Merge pull request #16 from bmachek/7-feature-improve-ai-photo-captio…

…ns-with-optional-context-input-or-metadata 7 feature improve ai photo captions with optional context input or metadata
bmachek · Jan 1, 2025 · 8208f7d · 8208f7d
2 parents 53ccb40 + 581ef71
commit 8208f7d
Show file tree

Hide file tree

Showing 5 changed files with 27 additions and 8 deletions.
diff --git a/lrc-ai-assistant.lrplugin/AiModelAPI.lua b/lrc-ai-assistant.lrplugin/AiModelAPI.lua
@@ -27,6 +27,6 @@ function AiModelAPI:new()
     return instance
 end
 
-function AiModelAPI:analyzeImage(filePath)
-    return self.usedApi:analyzeImage(filePath)
+function AiModelAPI:analyzeImage(filePath, metadata)
+    return self.usedApi:analyzeImage(filePath, metadata)
 end
diff --git a/lrc-ai-assistant.lrplugin/GeminiAPI.lua b/lrc-ai-assistant.lrplugin/GeminiAPI.lua
@@ -114,8 +114,18 @@ function GeminiAPI:doRequest(filePath, task, systemInstruction, generationConfig
 end
 
 
-function GeminiAPI:analyzeImage(filePath)
-    local success, result, inputTokenCount, outputTokenCount = GeminiAPI:doRequest(filePath, Defaults.defaultTask, Defaults.defaultSystemInstruction, Defaults.getDefaultGenerationConfig())
+function GeminiAPI:analyzeImage(filePath, metadata)
+    local task = Defaults.defaultTask
+    if metadata ~= nil then
+        if metadata.gps ~= nil then
+            task = task .. " " .. LOC "$$$/lrc-ai-assistant/GeminiAPI/gpsAddon=This photo was taken at the following coordinates:" .. metadata.gps.latitude .. ", " .. metadata.gps.longitude
+        end
+        if metadata.keywords ~= nil then
+            task = task .. " " .. LOC "$$$/lrc-ai-assistant/GeminiAPI/keywordAddon=Some keywords are:" .. metadata.keywords
+        end
+    end
+
+    local success, result, inputTokenCount, outputTokenCount = GeminiAPI:doRequest(filePath, task, Defaults.defaultSystemInstruction, Defaults.getDefaultGenerationConfig())
     if success and result ~= nil then
         result = string.gsub(result, Defaults.geminiKeywordsGarbageAtStart, '')
         result = string.gsub(result, Defaults.geminiKeywordsGarbageAtEnd, '')

diff --git a/lrc-ai-assistant.lrplugin/GenerateImageInfo.lua b/lrc-ai-assistant.lrplugin/GenerateImageInfo.lua
@@ -170,7 +170,7 @@ local function exportAndAnalyzePhoto(photo, progressScope)
         LR_size_units = 'pixels',
         LR_collisionHandling = 'rename',
         LR_includeVideoFiles = false,
-        LR_removeLocationMetadata = false,
+        LR_removeLocationMetadata = true,
         LR_embeddedMetadataOption = "all",
     }
 
@@ -188,8 +188,13 @@ local function exportAndAnalyzePhoto(photo, progressScope)
 
     for _, rendition in exportSession:renditions() do
         local success, path = rendition:waitForRender()
+        local metadata = {}
+
+        metadata.gps = photo:getRawMetadata("gps")
+        metadata.keywords = photo:getFormattedMetadata("keywordTagsForExport")
+
         if success then
-            local analyzeSuccess, result, inputTokens, outputTokens = ai:analyzeImage(path)
+            local analyzeSuccess, result, inputTokens, outputTokens = ai:analyzeImage(path, metadata)
 
             if result == 'RATE_LIMIT_EXHAUSTED' then
                 LrDialogs.showError(LOC "$$$/lrc-ai-assistant/GenerateImageInfo/rateLimit=Quota exhausted, set up pay as you go at Google, or wait for some hours.")

diff --git a/lrc-ai-assistant.lrplugin/TranslatedStrings_de.txt b/lrc-ai-assistant.lrplugin/TranslatedStrings_de.txt
@@ -18,6 +18,8 @@
 "$$$/lrc-ai-assistant/Defaults/defaultSystemInstruction=Du analysierst Bilder für Photo Management. Sei spezifisch und exakt. Generiere maximal 25 einzigartige Stichwörter."
 "$$$/lrc-ai-assistant/Defaults/defaultTask=Beschreibe den Bildinhalt inklusive aller erkannten Gegenstände"
 "$$$/lrc-ai-assistant/GeminiAPI/NoAPIkey=Kein Gemini API key eingetragen im Zusatzmodul-Manager."
+"$$$/lrc-ai-assistant/GeminiAPI/gpsAddon=Dieses Foto wurde an den folgenden Koordinaten aufgenommen: "
+"$$$/lrc-ai-assistant/GeminiAPI/keywordAddon=Einige Stichwörter sind:"
 "$$$/lrc-ai-assistant/GenerateImageInfo/SkipFromHere=Die folgenden ohne Überprüfung speichern."
 "$$$/lrc-ai-assistant/GenerateImageInfo/UsedTokenDialog/GeneratedCosts=Erzeugte Kosten"
 "$$$/lrc-ai-assistant/GenerateImageInfo/UsedTokenDialog/Title=Verursachte Kosten"
@@ -44,4 +46,4 @@
 "$$$/lrc-ai-assistant/PluginInfoDialogSections/keywords=Stichwörter"
 "$$$/lrc-ai-assistant/PluginInfoDialogSections/showCosts=Kosten anzeigen (Ohne jegliche Garantie!!!)"
 "$$$/lrc-ai-assistant/PluginInfoDialogSections/title=Titel"
-"$$$/lrc-ai-assistant/PluginInfoDialogSections/validateBeforeSaving=Vor dem Speichern überprüfen","
+"$$$/lrc-ai-assistant/PluginInfoDialogSections/validateBeforeSaving=Vor dem Speichern überprüfen"
diff --git a/lrc-ai-assistant.lrplugin/TranslatedStrings_en.txt b/lrc-ai-assistant.lrplugin/TranslatedStrings_en.txt
@@ -18,6 +18,8 @@
 "$$$/lrc-ai-assistant/Defaults/defaultSystemInstruction=You are classifying images for photo management. Be very specific an detailed. Do not return more than 25 unique keywords."
 "$$$/lrc-ai-assistant/Defaults/defaultTask=Describe the image contents, including all recognized objects."
 "$$$/lrc-ai-assistant/GeminiAPI/NoAPIkey=No Gemini API key configured in add-ons manager."
+"$$$/lrc-ai-assistant/GeminiAPI/gpsAddon=This photo was taken at the following coordinates: "
+"$$$/lrc-ai-assistant/GeminiAPI/keywordAddon=Some keywords are:"
 "$$$/lrc-ai-assistant/GenerateImageInfo/SkipFromHere=Save following without reviewing."
 "$$$/lrc-ai-assistant/GenerateImageInfo/UsedTokenDialog/GeneratedCosts=Generated costs"
 "$$$/lrc-ai-assistant/GenerateImageInfo/UsedTokenDialog/Title=Generation costs"
@@ -44,4 +46,4 @@
 "$$$/lrc-ai-assistant/PluginInfoDialogSections/keywords=Keywords"
 "$$$/lrc-ai-assistant/PluginInfoDialogSections/showCosts=Show costs (without any warranty!!!)"
 "$$$/lrc-ai-assistant/PluginInfoDialogSections/title=Title"
-"$$$/lrc-ai-assistant/PluginInfoDialogSections/validateBeforeSaving=Validate before saving"
+"$$$/lrc-ai-assistant/PluginInfoDialogSections/validateBeforeSaving=Validate before saving"