From 773cb610e68a9c40f67c0ba877e61151482b26de Mon Sep 17 00:00:00 2001
From: Travis Van Nimwegen <github@travis.engineer>
Date: Thu, 20 Mar 2025 17:22:57 -0400
Subject: [PATCH] Enhance configuration options

- Introduced new configuration options for vision and notes LLM prompts
- Added category detection feature toggle in ConfigModal
- Updated ScreenshotProcessor to utilize new prompt configurations
- Cleaned up HelpModal by removing outdated forum links
---
 manifest.json                                 |   2 +-
 package.json                                  |   2 +-
 src/components/MainViewHeader.tsx             |  12 +-
 src/components/modals/ConfigModal.tsx         | 103 +++++++++++++++++-
 src/components/modals/HelpModal.tsx           |  12 --
 ...TestConfigModal.tsx => TestSetupModal.tsx} |  10 +-
 src/services/llm-service.ts                   |  19 ++++
 src/services/screenshot-processor.ts          |  25 +++--
 src/types/config-types.ts                     |  13 ++-
 versions.json                                 |   3 +-
 10 files changed, 163 insertions(+), 38 deletions(-)
 rename src/components/modals/{TestConfigModal.tsx => TestSetupModal.tsx} (95%)

diff --git a/manifest.json b/manifest.json
index 5027d5a..724ef8d 100644
--- a/manifest.json
+++ b/manifest.json
@@ -1,7 +1,7 @@
 {
 	"id": "vision-recall",
 	"name": "Vision Recall",
-	"version": "1.1.0",
+	"version": "1.2.0",
 	"minAppVersion": "1.8.3",
 	"description": "Transform screenshots into searchable notes using AI vision and text analysis.",
 	"author": "Travis Van Nimwegen",
diff --git a/package.json b/package.json
index b216938..3738647 100644
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
   "name": "obsidian-vision-recall",
-  "version": "1.1.0",
+  "version": "1.2.0",
   "description": "Transform screenshots into searchable Obsidian notes using AI vision and text analysis.",
   "main": "main.js",
   "scripts": {
diff --git a/src/components/MainViewHeader.tsx b/src/components/MainViewHeader.tsx
index 2bdcdfd..ca5c372 100644
--- a/src/components/MainViewHeader.tsx
+++ b/src/components/MainViewHeader.tsx
@@ -7,7 +7,7 @@ import { DocViewerModal } from '@/components/modals/DocViewerModal';
 import { ConfigModal } from '@/components/modals/ConfigModal';
 import { cn } from '@/lib/utils';
 import { useDataContext } from '@/data/DataContext';
-import { TestConfigModal } from './modals/TestConfigModal';
+import { TestSetupModal } from './modals/TestSetupModal';
 import { ProcessingQueueModal } from './modals/ProcessingQueueModal';
 import { useQueueStore } from '@/stores/queueStore';
 import { DebugOperationsModal } from './modals/DebugOperationsModal';
@@ -166,10 +166,10 @@ export const MainViewHeader = ({ metadata, refreshMetadata, viewMode = 'list', o
           </button>
 
           <button
-            aria-label='Test config'
+            aria-label='Test setup'
             className='cursor-pointer flex flex-row items-center gap-2'
             onClick={() => {
-              new TestConfigModal(
+              new TestSetupModal(
                 app,
                 plugin,
                 dataManager
@@ -178,7 +178,7 @@ export const MainViewHeader = ({ metadata, refreshMetadata, viewMode = 'list', o
           >
             <Bug className='w-4 h-4' />
             <div className='hidden @3xl/subheader:block'>
-              Test config
+              Test setup
             </div>
           </button>
 
@@ -264,7 +264,7 @@ export const MainViewHeader = ({ metadata, refreshMetadata, viewMode = 'list', o
         </div>
 
         <button
-          aria-label='Advanced settings'
+          aria-label='Advanced configuration'
           className='cursor-pointer flex flex-row items-center gap-2'
           onClick={() => {
             new ConfigModal(app, plugin, plugin.dataManager).open();
@@ -272,7 +272,7 @@ export const MainViewHeader = ({ metadata, refreshMetadata, viewMode = 'list', o
         >
           <Settings2 className='w-4 h-4' />
           <div className='hidden @3xl/subheader:block'>
-            Settings
+            Config
           </div>
         </button>
       </div>
diff --git a/src/components/modals/ConfigModal.tsx b/src/components/modals/ConfigModal.tsx
index d766d91..452cd12 100644
--- a/src/components/modals/ConfigModal.tsx
+++ b/src/components/modals/ConfigModal.tsx
@@ -1,9 +1,10 @@
-import React, { useEffect, useState } from 'react';
+import React, { useEffect, useRef, useState } from 'react';
 import { Modal } from 'obsidian';
 import { DataManager } from '@/data/DataManager';
 import { Plugin } from 'obsidian';
 import { Config, DefaultConfig } from '@/types/config-types';
 import { createRoot } from 'react-dom/client';
+import { cn } from '@/lib/utils';
 
 const ConfigModalContent = ({
   dataManager,
@@ -16,11 +17,17 @@ const ConfigModalContent = ({
   const [config, setConfig] = useState<Config>({});
   const [dirty, setDirty] = useState(false);
 
+  const [enableCategoryDetection, setEnableCategoryDetection] = useState(false);
+
+  const promptTextareaRef = useRef<HTMLTextAreaElement>(null);
+  const notesPromptTextareaRef = useRef<HTMLTextAreaElement>(null);
+
   useEffect(() => {
     if (initialized) return;
     const loadConfig = async () => {
       const currentConfig = dataManager.getConfig();
       setConfig(currentConfig);
+      setEnableCategoryDetection(currentConfig.enableCategoryDetection);
       setInitialized(true);
     };
     loadConfig();
@@ -34,9 +41,26 @@ const ConfigModalContent = ({
 
   const updateConfig = (updates: Partial<Config>) => {
     setConfig(prev => ({ ...prev, ...updates }));
+    if (updates.enableCategoryDetection !== undefined) {
+      setEnableCategoryDetection(updates.enableCategoryDetection);
+    }
     setDirty(true);
   };
 
+  const resetVisionLLMPrompt = () => {
+    if (promptTextareaRef.current) {
+      promptTextareaRef.current.value = DefaultConfig.visionLLMPrompt;
+    }
+    updateConfig({ visionLLMPrompt: DefaultConfig.visionLLMPrompt });
+  };
+
+  const resetNotesLLMPrompt = () => {
+    if (notesPromptTextareaRef.current) {
+      notesPromptTextareaRef.current.value = DefaultConfig.notesLLMPrompt;
+    }
+    updateConfig({ notesLLMPrompt: DefaultConfig.notesLLMPrompt });
+  };
+
   if (!initialized) {
     return <div>Loading...</div>;
   }
@@ -96,6 +120,83 @@ const ConfigModalContent = ({
           </label>
         </div>
 
+        <div className='config-field'>
+          <div className="flex flex-col gap-1 w-full">
+            <div className='flex flex-row gap-1 w-full items-center'>
+              <label className='text-sm font-bold flex-1'>
+                Vision LLM prompt
+              </label>
+              <button
+                aria-label="Reset vision LLM prompt to default"
+                type="button"
+                className="mod-warning cursor-pointer opacity-50 hover:opacity-100 duration-300 text-xs px-2 py-1"
+                onClick={resetVisionLLMPrompt}
+              >
+                Reset
+              </button>
+            </div>
+            <textarea
+              ref={promptTextareaRef}
+              className='w-full flex-1'
+              defaultValue={config.visionLLMPrompt || DefaultConfig.visionLLMPrompt}
+              onChange={e => updateConfig({ visionLLMPrompt: e.target.value })}
+            />
+          </div>
+        </div>
+
+
+
+
+        <div className="config-field">
+          <label
+            className="config-field-checkbox-label"
+            aria-label={`Attempts to categorize screenshots based on vision LLM response. If enabled, the notes LLM prompt will be modified to include the category.`}
+          >
+            <input
+              type="checkbox"
+              defaultChecked={config.enableCategoryDetection || DefaultConfig.enableCategoryDetection}
+              onChange={e => updateConfig({ enableCategoryDetection: e.target.checked })}
+            />
+            Enable category detection
+          </label>
+        </div>
+
+
+        <div className='config-field'>
+          <div className="flex flex-col gap-1 w-full">
+            <div className='flex flex-row gap-1 w-full items-center'>
+              <label
+                className={cn(
+                  'text-sm font-bold flex-1',
+                  enableCategoryDetection && 'opacity-30 hover:opacity-70 duration-300'
+                )}
+                aria-label={!enableCategoryDetection ? 'Notes LLM prompt' : 'Notes LLM prompt (disabled because category detection is enabled)'}
+              >
+                Notes LLM prompt
+              </label>
+              <button
+                aria-label="Reset notes LLM prompt to default"
+                type="button"
+                className="mod-warning cursor-pointer opacity-30 hover:opacity-70 duration-300 text-xs px-2 py-1"
+                onClick={resetNotesLLMPrompt}
+              >
+                Reset
+              </button>
+            </div>
+            <textarea
+              ref={notesPromptTextareaRef}
+              aria-label={!enableCategoryDetection ? 'Notes LLM prompt' : 'Notes LLM prompt (disabled because category detection is enabled)'}
+              className={cn(
+                'w-full flex-1',
+                enableCategoryDetection && 'opacity-30 hover:opacity-70 duration-300'
+              )}
+              disabled={enableCategoryDetection}
+              defaultValue={config.notesLLMPrompt || DefaultConfig.notesLLMPrompt}
+              onChange={e => updateConfig({ notesLLMPrompt: e.target.value })}
+            />
+          </div>
+        </div>
+
 
         {/* <div className="config-field">
           <label className="config-field-checkbox-label">
diff --git a/src/components/modals/HelpModal.tsx b/src/components/modals/HelpModal.tsx
index dfa1a26..7446ea9 100644
--- a/src/components/modals/HelpModal.tsx
+++ b/src/components/modals/HelpModal.tsx
@@ -41,18 +41,6 @@ const HelpView: React.FC<HelpModalProps> = ({ plugin, onClose }) => {
         >
           Join the new Discord server <ExternalLink className='w-3 h-3' />
         </a>
-
-        <div className='text-text-muted'>or</div>
-
-        <a
-          href='https://forum.visionrecall.com'
-          target='_blank'
-          rel='noopener noreferrer'
-          className='cursor-pointer inline-flex flex-row items-center gap-1'
-          aria-label={'Open the forum'}
-        >
-          Visit & chat on the forum <ExternalLink className='w-3 h-3' />
-        </a>
       </div>
 
       <div className="flex flex-row justify-end items-center w-full">
diff --git a/src/components/modals/TestConfigModal.tsx b/src/components/modals/TestSetupModal.tsx
similarity index 95%
rename from src/components/modals/TestConfigModal.tsx
rename to src/components/modals/TestSetupModal.tsx
index 19cfae9..d0e2a5f 100644
--- a/src/components/modals/TestConfigModal.tsx
+++ b/src/components/modals/TestSetupModal.tsx
@@ -6,13 +6,13 @@ import { DateTime } from 'luxon';
 import { DataManager } from '@/data/DataManager';
 import { getModels } from '@/services/llm-service';
 
-interface TestConfigModalProps {
+interface TestSetupModalProps {
   dataManager: DataManager;
   plugin: VisionRecallPlugin;
   onClose: () => void;
 }
 
-const TestConfigView: React.FC<TestConfigModalProps> = ({ dataManager, plugin, onClose }) => {
+const TestSetupView: React.FC<TestSetupModalProps> = ({ dataManager, plugin, onClose }) => {
   const settings = plugin.settings;
 
   const [models, setModels] = useState<string[]>([]);
@@ -127,7 +127,7 @@ const TestConfigView: React.FC<TestConfigModalProps> = ({ dataManager, plugin, o
   );
 };
 
-export class TestConfigModal extends Modal {
+export class TestSetupModal extends Modal {
   private metadata: any;
   private plugin: VisionRecallPlugin;
 
@@ -139,11 +139,11 @@ export class TestConfigModal extends Modal {
 
   onOpen() {
     const { contentEl, titleEl } = this;
-    titleEl.setText('Test config');
+    titleEl.setText('Test your setup');
     const root = createRoot(contentEl);
 
     root.render(
-      <TestConfigView
+      <TestSetupView
         dataManager={this.plugin.dataManager}
         plugin={this.plugin}
         onClose={() => this.close()}
diff --git a/src/services/llm-service.ts b/src/services/llm-service.ts
index d9e8d9c..f27ddf7 100644
--- a/src/services/llm-service.ts
+++ b/src/services/llm-service.ts
@@ -1,11 +1,30 @@
 import { OPENROUTER_HEADERS } from '@/constants';
 import { getLanguagePromptModifierIfIndicated } from '@/lib/languages';
 import { tagsJsonSchema, TagsSchema } from '@/lib/tag-utils';
+import { Config } from '@/types/config-types';
 import { VisionRecallPluginSettings } from '@/types/settings-types';
 import { requestUrl, RequestUrlParam, RequestUrlResponse } from 'obsidian';
 
 export const VISION_LLM_PROMPT = "Analyze this screenshot and describe its content and identify the type of screenshot if possible.";
 
+export const NOTES_LLM_PROMPT = "The following OCR text and vision analysis are from a screenshot. Summarize and synthesize the text and vision analysis and identify key information.";
+
+export const getVisionLLMPrompt = (config: Config) => {
+  if (config.visionLLMPrompt && config.visionLLMPrompt.trim() !== '') {
+    return config.visionLLMPrompt;
+  }
+
+  return VISION_LLM_PROMPT;
+}
+
+export const getNotesLLMPrompt = (config: Config) => {
+  if (config.notesLLMPrompt && config.notesLLMPrompt.trim() !== '') {
+    return config.notesLLMPrompt;
+  }
+
+  return NOTES_LLM_PROMPT;
+}
+
 export function adjustEndpoint(apiEndpointUrl: string, shouldRemove: boolean): string {
   const suffix = '/v1'
 
diff --git a/src/services/screenshot-processor.ts b/src/services/screenshot-processor.ts
index 0be8f5a..926d3af 100644
--- a/src/services/screenshot-processor.ts
+++ b/src/services/screenshot-processor.ts
@@ -1,7 +1,7 @@
 import { App, FileManager, Notice, TFile, TFolder, arrayBufferToBase64, normalizePath } from 'obsidian';
 import Tesseract, { createWorker, Worker } from 'tesseract.js';
 import { VisionRecallPluginSettings } from '@/types/settings-types';
-import { DEFAULT_TAGS_AND_TITLE, TagsAndTitle, VISION_LLM_PROMPT, callLLMAPI } from '@/services/llm-service';
+import { DEFAULT_TAGS_AND_TITLE, TagsAndTitle, VISION_LLM_PROMPT, callLLMAPI, getNotesLLMPrompt, getVisionLLMPrompt } from '@/services/llm-service';
 import VisionRecallPlugin from '@/main';
 import { checkOCRText } from '@/lib/ocr-validation';
 import { formatTags, sanitizeObsidianTag, tagsToCommaString } from '@/lib/tag-utils';
@@ -171,7 +171,7 @@ export class ScreenshotProcessor {
     if (!ocrText || this.progressManager.isStoppedByUser()) return null;
 
     ocrText = cleanOCRResultLanguageSpecific(ocrText, currentLanguageSetting || 'eng') || '';
-    console.log('ocrText (after cleaning)', ocrText);
+    this.plugin.logger.debug('ocrText (after cleaning)', ocrText);
     if (!ocrText) return null;
 
     let validOCRText = await checkOCRText(ocrText, currentLanguageSetting || 'eng');
@@ -363,7 +363,7 @@ export class ScreenshotProcessor {
         languagePromptModifier = getLanguagePromptModifierIfIndicated(this.settings, true);
       }
 
-      let textPayload = VISION_LLM_PROMPT + languagePromptModifier;
+      let textPayload = getVisionLLMPrompt(this.plugin.dataManager.getConfig()) + languagePromptModifier;
       this.plugin.logger.debug('Vision LLM prompt:', textPayload);
 
       const visionPayload = {
@@ -399,19 +399,23 @@ export class ScreenshotProcessor {
 
   private async generateNotes(ocrText: string, visionLLMResponse: string, settings: VisionRecallPluginSettings): Promise<string | null> {
     try {
-
+      const config = this.plugin.dataManager.getConfig();
       const languagePromptModifier = getLanguagePromptModifierIfIndicated(settings, true);
 
-      let endpointPrompt = `The following text is from a screenshot. Summarize the text and identify key information.`;
+      let endpointPrompt = getNotesLLMPrompt(config);
 
-      const visionLLMCategories = Object.keys(visionLLMResponseCategoriesMap);
+      if (config.enableCategoryDetection) {
+        const visionLLMCategories = Object.keys(visionLLMResponseCategoriesMap);
 
-      const visionLLMCategory = visionLLMCategories.find(category => visionLLMResponse.toLowerCase().includes(category));
+        const visionLLMCategory = visionLLMCategories.find(category => visionLLMResponse.toLowerCase().includes(category));
 
-      if (visionLLMCategory) {
-        endpointPrompt = visionLLMResponseCategoriesMap[visionLLMCategory];
+        if (visionLLMCategory) {
+          endpointPrompt = visionLLMResponseCategoriesMap[visionLLMCategory];
+        }
       }
 
+      this.plugin.logger.debug('Generate Notes LLM prompt (without added language modifier, OCR text, or vision analysis that are added later):', endpointPrompt);
+
       endpointPrompt += `${languagePromptModifier}\n\nOCR text:\n${ocrText}\n\nVision analysis:\n${visionLLMResponse}`
 
       const endpointPayload = {
@@ -430,8 +434,8 @@ export class ScreenshotProcessor {
       );
 
       this.plugin.logger.debug('Generated notes:', generatedNotes ? generatedNotes.substring(0, 300) + '...' : 'API call failed or no notes generated');
-      return generatedNotes;
 
+      return generatedNotes;
     } catch (error) {
       this.plugin.logger.error('Endpoint LLM API error:', error);
       new Notice('Endpoint LLM API call failed. See console for details.');
@@ -502,6 +506,7 @@ export class ScreenshotProcessor {
 
       let counter = 1;
       while (this.app.vault.getAbstractFileByPath(notePath) && counter < 100) {
+        this.plugin.logger.debug(`Note title already exists: ${noteTitle}. Incrementing counter and trying again.`);
         noteTitle = `${noteTitleBase} Notes (${counter}).md`;
         notePath = normalizePath(`${outputNotesFolder}/${noteTitle}`);
         counter++;
diff --git a/src/types/config-types.ts b/src/types/config-types.ts
index a91bf9b..01f3a1b 100644
--- a/src/types/config-types.ts
+++ b/src/types/config-types.ts
@@ -1,3 +1,5 @@
+import { VISION_LLM_PROMPT, NOTES_LLM_PROMPT } from '@/services/llm-service';
+
 export interface Config {
   enableAutoIntakeFolderProcessing?: boolean;
 
@@ -10,6 +12,12 @@ export interface Config {
 
   experimentalFeatures?: string[];
 
+  visionLLMPrompt?: string;
+
+  enableCategoryDetection?: boolean; // this is rudimentarily implemented, so defaulting to false
+
+  notesLLMPrompt?: string;
+
   [key: string]: unknown;
 }
 
@@ -18,5 +26,8 @@ export const DefaultConfig: Config = {
   enablePeriodicIntakeFolderProcessing: false,
   intakeFolderPollingInterval: 300, // 300 seconds, or 5 minutes
   defaultMinimizedProgressDisplay: false, // Currently not functional
-  experimentalFeatures: []
+  experimentalFeatures: [],
+  visionLLMPrompt: VISION_LLM_PROMPT,
+  enableCategoryDetection: false,
+  notesLLMPrompt: NOTES_LLM_PROMPT,
 }; 
\ No newline at end of file
diff --git a/versions.json b/versions.json
index 923d4b8..7fc4684 100644
--- a/versions.json
+++ b/versions.json
@@ -14,5 +14,6 @@
 	"1.0.12": "1.8.3",
 	"1.0.13": "1.8.3",
 	"1.0.14": "1.8.3",
-	"1.1.0": "1.8.3"
+	"1.1.0": "1.8.3",
+	"1.2.0": "1.8.3"
 }
\ No newline at end of file