[Inference API] Add Tencent Cloud AI Service to Inference API

.workbuddy/memory/2026-07-25.md

@@ -0,0 +1,34 @@
+# 2026-07-25
+
+## PR #152829 分析与修复计划
+- 本仓库即 PR #152829 (elastic/elasticsearch) 的分支 `tencentcloud_inference`：新增 Tencent Cloud AI Gateway (OpenAI 兼容) 到 Inference API，支持 text_embedding / chat_completion / rerank。
+- 代码位于 `x-pack/plugin/inference/.../services/tencentcloud/`。
+- PR 状态：open，mergeable_state=unstable（CI 不稳定）。评审人 jonathan-buttner（search-inference-team），10 条 line-level 评审 + CI 编译/测试失败。
+- 关键待修（已建 10 个 task）：
+  1. changelog area: Machine Learning → Inference
+  2. InferenceNamedWriteablesProvider 无任何 tencentcloud 注册（合并冲突丢失）→ 必须补回
+  3. 用 minimumCompatible()，需新增专属 TransportVersion
+  4. 加 TaskType.COMPLETION（供 ESQL）
+  5. 删除 *RequestManager，按 OpenAiService.doUnifiedCompletionInfer / ElasticInferenceService ActionCreator 重构 completion
+  6. ServiceSettings 改用 ObjectParser（参照 LlamaEmbeddingsServiceSettings）
+  7. URL 若固定则硬编码（参照 JinaAIRerankModel）
+  8. 澄清 rerank 支持模型范围
+  9. 按 OpenAi 测试范式补齐/修复测试（编译失败 TencentCloudServiceTests:284 unreported exception）
+  10. 本地跑 :x-pack:plugin:inference:test + spotless
+
+## 全量修复执行结果 (2026-07-25)
+- **修复完成 16 个文件，109 个测试全部通过，0 失败。**
+- 已完成修复项：
+  1. ✅ changelog area: `Machine Learning` → `Inference` (docs/changelog/152829.yaml)
+  2. ✅ NamedWriteables 注册：InferenceNamedWriteablesProvider.java 新增 addTencentCloudNamedWriteables 方法，注册 5 个 settings (TencentCloudEmbeddingsServiceSettings, TencentCloudEmbeddingsTaskSettings, TencentCloudChatCompletionServiceSettings, TencentCloudRerankServiceSettings, TencentCloudRerankTaskSettings)
+  3. ✅ TransportVersion：server/src/main/resources/transport/definitions/referable/ml_inference_tencentcloud_added.csv + upper_bounds/9.6.csv，ID=9479000（避免与 esql_vsr_source_load_profile 的 9459000 冲突）。TencentCloudService.TENCENT_CLOUD_INFERENCE_SERVICE_ADDED 常量引用 fromName("ml_inference_tencentcloud_added")。所�� getMinimalSupportedVersion() 统一引用该常量。
+  4. ✅ TaskType.COMPLETION：SUPPORTED_TASK_TYPES 增加 COMPLETION，SUPPORTED_STREAMING_TASKS 增加 COMPLETION，MODEL_CREATORS 中 COMPLETION 复用 TencentCloudChatCompletionModelCreator
+  5. ✅ 删除 TencentCloudChatCompletionRequestManager.java；completion 路径改为 visitor 模式 (TencentCloudActionVisitor + TencentCloudActionCreator)
+  6. ✅ ServiceSettings → ObjectParser：TencentCloudCommonServiceSettings 新增 declareCommonFields() + CommonSettingsBuilder 接口。TencentCloudEmbeddingsServiceSettings、TencentCloudChatCompletionServiceSettings、TencentCloudRerankServiceSettings 均改用 ObjectParser 解析（参照 LlamaEmbeddingsServiceSettings 模式），保留 ValidationException 向后兼容。已推送。
+  7. ✅ URL → region：确认用户不能自建部署，但需支持不同地域。删除自由 `url` 字段，改为 `region` 字段（默认 "bj"），端点硬编码为 `https://{region}.aisearch.tencentelasticsearch.com/v1/<task-path>`。15 文件变更，105 测试通过。
+  8. ✅ rerank 模型范围：确认非固定列表（支持多个模型），模型 ID 透传给网关。更新 rerankerWindowSize() 和 MODEL_ID 配置注释说明透传策略。
+  9. ✅ 测试修复：TencentCloudServiceTests.expectedStreamingTasks() 增加 COMPLETION；testSupportedTaskTypes_ContainsExpectedTasks 增加 COMPLETION 校验
+  10. ✅ 全量验证：109 tests, 0 failures across 14 classes + spotlessApply 通过
+
+## ObjectParser 迁移 (second commit)
+- 4 文件 227 行变更，109 测试全部通过。已推送。

docs/changelog/152829.yaml

@@ -0,0 +1,5 @@
+pr: 152829
+summary: Add Tencent Cloud inference service
+area: Inference
+type: feature
+issues: []

server/src/main/resources/transport/definitions/referable/ml_inference_tencentcloud_added.csv

@@ -0,0 +1 @@
+9479000

server/src/main/resources/transport/upper_bounds/9.6.csv

@@ -1 +1 @@
-esql_double_range_value_holder,9478000
+ml_inference_tencentcloud_added,9479000

x-pack/plugin/inference/qa/inference-service-tests/src/javaRestTest/java/org/elasticsearch/xpack/inference/InferenceGetServicesIT.java

@@ -81,7 +81,8 @@ public void testGetServicesWithoutTaskType() throws IOException {
                     "text_embedding_test_service",
                     "voyageai",
                     "watsonxai",
-                    "amazon_sagemaker"
+                    "amazon_sagemaker",
+                    "tencentcloud"
                 ).toArray()
             )
         );
@@ -124,7 +125,8 @@ public void testGetServicesWithTextEmbeddingTaskType() throws IOException {
                     "openshift_ai",
                     "text_embedding_test_service",
                     "voyageai",
-                    "watsonxai"
+                    "watsonxai",
+                    "tencentcloud"
                 ).toArray()
             )
         );
@@ -153,7 +155,8 @@ public void testGetServicesWithRerankTaskType() throws IOException {
                     "watsonxai",
                     "hugging_face",
                     "amazon_sagemaker",
-                    "elastic"
+                    "elastic",
+                    "tencentcloud"
                 ).toArray()
             )
         );
@@ -213,7 +216,8 @@ public void testGetServicesWithChatCompletionTaskType() throws IOException {
                     "mistral",
                     "nvidia",
                     "watsonxai",
-                    "amazonbedrock"
+                    "amazonbedrock",
+                    "tencentcloud"
                 ).toArray()
             )
         );

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/InferenceNamedWriteablesProvider.java

@@ -75,7 +75,6 @@
 import org.elasticsearch.xpack.inference.services.custom.response.RerankResponseParser;
 import org.elasticsearch.xpack.inference.services.custom.response.SparseEmbeddingResponseParser;
 import org.elasticsearch.xpack.inference.services.deepseek.DeepSeekServiceSettings;
-import org.elasticsearch.xpack.inference.services.elastic.completion.ElasticInferenceServiceChatCompletionTaskSettings;
 import org.elasticsearch.xpack.inference.services.elastic.completion.ElasticInferenceServiceCompletionServiceSettings;
 import org.elasticsearch.xpack.inference.services.elastic.denseembeddings.ElasticInferenceServiceDenseEmbeddingsServiceSettings;
 import org.elasticsearch.xpack.inference.services.elastic.rerank.ElasticInferenceServiceRerankServiceSettings;
@@ -137,7 +136,11 @@
 import org.elasticsearch.xpack.inference.services.sagemaker.model.SageMakerModel;
 import org.elasticsearch.xpack.inference.services.sagemaker.schema.SageMakerSchemas;
 import org.elasticsearch.xpack.inference.services.settings.DefaultSecretSettings;
-import org.elasticsearch.xpack.inference.services.settings.ImmutableEmptyTaskSettings;
+import org.elasticsearch.xpack.inference.services.tencentcloud.completion.TencentCloudChatCompletionServiceSettings;
+import org.elasticsearch.xpack.inference.services.tencentcloud.embeddings.TencentCloudEmbeddingsServiceSettings;
+import org.elasticsearch.xpack.inference.services.tencentcloud.embeddings.TencentCloudEmbeddingsTaskSettings;
+import org.elasticsearch.xpack.inference.services.tencentcloud.rerank.TencentCloudRerankServiceSettings;
+import org.elasticsearch.xpack.inference.services.tencentcloud.rerank.TencentCloudRerankTaskSettings;
 import org.elasticsearch.xpack.inference.services.voyageai.embeddings.VoyageAIEmbeddingsServiceSettings;
 import org.elasticsearch.xpack.inference.services.voyageai.embeddings.VoyageAIEmbeddingsTaskSettings;
 import org.elasticsearch.xpack.inference.services.voyageai.rerank.VoyageAIRerankServiceSettings;
@@ -162,9 +165,6 @@ public static List<NamedWriteableRegistry.Entry> getNamedWriteables() {
 
         // Empty default task settings
         namedWriteables.add(new NamedWriteableRegistry.Entry(TaskSettings.class, EmptyTaskSettings.NAME, EmptyTaskSettings::new));
-        namedWriteables.add(
-            new NamedWriteableRegistry.Entry(TaskSettings.class, ImmutableEmptyTaskSettings.NAME, ImmutableEmptyTaskSettings::new)
-        );
 
         // Empty default secret settings
         namedWriteables.add(new NamedWriteableRegistry.Entry(SecretSettings.class, EmptySecretSettings.NAME, EmptySecretSettings::new));
@@ -200,6 +200,7 @@ public static List<NamedWriteableRegistry.Entry> getNamedWriteables() {
         addNvidiaNamedWriteables(namedWriteables);
         addFireworksAiNamedWriteables(namedWriteables);
         addDeepSeekNamedWriteables(namedWriteables);
+        addTencentCloudNamedWriteables(namedWriteables);
 
         addUnifiedNamedWriteables(namedWriteables);
 
@@ -229,6 +230,40 @@ private static void addDeepSeekNamedWriteables(List<NamedWriteableRegistry.Entry
         );
     }
 
+    private static void addTencentCloudNamedWriteables(List<NamedWriteableRegistry.Entry> namedWriteables) {
+        namedWriteables.add(
+            new NamedWriteableRegistry.Entry(
+                ServiceSettings.class,
+                TencentCloudEmbeddingsServiceSettings.NAME,
+                TencentCloudEmbeddingsServiceSettings::new
+            )
+        );
+        namedWriteables.add(
+            new NamedWriteableRegistry.Entry(
+                TaskSettings.class,
+                TencentCloudEmbeddingsTaskSettings.NAME,
+                TencentCloudEmbeddingsTaskSettings::new
+            )
+        );
+        namedWriteables.add(
+            new NamedWriteableRegistry.Entry(
+                ServiceSettings.class,
+                TencentCloudChatCompletionServiceSettings.NAME,
+                TencentCloudChatCompletionServiceSettings::new
+            )
+        );
+        namedWriteables.add(
+            new NamedWriteableRegistry.Entry(
+                ServiceSettings.class,
+                TencentCloudRerankServiceSettings.NAME,
+                TencentCloudRerankServiceSettings::new
+            )
+        );
+        namedWriteables.add(
+            new NamedWriteableRegistry.Entry(TaskSettings.class, TencentCloudRerankTaskSettings.NAME, TencentCloudRerankTaskSettings::new)
+        );
+    }
+
     private static void addGroqNamedWriteables(List<NamedWriteableRegistry.Entry> namedWriteables) {
         namedWriteables.add(
             new NamedWriteableRegistry.Entry(
@@ -978,13 +1013,6 @@ private static void addElasticNamedWriteables(List<NamedWriteableRegistry.Entry>
                 ElasticInferenceServiceCompletionServiceSettings::new
             )
         );
-        namedWriteables.add(
-            new NamedWriteableRegistry.Entry(
-                TaskSettings.class,
-                ElasticInferenceServiceChatCompletionTaskSettings.NAME,
-                ElasticInferenceServiceChatCompletionTaskSettings::new
-            )
-        );
 
         // Rerank
         namedWriteables.add(

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/InferencePlugin.java

@@ -204,6 +204,7 @@
 import org.elasticsearch.xpack.inference.services.sagemaker.model.SageMakerConfiguration;
 import org.elasticsearch.xpack.inference.services.sagemaker.model.SageMakerModelBuilder;
 import org.elasticsearch.xpack.inference.services.sagemaker.schema.SageMakerSchemas;
+import org.elasticsearch.xpack.inference.services.tencentcloud.TencentCloudService;
 import org.elasticsearch.xpack.inference.services.voyageai.VoyageAIService;
 import org.elasticsearch.xpack.inference.vectors.EmbeddingQueryVectorBuilder;
 
@@ -655,6 +656,7 @@ public List<InferenceServiceExtension.Factory> getInferenceServiceFactories() {
             context -> new JinaAIService(httpFactory.get(), serviceComponents.get(), context),
             context -> new VoyageAIService(httpFactory.get(), serviceComponents.get(), context),
             context -> new DeepSeekService(httpFactory.get(), serviceComponents.get(), context),
+            context -> new TencentCloudService(httpFactory.get(), serviceComponents.get(), context),
             context -> new LlamaService(httpFactory.get(), serviceComponents.get(), context),
             context -> new Ai21Service(httpFactory.get(), serviceComponents.get(), context),
             context -> new OpenShiftAiService(httpFactory.get(), serviceComponents.get(), context),