hack/tools/scripts/llm-key-hunter/results/_cache/content/bigchuidw3_spring-ai-agentx.json

{"repo": "bigchuidw3/spring-ai-agentx", "files": {"spring-ai-agentx-samples/src/main/java/com/agentx/ai/samples/TestConfig.java": {"74206edcf6fb207f25bbc52594e033d681f9f2d9": {"t": 1786068550.2060003, "n": 18306, "h": "cc7f7a49254e52b7", "c": "package com.agentx.ai.samples;\n\nimport com.agentx.ai.core.agent.ReactAgent;\nimport com.agentx.ai.core.chatmodels.DeepSeekV4ChatModel;\nimport com.agentx.ai.core.model.AgentStreamEvent;\nimport com.agentx.ai.core.model.PauseState;\nimport com.agentx.ai.core.model.RunnableParams;\nimport com.agentx.ai.core.model.SubAgentSource;\nimport com.zaxxer.hikari.HikariDataSource;\nimport org.springframework.ai.chat.model.ChatModel;\nimport org.springframework.ai.deepseek.DeepSeekChatOptions;\nimport org.springframework.ai.deepseek.api.DeepSeekApi;\nimport org.springframework.ai.document.MetadataMode;\nimport org.springframework.ai.embedding.EmbeddingModel;\nimport org.springframework.ai.openai.OpenAiChatModel;\nimport org.springframework.ai.openai.OpenAiChatOptions;\nimport org.springframework.ai.openai.OpenAiEmbeddingModel;\nimport org.springframework.ai.openai.OpenAiEmbeddingOptions;\nimport org.springframework.ai.openai.api.OpenAiApi;\nimport org.springframework.ai.zhipuai.ZhiPuAiChatModel;\nimport org.springframework.ai.zhipuai.ZhiPuAiChatOptions;\nimport org.springframework.ai.zhipuai.api.ZhiPuAiApi;\nimport org.springframework.ai.vectorstore.pgvector.PgVectorStore;\nimport org.springframework.http.client.ReactorClientHttpRequestFactory;\nimport org.springframework.jdbc.core.JdbcTemplate;\n\nimport org.springframework.http.client.reactive.ReactorClientHttpConnector;\nimport org.springframework.retry.policy.NeverRetryPolicy;\nimport org.springframework.retry.support.RetryTemplate;\nimport org.springframework.web.client.RestClient;\nimport org.springframework.web.reactive.function.client.WebClient;\nimport reactor.core.publisher.Flux;\nimport reactor.netty.http.client.HttpClient;\n\nimport javax.sql.DataSource;\nimport java.io.IOException;\nimport java.io.InputStream;\nimport java.time.Duration;\nimport java.util.HashMap;\nimport java.util.Map;\nimport java.util.Properties;\nimport java.util.UUID;\n\n/**\n * 测试公共配置 — 所有测试类共享的工厂方法和常量。\n *\n * @author bigchui\n *\n */\npublic final class TestConfig {\n\n    private static final Properties secrets = loadSecrets();\n\n    // 便捷访问方法\n    private static String s(String key) { return secrets.getProperty(key); }\n    private static String s(String key, String defaultValue) { return secrets.getProperty(key, defaultValue); }\n    private static int si(String key, int defaultValue) { return Integer.parseInt(s(key, String.valueOf(defaultValue))); }\n\n    // ===== 公开常量（用于测试打印）=====\n    public static final String CHAT_MODEL = s(\"dashscope.chat.model\", \"qwen-plus\");\n    public static final String EMBEDDING_MODEL = s(\"embedding.model\", \"text-embedding-v3\");\n    public static final String SKILLS_DIR = s(\"skills.dir\", \"\");\n\n    /** HTTP 响应超时（5 分钟），思考模型非流式调用需要较长等待 */\n    private static final Duration HTTP_TIMEOUT = Duration.ofSeconds(300);\n\n    private TestConfig() {\n    }\n\n    private static Properties loadSecrets() {\n        Properties props = new Properties();\n        try (InputStream is = TestConfig.class.getClassLoader().getResourceAsStream(\"secrets.properties\")) {\n            if (is != null) {\n                props.load(is);\n            } else {\n                System.err.println(\"[TestConfig] 未找到 secrets.properties，请复制 secrets.properties.example 并填入真实值\");\n            }\n        } catch (IOException e) {\n            throw new RuntimeException(\"Failed to load secrets.properties\", e);\n        }\n        return props;\n    }\n\n    // ---------- Factory Methods ----------\n\n    public static ChatModel createChatModel() {\n        OpenAiChatOptions opts = new OpenAiChatOptions();\n        opts.setModel(s(\"dashscope.chat.model\", \"qwen-plus\"));\n        opts.setTemperature(0.7);\n        Map<String,Object> map = new HashMap<>();\n        Map<String,Object> submap = new HashMap<>();\n        submap.put(\"enable_thinking\",true);\n        map.put(\"chat_template_kwargs\",submap);\n        opts.setExtraBody(map);\n\n        HttpClient httpClient = HttpClient.create().responseTimeout(HTTP_TIMEOUT);\n\n        return OpenAiChatModel.builder()\n                .openAiApi(OpenAiApi.builder()\n                        .baseUrl(s(\"dashscope.base.url\", \"https://dashscope.aliyuncs.com/compatible-mode/\"))\n                        .apiKey(s(\"dashscope.api.key\"))\n                        .restClientBuilder(RestClient.builder()\n                                .requestFactory(new ReactorClientHttpRequestFactory(httpClient)))\n                        .webClientBuilder(WebClient.builder()\n                                .clientConnector(new ReactorClientHttpConnector(httpClient)))\n                        .build())\n                .defaultOptions(opts)\n                .build();\n    }\n\n    public static ChatModel createMiniMaxChatModel() {\n        OpenAiChatOptions opts = new OpenAiChatOptions();\n        opts.setModel(s(\"minimax.chat.model\", \"MiniMax-M2.7\"));\n        opts.setTemperature(0.7);\n\n        HttpClient httpClient = HttpClient.create().responseTimeout(HTTP_TIMEOUT);\n\n        RetryTemplate noRetryTemplate = new RetryTemplate();\n        noRetryTemplate.setRetryPolicy(new NeverRetryPolicy());\n\n        return OpenAiChatModel.builder()\n                .openAiApi(OpenAiApi.builder()\n                        .baseUrl(s(\"minimax.base.url\", \"https://api.minimaxi.com/\"))\n                        .apiKey(s(\"minimax.api.key\"))\n                        .restClientBuilder(RestClient.builder()\n                                .requestFactory(new ReactorClientHttpRequestFactory(httpClient)))\n                        .webClientBuilder(WebClient.builder()\n                                .clientConnector(new ReactorClientHttpConnector(httpClient)))\n                        .build())\n                .defaultOptions(opts)\n                .retryTemplate(noRetryTemplate)\n                .build();\n    }\n\n    /**\n     * 错误用法，deepseek v4不兼容openai chatmodel\n     * @return\n     */\n    public static ChatModel createDeepSeekChatModel() {\n        OpenAiChatOptions opts = new OpenAiChatOptions();\n        opts.setModel(s(\"deepseek.chat.model\", \"deepseek-chat\"));\n        opts.setTemperature(0.7);\n\n        HttpClient httpClient = HttpClient.create().responseTimeout(HTTP_TIMEOUT);\n\n        RetryTemplate noRetryTemplate = new RetryTemplate();\n        noRetryTemplate.setRetryPolicy(new NeverRetryPolicy());\n\n        return OpenAiChatModel.builder()\n                .openAiApi(OpenAiApi.builder()\n                        .baseUrl(s(\"deepseek.base.url\", \"https://api.deepseek.com\"))\n                        .apiKey(s(\"deepseek.api.key\"))\n                        .restClientBuilder(RestClient.builder()\n                                .requestFactory(new ReactorClientHttpRequestFactory(httpClient)))\n                        .webClientBuilder(WebClient.builder()\n                                .clientConnector(new ReactorClientHttpConnector(httpClient)))\n                        .build())\n                .defaultOptions(opts)\n                .retryTemplate(noRetryTemplate)\n                .build();\n    }\n\n    /**\n     * 正确用法\n     * @return\n     */\n    public static ChatModel createDeepSeekV4ChatModel(){\n        DeepSeekApi deepSeekApi = DeepSeekApi.builder()\n                .apiKey(s(\"deepseek.api.key\"))\n                .baseUrl(s(\"deepseek.base.url\"))\n                .build();\n        DeepSeekChatOptions options = DeepSeekChatOptions.builder()\n                .model(s(\"deepseek.chat.model\"))\n                .temperature(0.7)\n                .build();\n        DeepSeekV4ChatModel chatModel = DeepSeekV4ChatModel.builder()\n                .deepSeekApi(deepSeekApi)\n                .defaultOptions(options)\n                .build();\n        return chatModel;\n    }\n\n    public static ChatModel createZhiPuChatModel() {\n        ZhiPuAiApi api = ZhiPuAiApi.builder()\n                .baseUrl(\"https://open.bigmodel.cn/api/coding/paas/\")\n                .apiKey(s(\"zhipu.api.key\"))\n                .build();\n\n        ZhiPuAiChatOptions options = ZhiPuAiChatOptions.builder()\n                .model(s(\"zhipu.chat.model\", \"glm-4.7\"))\n                .temperature(0.7)\n                .thinking(ZhiPuAiApi.ChatCompletionRequest.Thinking.enabled())\n                .build();\n\n        // 禁用 Spring AI 内部重试，由 AgentX 框架统一处理重试\n        RetryTemplate noRetryTemplate = new RetryTemplate();\n        noRetryTemplate.setRetryPolicy(new NeverRetryPolicy());\n\n        return new ZhiPuAiChatModel(api, options, noRetryTemplate);\n    }\n\n    /**\n     * 创建 OpenAI 兼容风格的 ZhiPu ChatModel。\n     * <p>\n     * 通过 OpenAI 兼容接口访问 GLM 模型，支持 {@code reasoning_content} 字段提取。\n     * <p>\n     * 与 {@link #createZhiPuChatModel()} 的区别：\n     * <ul>\n     *   <li>原生 ZhiPu 实现 — 不提取 reasoning_content，无法使用 REASONING_CONTENT 模式</li>\n     *   <li>OpenAI 兼容接口 — 正确提取 reasoning_content 到 metadata，支持 REASONING_CONTENT 模式</li>\n     * </ul>\n     */\n    public static ChatModel createZhiPuOpenAiChatModel() {\n        OpenAiChatOptions opts = new OpenAiChatOptions();\n        opts.setModel(s(\"zhipu.chat.model\", \"glm-4.7\"));\n        opts.setTemperature(0.7);\n\n        // 思考模型非流式调用需要更长超时（模型生成完整响应后才开始发送）\n        HttpClient httpClient = HttpClient.create()\n                .responseTimeout(HTTP_TIMEOUT);\n\n        RestClient.Builder restClientBuilder = RestClient.builder()\n                .requestFactory(new ReactorClientHttpRequestFactory(httpClient));\n        WebClient.Builder webClientBuilder = WebClient.builder()\n                .clientConnector(new ReactorClientHttpConnector(httpClient));\n\n        RetryTemplate noRetryTemplate = new RetryTemplate();\n        noRetryTemplate.setRetryPolicy(new NeverRetryPolicy());\n\n        return OpenAiChatModel.builder()\n                .openAiApi(OpenAiApi.builder()\n                        .baseUrl(s(\"zhipu.openai.base.url\", \"https://open.bigmodel.cn/api/coding/paas/v4\"))\n                        .apiKey(s(\"zhipu.api.key\"))\n                        .completionsPath(\"/chat/completions\")\n                        .restClientBuilder(restClientBuilder)\n                        .webClientBuilder(webClientBuilder)\n                        .build())\n                .defaultOptions(opts)\n                .retryTemplate(noRetryTemplate)\n                .build();\n    }\n\n    public static EmbeddingModel createEmbeddingModel() {\n        OpenAiEmbeddingOptions options = OpenAiEmbeddingOptions.builder()\n                .model(EMBEDDING_MODEL)\n                .dimensions(1024)\n                .build();\n\n        return new OpenAiEmbeddingModel(\n                OpenAiApi.builder()\n                        .baseUrl(s(\"dashscope.base.url\", \"https://dashscope.aliyuncs.com/compatible-mode/\"))\n                        .apiKey(s(\"dashscope.api.key\"))\n                        .build(),\n                MetadataMode.EMBED,\n                options\n        );\n    }\n\n    /**\n     * 创建使用错误 API Key 的 ChatModel，用于测试 ErrorEvent。\n     */\n    public static ChatModel createBadApiKeyChatModel() {\n        DeepSeekApi deepSeekApi = DeepSeekApi.builder()\n                .apiKey(\"sk-bad-api-key-for-testing-error-event\")\n                .baseUrl(s(\"deepseek.base.url\", \"https://api.deepseek.com\"))\n                .build();\n        DeepSeekChatOptions options = DeepSeekChatOptions.builder()\n                .model(s(\"deepseek.chat.model\", \"deepseek-chat\"))\n                .temperature(0.7)\n                .build();\n        return DeepSeekV4ChatModel.builder()\n                .deepSeekApi(deepSeekApi)\n                .defaultOptions(options)\n                .build();\n    }\n\n    public static DataSource createMySqlDataSource() {\n        HikariDataSource ds = new HikariDataSource();\n        ds.setJdbcUrl(s(\"mysql.url\", \"jdbc:mysql://localhost:3306/agent_test?useSSL=false&allowPublicKeyRetrieval=true\"));\n        ds.setUsername(s(\"mysql.user\", \"root\"));\n        ds.setPassword(s(\"mysql.password\", \"root\"));\n        ds.setDriverClassName(\"com.mysql.cj.jdbc.Driver\");\n        return ds;\n    }\n\n    public static DataSource createPgDataSource() {\n        HikariDataSource ds = new HikariDataSource();\n        ds.setJdbcUrl(String.format(\"jdbc:postgresql://%s:%d/%s\",\n                s(\"pg.host\", \"192.168.11.163\"),\n                si(\"pg.port\", 5433),\n                s(\"pg.database\", \"vector_store\")));\n        ds.setUsername(s(\"pg.user\", \"postgres\"));\n        ds.setPassword(s(\"pg.password\", \"postgres\"));\n        ds.setDriverClassName(\"org.postgresql.Driver\");\n        return ds;\n    }\n\n    public static PgVectorStore createPgVectorStore(DataSource pgDataSource, EmbeddingModel embeddingModel) {\n        JdbcTemplate jdbcTemplate = new JdbcTemplate(pgDataSource);\n        PgVectorStore store = PgVectorStore.builder(jdbcTemplate, embeddingModel)\n                .dimensions(1024)\n                .distanceType(PgVectorStore.PgDistanceType.COSINE_DISTANCE)\n                .indexType(PgVectorStore.PgIndexType.HNSW)\n                .initializeSchema(true)\n                .removeExistingVectorStoreTable(false)\n                .vectorTableName(s(\"pg.table.name\", \"agentx_long_term_memory\"))\n                .maxDocumentBatchSize(100)\n                .build();\n        try {\n            store.afterPropertiesSet();\n        } catch (Exception e) {\n            throw new RuntimeException(\"Failed to initialize PgVectorStore\", e);\n        }\n        return store;\n    }\n\n    // ---------- Helper Methods ----------\n\n    public static String randomConvId() {\n        return \"conv_\" + UUID.randomUUID();\n    }\n\n    public static String randomUserId(String prefix) {\n        return prefix + \"_\" + System.currentTimeMillis();\n    }\n\n    public static RunnableParams buildParams(String convId, String userId) {\n        return RunnableParams.builder()\n                .conversationId(convId)\n                .userId(userId)\n                .build();\n    }\n\n    /**\n     * 流式调用 Agent 并实时打印输出。\n     */\n    public static void streamAndPrint(ReactAgent agent, String query) {\n        streamAndPrint(agent, query, null);\n    }\n\n    public static void streamAndPrint(ReactAgent agent, String query, RunnableParams params) {\n        System.out.println(\"Q: \" + query);\n        System.out.print(\"A: \");\n        if (params != null) {\n            agent.stream(query, params)\n                    .doOnNext(chunk -> {\n                        System.out.print(chunk);\n                        System.out.flush();\n                    })\n                    .doOnError(err -> System.err.println(\"\\nError: \" + err.getMessage()))\n                    .blockLast();\n        } else {\n            agent.stream(query)\n                    .doOnNext(chunk -> {\n                        System.out.print(chunk);\n                        System.out.flush();\n                    })\n                    .doOnError(err -> System.err.println(\"\\nError: \" + err.getMessage()))\n                    .blockLast();\n        }\n        System.out.println(\"\\n\");\n    }\n\n    public static void printTestHeader(String title) {\n        System.out.println(\"\\n\" + \"=\".repeat(60));\n        System.out.println(\"  \" + title);\n        System.out.println(\"=\".repeat(60) + \"\\n\");\n    }\n\n    // ---------- 流式事件工具方法 ----------\n\n    /**\n     * 上一个事件是否为 Thinking，用于控制标记打印\n     */\n    private static boolean lastEventWasThinking = false;\n\n    /**\n     * 当前事件来源的 Agent 名称（null = 主 Agent），用于 SubAgent 分隔\n     */\n    private static String currentAgentName = null;\n\n    /**\n     * 从事件中提取 SubAgentSource。\n     */\n    private static SubAgentSource extractSource(AgentStreamEvent event) {\n        return switch (event) {\n            case AgentStreamEvent.Text e -> e.source();\n            case AgentStreamEvent.Thinking e -> e.source();\n            case AgentStreamEvent.ToolStart e -> e.source();\n            case AgentStreamEvent.ToolEnd e -> e.source();\n            case AgentStreamEvent.Complete e -> e.source();\n            default -> null;\n        };\n    }\n\n    /**\n     * 格式化打印单个流式事件。\n     * <p>\n     * - SubAgent 来源切换时打印 ━━━ [agent名] ━━━ 分隔线\n     * - Thinking 事件只在进入思考时打印一次标记，避免每个 chunk 重复前缀\n     */\n    public static void printEvent(AgentStreamEvent event) {\n        // SubAgent 来源切换检测\n        SubAgentSource source = extractSource(event);\n        String agentName = source != null ? source.agentName() : null;\n        if (!java.util.Objects.equals(agentName, currentAgentName)) {\n            if (currentAgentName != null) System.out.println();\n            currentAgentName = agentName;\n            lastEventWasThinking = false;\n            String label = agentName != null ? agentName : \"main\";\n            System.out.println(\"\\n━━━ [\" + label + \"] ━━━\");\n        }\n\n        switch (event) {\n            case AgentStreamEvent.Thinking t -> {\n                if (!lastEventWasThinking) {\n                    System.out.print(\"\\n[Thinking] \");\n                }\n                System.out.print(t.content());\n                lastEventWasThinking = true;\n            }\n            case AgentStreamEvent.Text t -> {\n                if (lastEventWasThinking) {\n                    System.out.println(\"\\n---\");\n                    System.out.println(\"\\n最终答案：\");\n                }\n                System.out.print(t.content());\n                lastEventWasThinking = false;\n            }\n            case AgentStreamEvent.ToolStart ts ->\n                    System.out.println(\"\\n[ToolStart] \" + ts.toolName() + \" id=\" + ts.toolCallId());\n            case AgentStreamEvent.ToolEnd te -> System.out.println(\"[ToolEnd] \" + te.toolName() + \" result=\"\n                    + (te.result() != null && te.result().length() > 80\n                    ? te.result().substring(0, 80) + \"...\"\n                    : te.result()));\n            case AgentStreamEvent.Error e ->\n                    System.out.println(\"\\n[Error:\" + e.code().code() + \"] \" + e.message() + \"\\n  \" + e.detail());\n            case AgentStreamEvent.Complete c -> System.out.println(\"\\n[Complete]：promptTokens：\"+c.totalPromptTokens()+ \"，completionTokens：\"+c.totalCompletionTokens());\n            case AgentStreamEvent.Paused p -> System.out.println(\"[Paused]\");\n        }\n    }\n\n    /**\n     * 收集流式事件，实时打印所有事件，返回 PauseState（如果暂停）。\n     */\n    public static PauseState collectStreamEvents(Flux<AgentStreamEvent> flux) {\n        PauseState[] holder = new PauseState[1];\n        flux.doOnNext(event -> {\n            if (event instanceof AgentStreamEvent.Paused p) {\n                holder[0] = p.state();\n            } else {\n                printEvent(event);\n            }\n        }).blockLast();\n        return holder[0];\n    }\n}\n"}}}}