#29281: Applying feedback and fixing dirty models bug

core-web/yarn.lock

@@ -22128,7 +22128,7 @@ string-length@^4.0.1:
     char-regex "^1.0.2"
     strip-ansi "^6.0.0"
 
-"string-width-cjs@npm:string-width@^4.2.0", "string-width@^1.0.2 || 2 || 3 || 4", string-width@^4.1.0, string-width@^4.2.0, string-width@^4.2.3:
+"string-width-cjs@npm:string-width@^4.2.0":
   version "4.2.3"
   resolved "https://registry.yarnpkg.com/string-width/-/string-width-4.2.3.tgz#269c7117d27b05ad2e536830a8ec895ef9c6d010"
   integrity sha512-wKyQRQpjJ0sIp62ErSZdGsjMJWsap5oRNihHhu6G7JVO/9jIB6UyevL+tXuOqrng8j/cxKTWyWUwvSTriiZz/g==
@@ -22146,6 +22146,15 @@ string-width@^1.0.1:
     is-fullwidth-code-point "^1.0.0"
     strip-ansi "^3.0.0"
 
+"string-width@^1.0.2 || 2 || 3 || 4", string-width@^4.1.0, string-width@^4.2.0, string-width@^4.2.3:
+  version "4.2.3"
+  resolved "https://registry.yarnpkg.com/string-width/-/string-width-4.2.3.tgz#269c7117d27b05ad2e536830a8ec895ef9c6d010"
+  integrity sha512-wKyQRQpjJ0sIp62ErSZdGsjMJWsap5oRNihHhu6G7JVO/9jIB6UyevL+tXuOqrng8j/cxKTWyWUwvSTriiZz/g==
+  dependencies:
+    emoji-regex "^8.0.0"
+    is-fullwidth-code-point "^3.0.0"
+    strip-ansi "^6.0.1"
+
 string-width@^2.0.0, string-width@^2.1.1:
   version "2.1.1"
   resolved "https://registry.yarnpkg.com/string-width/-/string-width-2.1.1.tgz#ab93f27a8dc13d28cac815c462143a6d9012ae9e"
@@ -22242,7 +22251,7 @@ stringify-package@^1.0.0, stringify-package@^1.0.1:
   resolved "https://registry.yarnpkg.com/stringify-package/-/stringify-package-1.0.1.tgz#e5aa3643e7f74d0f28628b72f3dad5cecfc3ba85"
   integrity sha512-sa4DUQsYciMP1xhKWGuFM04fB0LG/9DlluZoSVywUMRNvzid6XucHK0/90xGxRoHrAaROrcHK1aPKaijCtSrhg==
 
-"strip-ansi-cjs@npm:strip-ansi@^6.0.1", strip-ansi@^6.0.0, strip-ansi@^6.0.1:
+"strip-ansi-cjs@npm:strip-ansi@^6.0.1":
   version "6.0.1"
   resolved "https://registry.yarnpkg.com/strip-ansi/-/strip-ansi-6.0.1.tgz#9e26c63d30f53443e9489495b2105d37b67a85d9"
   integrity sha512-Y38VPSHcqkFrCpFnQ9vuSXmquuv5oXOKpGeT6aGrr3o3Gc9AlVa6JBfUSOCnbxGGZF+/0ooI7KrPuUSztUdU5A==
@@ -22270,6 +22279,13 @@ strip-ansi@^5.0.0, strip-ansi@^5.1.0, strip-ansi@^5.2.0:
   dependencies:
     ansi-regex "^4.1.0"
 
+strip-ansi@^6.0.0, strip-ansi@^6.0.1:
+  version "6.0.1"
+  resolved "https://registry.yarnpkg.com/strip-ansi/-/strip-ansi-6.0.1.tgz#9e26c63d30f53443e9489495b2105d37b67a85d9"
+  integrity sha512-Y38VPSHcqkFrCpFnQ9vuSXmquuv5oXOKpGeT6aGrr3o3Gc9AlVa6JBfUSOCnbxGGZF+/0ooI7KrPuUSztUdU5A==
+  dependencies:
+    ansi-regex "^5.0.1"
+
 strip-ansi@^7.0.1:
   version "7.1.0"
   resolved "https://registry.yarnpkg.com/strip-ansi/-/strip-ansi-7.1.0.tgz#d5b6568ca689d8561370b0707685d22434faff45"
@@ -24269,7 +24285,7 @@ worker-farm@^1.6.0, worker-farm@^1.7.0:
   dependencies:
     errno "~0.1.7"
 
-"wrap-ansi-cjs@npm:wrap-ansi@^7.0.0", wrap-ansi@^7.0.0:
+"wrap-ansi-cjs@npm:wrap-ansi@^7.0.0":
   version "7.0.0"
   resolved "https://registry.yarnpkg.com/wrap-ansi/-/wrap-ansi-7.0.0.tgz#67e145cff510a6a6984bdf1152911d69d2eb9e43"
   integrity sha512-YVGIj2kamLSTxw6NsZjoBxfSwsn0ycdesmc4p+Q21c5zPuZ1pl+NfxVdxPtdHvmNVOQ6XSYG4AUtyt/Fi7D16Q==
@@ -24304,6 +24320,15 @@ wrap-ansi@^6.2.0:
     string-width "^4.1.0"
     strip-ansi "^6.0.0"
 
+wrap-ansi@^7.0.0:
+  version "7.0.0"
+  resolved "https://registry.yarnpkg.com/wrap-ansi/-/wrap-ansi-7.0.0.tgz#67e145cff510a6a6984bdf1152911d69d2eb9e43"
+  integrity sha512-YVGIj2kamLSTxw6NsZjoBxfSwsn0ycdesmc4p+Q21c5zPuZ1pl+NfxVdxPtdHvmNVOQ6XSYG4AUtyt/Fi7D16Q==
+  dependencies:
+    ansi-styles "^4.0.0"
+    string-width "^4.1.0"
+    strip-ansi "^6.0.0"
+
 wrap-ansi@^8.1.0:
   version "8.1.0"
   resolved "https://registry.yarnpkg.com/wrap-ansi/-/wrap-ansi-8.1.0.tgz#56dc22368ee570face1b49819975d9b9a5ead214"

dotCMS/src/main/java/com/dotcms/ai/api/CompletionsAPIImpl.java

@@ -204,7 +204,7 @@ private String getTextPrompt(final String prompt, final String supportingContent
     }
 
     private int countTokens(final String testString) {
-        return EncodingUtil.REGISTRY
+        return EncodingUtil.get().registry
                 .getEncodingForModel(config.get().getModel().getCurrentModel())
                 .map(enc -> enc.countTokens(testString))
                 .orElseThrow(() -> new DotRuntimeException("Encoder not found"));

dotCMS/src/main/java/com/dotcms/ai/api/EmbeddingsAPIImpl.java

@@ -328,7 +328,10 @@ public Tuple2<Integer, List<Float>> pullOrGenerateEmbeddings(final String conten
             return cachedEmbeddings;
         }
 
-        final List<Integer> tokens = EncodingUtil.ENCODING.get().encode(content);
+        final List<Integer> tokens = EncodingUtil.get()
+                .getEncoding()
+                .map(encoding -> encoding.encode(content))
+                .orElse(List.of());
         if (tokens.isEmpty()) {
             debugLogger(this.getClass(), () -> String.format("No tokens for content ID '%s' were encoded: %s", contentId, content));
             return Tuple.of(0, List.of());

dotCMS/src/main/java/com/dotcms/ai/api/EmbeddingsRunner.java

@@ -68,7 +68,10 @@ public void run() {
             int totalTokens = 0;
             for (int end = iterator.next(); end != BreakIterator.DONE; start = end, end = iterator.next()) {
                 final String sentence = cleanContent.substring(start, end);
-                final int tokenCount = EncodingUtil.ENCODING.get().countTokens(sentence);
+                final int tokenCount = EncodingUtil.get()
+                        .getEncoding()
+                        .map(encoding -> encoding.countTokens(sentence))
+                        .orElse(0);
                 totalTokens += tokenCount;
 
                 if (totalTokens < splitAtTokens) {

dotCMS/src/main/java/com/dotcms/ai/app/AIAppUtil.java

@@ -117,7 +117,7 @@ public String discoverSecret(final Map<String, Secret> secrets, final AppKeys ke
      * @return the list of split secret values
      */
     public List<String> splitDiscoveredSecret(final Map<String, Secret> secrets, final AppKeys key) {
-        return Arrays.stream(discoverSecret(secrets, key).split(","))
+        return Arrays.stream(Optional.ofNullable(discoverSecret(secrets, key)).orElse(StringPool.BLANK).split(","))
                 .map(String::trim)
                 .map(String::toLowerCase)
                 .collect(Collectors.toList());

dotCMS/src/main/java/com/dotcms/ai/app/AIModels.java

@@ -131,18 +131,16 @@ public Optional<AIModel> findModel(final String host, final AIModelType type) {
      */
     public void resetModels(final Host host) {
         final String hostKey = host.getHostname();
-        synchronized (AIModels.class) {
-            Optional.ofNullable(internalModels.get(hostKey)).ifPresent(models -> {
-                models.clear();
-                internalModels.remove(hostKey);
-            });
-            modelsByName.keySet()
-                    .stream()
-                    .filter(key -> key._1.equals(hostKey))
-                    .collect(Collectors.toSet())
-                    .forEach(modelsByName::remove);
-            ConfigService.INSTANCE.config(host);
-        }
+        Optional.ofNullable(internalModels.get(hostKey)).ifPresent(models -> {
+            models.clear();
+            internalModels.remove(hostKey);
+        });
+        modelsByName.keySet()
+                .stream()
+                .filter(key -> key._1.equals(hostKey))
+                .collect(Collectors.toSet())
+                .forEach(modelsByName::remove);
+        ConfigService.INSTANCE.config(host);
     }
 
     /**
@@ -152,31 +150,29 @@ public void resetModels(final Host host) {
      * @return a list of supported model names
      */
     public List<String> getOrPullSupportedModels() {
-        synchronized (supportedModelsCache) {
-            final List<String> cached = supportedModelsCache.getIfPresent(SUPPORTED_MODELS_KEY);
-            if (CollectionUtils.isNotEmpty(cached)) {
-                return cached;
-            }
-
-            final AppConfig appConfig = appConfigSupplier.get();
-            if (!appConfig.isEnabled()) {
-                Logger.debug(this, "OpenAI is not enabled, returning empty list of supported models");
-                return List.of();
-            }
-
-            final List<String> supported = Try.of(() ->
-                            fetchOpenAIModels(appConfig)
-                                    .getResponse()
-                                    .getData()
-                                    .stream()
-                                    .map(OpenAIModel::getId)
-                                    .map(String::toLowerCase)
-                                    .collect(Collectors.toList()))
-                    .getOrElse(Optional.ofNullable(cached).orElse(List.of()));
-            supportedModelsCache.put(SUPPORTED_MODELS_KEY, supported);
-
-            return supported;
+        final List<String> cached = supportedModelsCache.getIfPresent(SUPPORTED_MODELS_KEY);
+        if (CollectionUtils.isNotEmpty(cached)) {
+            return cached;
+        }
+
+        final AppConfig appConfig = appConfigSupplier.get();
+        if (!appConfig.isEnabled()) {
+            Logger.debug(this, "OpenAI is not enabled, returning empty list of supported models");
+            return List.of();
         }
+
+        final List<String> supported = Try.of(() ->
+                        fetchOpenAIModels(appConfig)
+                                .getResponse()
+                                .getData()
+                                .stream()
+                                .map(OpenAIModel::getId)
+                                .map(String::toLowerCase)
+                                .collect(Collectors.toList()))
+                .getOrElse(Optional.ofNullable(cached).orElse(List.of()));
+        supportedModelsCache.put(SUPPORTED_MODELS_KEY, supported);
+
+        return supported;
     }
 
     /**

dotCMS/src/main/java/com/dotcms/ai/app/AppConfig.java

@@ -1,6 +1,5 @@
 package com.dotcms.ai.app;
 
-import com.dotcms.ai.util.OpenAIRequest;
 import com.dotcms.security.apps.Secret;
 import com.dotmarketing.exception.DotRuntimeException;
 import com.dotmarketing.util.Config;
@@ -44,7 +43,6 @@ public AppConfig(final String host, final Map<String, Secret> secrets) {
         this.host = host;
 
         final AIAppUtil aiAppUtil = AIAppUtil.get();
-        apiKey = aiAppUtil.discoverEnvSecret(secrets, AppKeys.API_KEY);
 
         AIModels.get().loadModels(
                 this.host,
@@ -53,13 +51,14 @@ public AppConfig(final String host, final Map<String, Secret> secrets) {
                         aiAppUtil.createImageModel(secrets),
                         aiAppUtil.createEmbeddingsModel(secrets)));
 
-        model = resolveModel(AIModelType.TEXT);
-        imageModel = resolveModel(AIModelType.IMAGE);
-        embeddingsModel = resolveModel(AIModelType.EMBEDDINGS);
-
         apiUrl = aiAppUtil.discoverEnvSecret(secrets, AppKeys.API_URL);
         apiImageUrl = aiAppUtil.discoverEnvSecret(secrets, AppKeys.API_IMAGE_URL);
         apiEmbeddingsUrl = discoverEmbeddingsApiUrl(secrets);
+        apiKey = aiAppUtil.discoverEnvSecret(secrets, AppKeys.API_KEY);
+
+        model = resolveModel(AIModelType.TEXT);
+        imageModel = resolveModel(AIModelType.IMAGE);
+        embeddingsModel = resolveModel(AIModelType.EMBEDDINGS);
 
         rolePrompt = aiAppUtil.discoverSecret(secrets, AppKeys.ROLE_PROMPT);
         textPrompt = aiAppUtil.discoverSecret(secrets, AppKeys.TEXT_PROMPT);
@@ -73,13 +72,13 @@ public AppConfig(final String host, final Map<String, Secret> secrets) {
         Logger.debug(getClass(), () -> "apiUrl: " + apiUrl);
         Logger.debug(getClass(), () -> "apiImageUrl: " + apiImageUrl);
         Logger.debug(getClass(), () -> "embeddingsUrl: " + apiEmbeddingsUrl);
-        Logger.debug(getClass(), () -> "model: " + model);
-        Logger.debug(getClass(), () -> "imageModel: " + imageModel);
-        Logger.debug(getClass(), () -> "embeddingsModel: " + embeddingsModel);
         Logger.debug(getClass(), () -> "rolePrompt: " + rolePrompt);
         Logger.debug(getClass(), () -> "textPrompt: " + textPrompt);
+        Logger.debug(getClass(), () -> "model: " + model);
         Logger.debug(getClass(), () -> "imagePrompt: " + imagePrompt);
+        Logger.debug(getClass(), () -> "imageModel: " + imageModel);
         Logger.debug(getClass(), () -> "imageSize: " + imageSize);
+        Logger.debug(getClass(), () -> "embeddingsModel: " + embeddingsModel);
         Logger.debug(getClass(), () -> "listerIndexer: " + listenerIndexer);
     }
 
@@ -263,24 +262,24 @@ public AIModel resolveModel(final AIModelType type) {
      * @param modelName the name of the model to find
      */
     public AIModel resolveModelOrThrow(final String modelName) {
-        final AIModel model = AIModels.get()
+        final AIModel aiModel = AIModels.get()
                 .findModel(host, modelName)
                 .orElseThrow(() -> {
                     final String supported = String.join(", ", AIModels.get().getOrPullSupportedModels());
                     return new DotRuntimeException(
                             "Unable to find model: [" + modelName + "]. Only [" + supported + "] are supported ");
                 });
 
-        if (!model.isOperational()) {
-            Logger.debug(
-                    OpenAIRequest.class,
-                    String.format(
+        if (!aiModel.isOperational()) {
+            debugLogger(
+                    AppConfig.class,
+                    () -> String.format(
                             "Resolved model [%s] is not operational, avoiding its usage",
-                            model.getCurrentModel()));
-            throw new DotRuntimeException(String.format("Model [%s] is not operational", model.getCurrentModel()));
+                            aiModel.getCurrentModel()));
+            throw new DotRuntimeException(String.format("Model [%s] is not operational", aiModel.getCurrentModel()));
         }
 
-        return model;
+        return aiModel;
     }
 
     /**

dotCMS/src/main/java/com/dotcms/ai/app/AppKeys.java

@@ -14,20 +14,20 @@ public enum AppKeys {
     IMAGE_PROMPT("imagePrompt", "Use 16:9 aspect ratio."),
     IMAGE_SIZE("imageSize", "1024x1024"),
     TEXT_MODEL_NAMES("textModelNames", "gpt-3.5-turbo-16k"),
-    TEXT_MODEL_TOKENS_PER_MINUTE("textModelTokensPerMinute", "1000"),
-    TEXT_MODEL_API_PER_MINUTE("textModelApiPerMinute", "1000"),
-    TEXT_MODEL_MAX_TOKENS("textModelMaxTokens", "1000"),
+    TEXT_MODEL_TOKENS_PER_MINUTE("textModelTokensPerMinute", "180000"),
+    TEXT_MODEL_API_PER_MINUTE("textModelApiPerMinute", "3500"),
+    TEXT_MODEL_MAX_TOKENS("textModelMaxTokens", "16384"),
     TEXT_MODEL_COMPLETION("textModelCompletion", "true"),
     IMAGE_MODEL_NAMES("imageModelNames", "dall-e-3"),
-    IMAGE_MODEL_TOKENS_PER_MINUTE("imageModelTokensPerMinute", "1000"),
-    IMAGE_MODEL_API_PER_MINUTE("imageModelApiPerMinute", "1000"),
-    IMAGE_MODEL_MAX_TOKENS("imageModelMaxTokens", "1000"),
-    IMAGE_MODEL_COMPLETION("imageModelCompletion", "true"),
+    IMAGE_MODEL_TOKENS_PER_MINUTE("imageModelTokensPerMinute", "0"),
+    IMAGE_MODEL_API_PER_MINUTE("imageModelApiPerMinute", "50"),
+    IMAGE_MODEL_MAX_TOKENS("imageModelMaxTokens", "0"),
+    IMAGE_MODEL_COMPLETION("imageModelCompletion", "false"),
     EMBEDDINGS_MODEL_NAMES("embeddingsModelNames", "text-embedding-ada-002"),
-    EMBEDDINGS_MODEL_TOKENS_PER_MINUTE("embeddingsModelTokensPerMinute", "1000"),
-    EMBEDDINGS_MODEL_API_PER_MINUTE("embeddingsModelApiPerMinute", "1000"),
-    EMBEDDINGS_MODEL_MAX_TOKENS("embeddingsModelMaxTokens", "1000"),
-    EMBEDDINGS_MODEL_COMPLETION("embeddingsModelCompletion", "true"),
+    EMBEDDINGS_MODEL_TOKENS_PER_MINUTE("embeddingsModelTokensPerMinute", "1000000"),
+    EMBEDDINGS_MODEL_API_PER_MINUTE("embeddingsModelApiPerMinute", "3000"),
+    EMBEDDINGS_MODEL_MAX_TOKENS("embeddingsModelMaxTokens", "8191"),
+    EMBEDDINGS_MODEL_COMPLETION("embeddingsModelCompletion", "false"),
     EMBEDDINGS_SPLIT_AT_TOKENS("com.dotcms.ai.embeddings.split.at.tokens", "512"),
     EMBEDDINGS_MINIMUM_TEXT_LENGTH_TO_INDEX("com.dotcms.ai.embeddings.minimum.text.length", "64"),
     EMBEDDINGS_MINIMUM_FILE_SIZE_TO_INDEX("com.dotcms.ai.embeddings.minimum.file.size", "1024"),
@@ -57,7 +57,7 @@ public enum AppKeys {
     public final String key;
     public final String defaultValue;
 
-    AppKeys(String key, String defaultValue) {
+    AppKeys(final String key, final String defaultValue) {
         this.key = key;
         this.defaultValue = defaultValue;
     }

dotCMS/src/main/java/com/dotcms/ai/listener/AIAppListener.java

@@ -49,7 +49,7 @@ public void notify(final AppSecretSavedEvent event) {
         final String hostId = event.getHostIdentifier();
         final Host host = Try.of(() -> hostAPI.find(hostId, APILocator.systemUser(), false)).getOrNull();
 
-        Optional.ofNullable(host).ifPresent(found ->  AIModels.get().resetModels(found));
+        Optional.ofNullable(host).ifPresent(found -> AIModels.get().resetModels(found));
     }
 
     @Override

dotCMS/src/main/java/com/dotcms/ai/listener/EmbeddingContentListener.java

@@ -2,7 +2,6 @@
 
 import com.dotcms.ai.api.EmbeddingsAPI;
 import com.dotcms.ai.app.AppConfig;
-import com.dotcms.ai.app.AppKeys;
 import com.dotcms.ai.app.ConfigService;
 import com.dotcms.ai.db.EmbeddingsDTO;
 import com.dotcms.content.elasticsearch.business.event.ContentletArchiveEvent;
@@ -18,7 +17,6 @@
 import com.dotmarketing.util.Logger;
 import com.dotmarketing.util.json.JSONObject;
 import io.vavr.control.Try;
-import org.apache.commons.lang3.StringUtils;
 
 import java.util.List;
 import java.util.Map;
@@ -85,7 +83,7 @@ private AppConfig getAppConfig(final String hostId) {
                 .getOrElse(APILocator.systemHost());
 
         final AppConfig appConfig = ConfigService.INSTANCE.config(host);
-        if (StringUtils.isBlank(appConfig.getApiKey())) {
+        if (!appConfig.isEnabled()) {
             throw new DotRuntimeException("No API key found in app config");
         }
 
@@ -127,10 +125,7 @@ private void addToIndexesIfNeeded(final Contentlet contentlet) {
                     .stream()
                     .filter(typeFields -> contentType.equalsIgnoreCase(typeFields.getKey()))
                     .forEach(e -> EmbeddingsAPI.impl()
-                            .generateEmbeddingsForContent(
-                                    contentlet,
-                                    e.getValue(),
-                                    indexName));
+                            .generateEmbeddingsForContent(contentlet, e.getValue(), indexName));
         }
     }