saveSettings();
+				context.redisplay();
+			})
+		);
+
+	if (plugin.settings.loopDetectionEnabled) {
+		new Setting(sectionEl)
+			.setName('Loop threshold')
+			.setDesc('Number of identical tool calls before considering it a loop (default: 3).')
+			.addSlider((slider) =>
+				slider
+					.setLimits(2, 10, 1)
+					.setValue(plugin.settings.loopDetectionThreshold)
+					.setDynamicTooltip()
+					.onChange(async (value) => {
+						plugin.settings.loopDetectionThreshold = value;
+						await plugin.saveSettings();
+					})
+			);
 
-	// Create topP setting with dynamic ranges
-	await createTopPSetting(containerEl, plugin);
+		new Setting(sectionEl)
+			.setName('Time window (seconds)')
+			.setDesc('Time window to check for repeated calls (default: 30 seconds).')
+			.addSlider((slider) =>
+				slider
+					.setLimits(10, 120, 5)
+					.setValue(plugin.settings.loopDetectionTimeWindowSeconds)
+					.setDynamicTooltip()
+					.onChange(async (value) => {
+						plugin.settings.loopDetectionTimeWindowSeconds = value;
+						await plugin.saveSettings();
+					})
+			);
+	}
 }
 
 async function createTemperatureSetting(containerEl: HTMLElement, plugin: ObsidianGemini): Promise<void> {
@@ -120,18 +193,14 @@ async function createTemperatureSetting(containerEl: HTMLElement, plugin: Obsidi
 				.setValue(plugin.settings.temperature)
 				.setDynamicTooltip()
 				.onChange(async (value) => {
-					// Clear previous timeout
 					if (temperatureDebounceTimer) {
 						clearTimeout(temperatureDebounceTimer);
 					}
 
-					// Set immediate value for responsive UI
 					plugin.settings.temperature = value;
 
-					// Capture the run ID upfront so we can discard stale async results.
 					const runId = ++temperatureRunId;
 
-					// Debounce validation and saving
 					temperatureDebounceTimer = setTimeout(async () => {
 						try {
 							// Validate the current value against model capabilities. Read from
@@ -189,30 +258,21 @@ async function createTopPSetting(containerEl: HTMLElement, plugin: ObsidianGemin
 				.setValue(plugin.settings.topP)
 				.setDynamicTooltip()
 				.onChange(async (value) => {
-					// Clear previous timeout
 					if (topPDebounceTimer) {
 						clearTimeout(topPDebounceTimer);
 					}
 
-					// Set immediate value for responsive UI
 					plugin.settings.topP = value;
 
-					// Capture the run ID upfront so we can discard stale async results.
 					const runId = ++topPRunId;
 
-					// Debounce validation and saving
 					topPDebounceTimer = setTimeout(async () => {
 						try {
-							// Validate the current value against model capabilities. Read from
-							// settings rather than the captured `value` so the validation always
-							// matches the most recent user input.
 							const validation = await modelManager.validateParameters(
 								plugin.settings.temperature,
 								plugin.settings.topP
 							);
 
-							// A newer slider change has superseded this run — discard the
-							// stale result instead of clobbering the current slider/value.
 							if (runId !== topPRunId) {
 								return;
 							}
@@ -227,8 +287,6 @@ async function createTopPSetting(containerEl: HTMLElement, plugin: ObsidianGemin
 
 							await plugin.saveSettings();
 						} catch (error) {
-							// If a newer run has superseded us, drop this stale failure silently —
-							// surfacing it would contradict whatever the current run is doing.
 							if (runId !== topPRunId) {
 								return;
 							}