Merge pull request #507 from jmartisk/fault-tolerance

Turn off built-in fault tolerance by default
quarkiverse · Apr 25, 2024 · e1c980f · e1c980f
2 parents c61b2ea + 734c463
commit e1c980f
Show file tree

Hide file tree

Showing 11 changed files with 43 additions and 19 deletions.
diff --git a/...me/src/main/java/io/quarkiverse/langchain4j/anthropic/runtime/config/ChatModelConfig.java b/...me/src/main/java/io/quarkiverse/langchain4j/anthropic/runtime/config/ChatModelConfig.java
@@ -50,9 +50,12 @@ public interface ChatModelConfig {
     Integer topK();
 
     /**
-     * The maximum number of retries for API requests.
+     * The maximum number of times to retry. 1 means exactly one attempt, with retrying disabled.
+     *
+     * @deprecated Using the fault tolerance mechanisms built in Langchain4j is not recommended. If possible,
+     *             use MicroProfile Fault Tolerance instead.
      */
-    @WithDefault("3")
+    @WithDefault("1")
     Integer maxRetries();
 
     /**

diff --git a/...me/src/main/java/io/quarkiverse/langchain4j/cohere/runtime/QuarkusCohereScoringModel.java b/...me/src/main/java/io/quarkiverse/langchain4j/cohere/runtime/QuarkusCohereScoringModel.java
@@ -40,6 +40,9 @@ public QuarkusCohereScoringModel(
             Integer maxRetries) {
         this.model = model;
         this.maxRetries = maxRetries;
+        if (this.maxRetries < 1) {
+            throw new IllegalArgumentException("max-retries must be at least 1");
+        }
         ClientHeadersFactory factory = new ClientHeadersFactory() {
             @Override
             public MultivaluedMap<String, String> update(MultivaluedMap<String, String> incomingHeaders,

diff --git a/.../main/java/io/quarkiverse/langchain4j/cohere/runtime/config/CohereScoringModelConfig.java b/.../main/java/io/quarkiverse/langchain4j/cohere/runtime/config/CohereScoringModelConfig.java
@@ -20,9 +20,12 @@ public interface CohereScoringModelConfig {
     Duration timeout();
 
     /**
-     * Maximum number of retries for Cohere API invocations.
+     * The maximum number of times to retry. 1 means exactly one attempt, with retrying disabled.
+     *
+     * @deprecated Using the fault tolerance mechanisms built in Langchain4j is not recommended. If possible,
+     *             use MicroProfile Fault Tolerance instead.
      */
-    @WithDefault("3")
+    @WithDefault("1")
     Integer maxRetries();
 
 }
diff --git a/docs/modules/ROOT/pages/includes/quarkus-langchain4j-anthropic.adoc b/docs/modules/ROOT/pages/includes/quarkus-langchain4j-anthropic.adoc
@@ -243,7 +243,7 @@ a| [[quarkus-langchain4j-anthropic_quarkus-langchain4j-anthropic-chat-model-max-
 
 [.description]
 --
-The maximum number of retries for API requests.
+The maximum number of times to retry. 1 means exactly one attempt, with retrying disabled.
 
 ifdef::add-copy-button-to-env-var[]
 Environment variable: env_var_with_copy_button:+++QUARKUS_LANGCHAIN4J_ANTHROPIC_CHAT_MODEL_MAX_RETRIES+++[]
@@ -252,7 +252,7 @@ ifndef::add-copy-button-to-env-var[]
 Environment variable: `+++QUARKUS_LANGCHAIN4J_ANTHROPIC_CHAT_MODEL_MAX_RETRIES+++`
 endif::add-copy-button-to-env-var[]
 --|int 
-|`3`
+|`1`
 
 
 a| [[quarkus-langchain4j-anthropic_quarkus-langchain4j-anthropic-chat-model-stop-sequences]]`link:#quarkus-langchain4j-anthropic_quarkus-langchain4j-anthropic-chat-model-stop-sequences[quarkus.langchain4j.anthropic.chat-model.stop-sequences]`
@@ -527,7 +527,7 @@ a| [[quarkus-langchain4j-anthropic_quarkus-langchain4j-anthropic-model-name-chat
 
 [.description]
 --
-The maximum number of retries for API requests.
+The maximum number of times to retry. 1 means exactly one attempt, with retrying disabled.
 
 ifdef::add-copy-button-to-env-var[]
 Environment variable: env_var_with_copy_button:+++QUARKUS_LANGCHAIN4J_ANTHROPIC__MODEL_NAME__CHAT_MODEL_MAX_RETRIES+++[]
@@ -536,7 +536,7 @@ ifndef::add-copy-button-to-env-var[]
 Environment variable: `+++QUARKUS_LANGCHAIN4J_ANTHROPIC__MODEL_NAME__CHAT_MODEL_MAX_RETRIES+++`
 endif::add-copy-button-to-env-var[]
 --|int 
-|`3`
+|`1`
 
 
 a| [[quarkus-langchain4j-anthropic_quarkus-langchain4j-anthropic-model-name-chat-model-stop-sequences]]`link:#quarkus-langchain4j-anthropic_quarkus-langchain4j-anthropic-model-name-chat-model-stop-sequences[quarkus.langchain4j.anthropic."model-name".chat-model.stop-sequences]`

diff --git a/docs/modules/ROOT/pages/includes/quarkus-langchain4j-cohere.adoc b/docs/modules/ROOT/pages/includes/quarkus-langchain4j-cohere.adoc
@@ -84,7 +84,7 @@ a| [[quarkus-langchain4j-cohere_quarkus-langchain4j-cohere-scoring-model-max-ret
 
 [.description]
 --
-Maximum number of retries for Cohere API invocations.
+The maximum number of times to retry. 1 means exactly one attempt, with retrying disabled.
 
 ifdef::add-copy-button-to-env-var[]
 Environment variable: env_var_with_copy_button:+++QUARKUS_LANGCHAIN4J_COHERE_SCORING_MODEL_MAX_RETRIES+++[]
@@ -93,7 +93,7 @@ ifndef::add-copy-button-to-env-var[]
 Environment variable: `+++QUARKUS_LANGCHAIN4J_COHERE_SCORING_MODEL_MAX_RETRIES+++`
 endif::add-copy-button-to-env-var[]
 --|int 
-|`3`
+|`1`
 
 |===
 ifndef::no-duration-note[]

diff --git a/docs/modules/ROOT/pages/includes/quarkus-langchain4j-openai.adoc b/docs/modules/ROOT/pages/includes/quarkus-langchain4j-openai.adoc
@@ -152,7 +152,7 @@ a| [[quarkus-langchain4j-openai_quarkus-langchain4j-openai-max-retries]]`link:#q
 
 [.description]
 --
-The maximum number of times to retry
+The maximum number of times to retry. 1 means exactly one attempt, with retrying disabled.
 
 ifdef::add-copy-button-to-env-var[]
 Environment variable: env_var_with_copy_button:+++QUARKUS_LANGCHAIN4J_OPENAI_MAX_RETRIES+++[]
@@ -161,7 +161,7 @@ ifndef::add-copy-button-to-env-var[]
 Environment variable: `+++QUARKUS_LANGCHAIN4J_OPENAI_MAX_RETRIES+++`
 endif::add-copy-button-to-env-var[]
 --|int 
-|`3`
+|`1`
 
 
 a| [[quarkus-langchain4j-openai_quarkus-langchain4j-openai-log-requests]]`link:#quarkus-langchain4j-openai_quarkus-langchain4j-openai-log-requests[quarkus.langchain4j.openai.log-requests]`
@@ -771,7 +771,7 @@ a| [[quarkus-langchain4j-openai_quarkus-langchain4j-openai-model-name-max-retrie
 
 [.description]
 --
-The maximum number of times to retry
+The maximum number of times to retry. 1 means exactly one attempt, with retrying disabled.
 
 ifdef::add-copy-button-to-env-var[]
 Environment variable: env_var_with_copy_button:+++QUARKUS_LANGCHAIN4J_OPENAI__MODEL_NAME__MAX_RETRIES+++[]
@@ -780,7 +780,7 @@ ifndef::add-copy-button-to-env-var[]
 Environment variable: `+++QUARKUS_LANGCHAIN4J_OPENAI__MODEL_NAME__MAX_RETRIES+++`
 endif::add-copy-button-to-env-var[]
 --|int 
-|`3`
+|`1`
 
 
 a| [[quarkus-langchain4j-openai_quarkus-langchain4j-openai-model-name-log-requests]]`link:#quarkus-langchain4j-openai_quarkus-langchain4j-openai-model-name-log-requests[quarkus.langchain4j.openai."model-name".log-requests]`

diff --git a/...i/runtime/src/main/java/io/quarkiverse/langchain4j/azure/openai/AzureOpenAiChatModel.java b/...i/runtime/src/main/java/io/quarkiverse/langchain4j/azure/openai/AzureOpenAiChatModel.java
@@ -93,7 +93,10 @@ public AzureOpenAiChatModel(String endpoint,
         this.maxTokens = maxTokens;
         this.presencePenalty = presencePenalty;
         this.frequencyPenalty = frequencyPenalty;
-        this.maxRetries = getOrDefault(maxRetries, 3);
+        this.maxRetries = getOrDefault(maxRetries, 1);
+        if (this.maxRetries < 1) {
+            throw new IllegalArgumentException("max-retries must be at least 1");
+        }
         this.tokenizer = tokenizer;
         this.responseFormat = responseFormat;
     }

diff --git a/...time/src/main/java/io/quarkiverse/langchain4j/azure/openai/AzureOpenAiEmbeddingModel.java b/...time/src/main/java/io/quarkiverse/langchain4j/azure/openai/AzureOpenAiEmbeddingModel.java
@@ -61,6 +61,9 @@ public AzureOpenAiEmbeddingModel(String endpoint,
             Boolean logResponses) {
 
         timeout = getOrDefault(timeout, ofSeconds(60));
+        if (maxRetries < 1) {
+            throw new IllegalArgumentException("max-retries must be at least 1");
+        }
 
         this.client = ((QuarkusOpenAiClient.Builder) OpenAiClient.builder()
                 .baseUrl(ensureNotBlank(endpoint, "endpoint"))

diff --git a/.../io/quarkiverse/langchain4j/azure/openai/runtime/config/LangChain4jAzureOpenAiConfig.java b/.../io/quarkiverse/langchain4j/azure/openai/runtime/config/LangChain4jAzureOpenAiConfig.java
@@ -93,9 +93,12 @@ interface AzureAiConfig {
         Duration timeout();
 
         /**
-         * The maximum number of times to retry
+         * The maximum number of times to retry. 1 means exactly one attempt, with retrying disabled.
+         *
+         * @deprecated Using the fault tolerance mechanisms built in Langchain4j is not recommended. If possible,
+         *             use MicroProfile Fault Tolerance instead.
          */
-        @WithDefault("3")
+        @WithDefault("1")
         Integer maxRetries();
 
         /**

diff --git a/...illa/runtime/src/main/java/io/quarkiverse/langchain4j/openai/QuarkusOpenAiImageModel.java b/...illa/runtime/src/main/java/io/quarkiverse/langchain4j/openai/QuarkusOpenAiImageModel.java
@@ -49,6 +49,9 @@ public QuarkusOpenAiImageModel(String baseUrl, String apiKey, String organizatio
         this.user = user;
         this.responseFormat = responseFormat;
         this.maxRetries = maxRetries;
+        if (this.maxRetries < 1) {
+            throw new IllegalArgumentException("max-retries must be at least 1");
+        }
         this.persistDirectory = persistDirectory;
 
         this.client = QuarkusOpenAiClient.builder()

diff --git a/...c/main/java/io/quarkiverse/langchain4j/openai/runtime/config/LangChain4jOpenAiConfig.java b/...c/main/java/io/quarkiverse/langchain4j/openai/runtime/config/LangChain4jOpenAiConfig.java
@@ -62,9 +62,12 @@ interface OpenAiConfig {
         Duration timeout();
 
         /**
-         * The maximum number of times to retry
+         * The maximum number of times to retry. 1 means exactly one attempt, with retrying disabled.
+         *
+         * @deprecated Using the built-in fault tolerance mechanisms is not recommended. If possible,
+         *             use MicroProfile Fault Tolerance instead.
          */
-        @WithDefault("3")
+        @WithDefault("1")
         Integer maxRetries();
 
         /**