diff --git a/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/agent/AgentRuntime.java b/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/agent/AgentRuntime.java
index 34c520c8..e2c4dc3a 100644
--- a/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/agent/AgentRuntime.java
+++ b/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/agent/AgentRuntime.java
@@ -1,5 +1,6 @@
 package cn.com.mfish.ai.agent;
 
+import cn.com.mfish.ai.service.FileParseService;
 import cn.com.mfish.ai.service.LlmModelRouter;
 import cn.com.mfish.common.ai.agent.EventBus;
 import cn.com.mfish.common.ai.agent.TenantContext;
@@ -10,6 +11,7 @@ import cn.com.mfish.common.ai.entity.EventType;
 import cn.com.mfish.common.ai.entity.PlanStep;
 import cn.com.mfish.common.core.utils.AuthInfoUtils;
 import cn.com.mfish.common.core.utils.ServletUtils;
+import cn.com.mfish.common.core.utils.StringUtils;
 import lombok.extern.slf4j.Slf4j;
 import org.springframework.stereotype.Component;
 import org.springframework.web.context.request.RequestAttributes;
@@ -48,11 +50,14 @@ public class AgentRuntime {
     private final Planner planner;
     private final Executor executor;
     private final LlmModelRouter llmModelRouter;
+    private final FileParseService fileParseService;
 
-    public AgentRuntime(Planner planner, Executor executor, LlmModelRouter llmModelRouter) {
+    public AgentRuntime(Planner planner, Executor executor, LlmModelRouter llmModelRouter,
+                        FileParseService fileParseService) {
         this.planner = planner;
         this.executor = executor;
         this.llmModelRouter = llmModelRouter;
+        this.fileParseService = fileParseService;
     }
 
     /**
@@ -67,11 +72,16 @@ public class AgentRuntime {
      * 需要用快照构建 ToolContext 和 ChatClient。
      * </p>
      * <p>
+     * 文件解析：若请求携带 fileIds，在请求线程同步加载文件内容（Feign 调用 mf-storage），
+     * 拼接到用户提示词前。必须在请求线程执行，因为 Feign 的 BearerTokenInterceptor 依赖
+     * RequestContextHolder 中继令牌，异步线程拿不到。
+     * </p>
+     * <p>
      * 所有返回的 {@link ChatResponseVo} 的 id 字段填充为 {@link AiRequest#getId()}，
      * 与普通聊天返回结构保持一致，前端可按 id 关联请求与响应。
      * </p>
      *
-     * @param aiRequest AI请求参数（含 id/sessionId/message）
+     * @param aiRequest AI请求参数（含 id/sessionId/message/fileIds）
      * @return 聊天响应流（与普通聊天统一结构）
      */
     public Flux<ChatResponseVo> run(AiRequest aiRequest) {
@@ -84,6 +94,9 @@ public class AgentRuntime {
         // 在请求线程捕获租户上下文快照，供异步编排使用
         TenantContext tenantContext = captureTenantContext();
 
+        // 文件解析必须在请求线程执行：Feign BearerTokenInterceptor 依赖 RequestContextHolder
+        prompt = resolveFileContents(prompt, aiRequest.getFileIds());
+
         // 后台启动编排流程
         runOrchestration(sessionId, prompt, eventBus, tenantContext);
 
@@ -91,6 +104,31 @@ public class AgentRuntime {
         return eventBus.asFlux();
     }
 
+    /**
+     * 加载文件内容并拼接到提示词前
+     * <p>
+     * 若 fileIds 为空或全部加载失败，返回原始 prompt。
+     * 文件内容加载失败时记录日志但不阻断流程，降级为纯文本对话。
+     * </p>
+     *
+     * @param prompt  原始用户提示词
+     * @param fileIds 文件fileKey列表
+     * @return 拼接后的提示词
+     */
+    private String resolveFileContents(String prompt, List<String> fileIds) {
+        if (fileIds == null || fileIds.isEmpty()) {
+            return prompt;
+        }
+        String fileContents = fileParseService.loadFileContents(fileIds);
+        if (StringUtils.isEmpty(fileContents)) {
+            log.warn("[AgentRuntime] 文件内容加载为空 fileIds={}", fileIds);
+            return prompt;
+        }
+        return "以下是用户上传的文件内容，请基于文件内容进行分析：\n\n"
+                + fileContents
+                + "\n用户需求：" + prompt;
+    }
+
     /**
      * 在请求线程捕获租户上下文快照
      * <p>
diff --git a/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/agent/BaseAssistant.java b/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/agent/BaseAssistant.java
index 0fae7126..198a1001 100644
--- a/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/agent/BaseAssistant.java
+++ b/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/agent/BaseAssistant.java
@@ -1,5 +1,6 @@
 package cn.com.mfish.ai.agent;
 
+import cn.com.mfish.ai.service.FileParseService;
 import cn.com.mfish.ai.service.LlmModelRouter;
 import cn.com.mfish.common.ai.client.IClientAssistant;
 import cn.com.mfish.common.ai.engine.ApiToolEngine;
@@ -17,6 +18,7 @@ import org.springframework.ai.chat.client.advisor.SimpleLoggerAdvisor;
 import org.springframework.ai.chat.memory.ChatMemory;
 import org.springframework.ai.chat.model.ChatResponse;
 import org.springframework.ai.tool.ToolCallbackProvider;
+import org.springframework.beans.factory.annotation.Autowired;
 import org.springframework.web.context.request.RequestAttributes;
 import org.springframework.web.context.request.RequestContextHolder;
 import org.springframework.web.server.ServerWebExchange;
@@ -26,6 +28,7 @@ import reactor.core.scheduler.Schedulers;
 
 import java.util.Collections;
 import java.util.HashMap;
+import java.util.List;
 import java.util.Map;
 import java.util.Objects;
 import java.util.Set;
@@ -51,6 +54,13 @@ public abstract class BaseAssistant implements IClientAssistant, ToolCapable {
     private final LlmModelRouter llmModelRouter;
     protected final ApiToolEngine apiToolEngine;
 
+    /**
+     * 文件解析服务：用于根据AiRequest.fileIds获取文件内容并注入提示词
+     * 采用字段注入避免修改所有子类构造函数
+     */
+    @Autowired
+    protected FileParseService fileParseService;
+
     public BaseAssistant(ChatMemory chatMemory, LlmModelRouter llmModelRouter, ApiToolEngine apiToolEngine) {
         this.llmModelRouter = llmModelRouter;
         this.chatMemory = chatMemory;
@@ -268,12 +278,26 @@ public abstract class BaseAssistant implements IClientAssistant, ToolCapable {
 
     /**
      * 聊天返回id
+     * <p>
+     * 若请求携带fileIds，会先通过FileParseService从文件服务获取文件内容，
+     * 将其拼接到用户提示词前，再交由子类chat(sessionId, prompt)处理。
      *
      * @return 聊天信息
      */
     @Override
     public Flux<ChatResponseVo> chat(AiRequest aiRequest) {
-        return chat(aiRequest.getSessionId(), aiRequest.getMessage().getContent())
+        String prompt = aiRequest.getMessage().getContent();
+        List<String> fileIds = aiRequest.getFileIds();
+        if (fileIds != null && !fileIds.isEmpty()) {
+            String fileContents = fileParseService.loadFileContents(fileIds);
+            if (StringUtils.isNotEmpty(fileContents)) {
+                prompt = "以下是用户上传的文件内容，请基于文件内容进行分析：\n\n"
+                        + fileContents
+                        + "\n用户问题：" + prompt;
+            }
+        }
+        final String finalPrompt = prompt;
+        return chat(aiRequest.getSessionId(), finalPrompt)
                 .filter(resp -> "STOP".equals(Objects.requireNonNull(resp.getResult()).getMetadata().getFinishReason())
                         || StringUtils.isNotEmpty(resp.getResult().getOutput().getText()))
                 .map(resp -> new ChatResponseVo().setId(aiRequest.getId())
diff --git a/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/service/FileParseService.java b/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/service/FileParseService.java
new file mode 100644
index 00000000..bb0d7fec
--- /dev/null
+++ b/mf-business/mf-ai/src/main/java/cn/com/mfish/ai/service/FileParseService.java
@@ -0,0 +1,313 @@
+package cn.com.mfish.ai.service;
+
+import cn.com.mfish.common.core.constants.RPCConstants;
+import cn.com.mfish.common.core.utils.StringUtils;
+import cn.com.mfish.common.storage.api.entity.StorageInfo;
+import cn.com.mfish.common.storage.api.remote.RemoteStorageService;
+import cn.com.mfish.common.core.web.Result;
+import lombok.extern.slf4j.Slf4j;
+import org.apache.poi.extractor.ExtractorFactory;
+import org.apache.poi.extractor.POITextExtractor;
+import org.springframework.core.io.Resource;
+import org.springframework.http.ResponseEntity;
+import org.springframework.stereotype.Service;
+
+import java.io.InputStream;
+import java.nio.charset.StandardCharsets;
+import java.util.List;
+import java.util.Locale;
+import java.util.Set;
+
+/**
+ * 文件解析服务
+ * <p>
+ * 根据文件fileKey列表，通过远程存储服务获取文件元数据与内容，
+ * 将文件内容提取为LLM可理解的提示词片段。
+ * <p>
+ * 支持的文件类型：
+ * <ul>
+ *   <li>文本类：txt/log/md/json/xml/yml/csv/sql/源代码等（直接UTF-8读取）</li>
+ *   <li>Office文档（仅OOXML新格式）：docx/xlsx/pptx（通过Apache POI提取文本）</li>
+ *   <li>其他二进制：图片/压缩包/旧版Office（.doc/.xls/.ppt）等（返回提示信息，不解析内容）</li>
+ * </ul>
+ * <p>
+ * 注意：旧版OLE2格式（.doc/.xls/.ppt）需要 poi-scratchpad 依赖，当前未引入，
+ * 前端应限制只允许上传 .docx/.xlsx/.pptx 格式。
+ * <p>
+ * 调用链：FileParseService → RemoteStorageService(Feign) → mf-storage 服务
+ * BearerTokenInterceptor 自动中继用户令牌，保证私有文件可访问。
+ *
+ * @author: mfish
+ * @date: 2026/07/20
+ */
+@Slf4j
+@Service
+public class FileParseService {
+    /**
+     * 单个文件读取的最大字符数，超过则截断，避免撑爆LLM上下文
+     */
+    private static final int MAX_CONTENT_CHARS = 51200;
+    /**
+     * 文本类contentType前缀集合，命中时按文本读取
+     */
+    private static final Set<String> TEXT_CONTENT_TYPE_PREFIXES = Set.of("text/");
+    /**
+     * 文本类contentType精确匹配集合
+     */
+    private static final Set<String> TEXT_CONTENT_TYPES = Set.of(
+            "application/json",
+            "application/xml",
+            "application/javascript",
+            "application/x-yaml",
+            "application/yaml",
+            "application/x-sh",
+            "application/sql",
+            "application/csv",
+            "application/x-java",
+            "application/x-tex",
+            "application/manifest+json",
+            "application/graphql-response+json"
+    );
+    /**
+     * 文本类文件后缀集合（按扩展名兜底识别）
+     */
+    private static final Set<String> TEXT_SUFFIXES = Set.of(
+            ".txt", ".log", ".md", ".markdown", ".json", ".xml", ".yaml", ".yml",
+            ".csv", ".sql", ".js", ".ts", ".java", ".py", ".go", ".rs", ".c",
+            ".cpp", ".h", ".hpp", ".cs", ".rb", ".php", ".swift", ".kt", ".scala",
+            ".sh", ".bat", ".ps1", ".properties", ".conf", ".ini", ".toml",
+            ".html", ".htm", ".css", ".scss", ".less", ".vue", ".tsx", ".jsx",
+            ".gradle", ".gitignore", ".dockerfile", ".env", ".proto"
+    );
+    /**
+     * Office文档后缀集合（仅OOXML新格式，通过Apache POI解析）
+     * <p>
+     * 旧版OLE2格式（.doc/.xls/.ppt）需要 poi-scratchpad 依赖，当前未引入，
+     * 前端应限制只允许上传新格式。
+     */
+    private static final Set<String> OFFICE_SUFFIXES = Set.of(
+            ".docx", ".xlsx", ".pptx", ".docm", ".dotm", ".xlsm", ".xltm", ".pptm", ".potm"
+    );
+    /**
+     * Office文档contentType精确匹配集合（仅OOXML新格式）
+     */
+    private static final Set<String> OFFICE_CONTENT_TYPES = Set.of(
+            "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+            "application/vnd.openxmlformats-officedocument.wordprocessingml.template",
+            "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+            "application/vnd.openxmlformats-officedocument.spreadsheetml.template",
+            "application/vnd.openxmlformats-officedocument.presentationml.presentation",
+            "application/vnd.openxmlformats-officedocument.presentationml.template"
+    );
+
+    private final RemoteStorageService remoteStorageService;
+
+    public FileParseService(RemoteStorageService remoteStorageService) {
+        this.remoteStorageService = remoteStorageService;
+    }
+
+    /**
+     * 加载文件内容并拼接为提示词片段
+     * <p>
+     * 遍历fileKey列表，逐个获取文件元数据与内容，将文本/Office文档内容拼接到StringBuilder。
+     * 单个文件读取失败不影响其他文件，异常会被记录日志并跳过。
+     *
+     * @param fileIds 文件fileKey列表
+     * @return 拼接好的文件内容提示词片段；列表为空或全部失败时返回空字符串
+     */
+    public String loadFileContents(List<String> fileIds) {
+        if (fileIds == null || fileIds.isEmpty()) {
+            return "";
+        }
+        StringBuilder sb = new StringBuilder();
+        int successCount = 0;
+        for (String fileKey : fileIds) {
+            if (StringUtils.isEmpty(fileKey)) {
+                continue;
+            }
+            try {
+                String segment = loadOneFile(fileKey);
+                if (segment != null && !segment.isEmpty()) {
+                    sb.append(segment).append("\n\n");
+                    successCount++;
+                }
+            } catch (Exception e) {
+                log.warn("加载文件内容失败 fileKey={} reason={}", fileKey, e.getMessage());
+            }
+        }
+        if (successCount == 0) {
+            return "";
+        }
+        return sb.toString();
+    }
+
+    /**
+     * 加载单个文件并构建内容片段
+     * <p>
+     * 按文件类型分发到不同解析路径：
+     * <ol>
+     *   <li>文本类：直接UTF-8读取</li>
+     *   <li>Office文档：通过POI提取文本</li>
+     *   <li>其他二进制：返回提示信息</li>
+     * </ol>
+     *
+     * @param fileKey 文件key
+     * @return 文件内容片段；文件不存在返回null
+     */
+    private String loadOneFile(String fileKey) {
+        Result<StorageInfo> infoResult = remoteStorageService.queryByKey(RPCConstants.INNER, fileKey);
+        if (infoResult == null || !infoResult.isSuccess() || infoResult.getData() == null) {
+            log.warn("文件信息查询失败 fileKey={} msg={}", fileKey,
+                    infoResult == null ? "result is null" : infoResult.getMsg());
+            return null;
+        }
+        StorageInfo storageInfo = infoResult.getData();
+        String fileName = StringUtils.isEmpty(storageInfo.getFileName()) ? fileKey : storageInfo.getFileName();
+        String fileType = storageInfo.getFileType();
+
+        if (isTextFile(fileType, fileName)) {
+            String content = readFileText(fileKey);
+            return content == null ? buildReadFailureHint(fileName) : buildTextFileSegment(fileName, truncate(content));
+        }
+        if (isOfficeDocument(fileType, fileName)) {
+            String content = extractOfficeText(fileKey, fileName);
+            return content == null ? buildReadFailureHint(fileName) : buildTextFileSegment(fileName, truncate(content));
+        }
+        return buildBinaryFileHint(fileName, fileType, storageInfo.getFileSize());
+    }
+
+    /**
+     * 通过Feign获取文件资源并读取为UTF-8文本
+     *
+     * @param fileKey 文件key
+     * @return 文件文本内容；读取失败返回null
+     */
+    private String readFileText(String fileKey) {
+        ResponseEntity<Resource> resp = remoteStorageService.fetch(RPCConstants.INNER, fileKey);
+        if (resp == null || !resp.getStatusCode().is2xxSuccessful() || resp.getBody() == null) {
+            log.warn("文件资源获取失败 fileKey={} status={}", fileKey,
+                    resp == null ? "null" : resp.getStatusCode());
+            return null;
+        }
+        try (InputStream is = resp.getBody().getInputStream()) {
+            return new String(is.readAllBytes(), StandardCharsets.UTF_8);
+        } catch (Exception e) {
+            log.warn("文件内容读取失败 fileKey={} reason={}", fileKey, e.getMessage());
+            return null;
+        }
+    }
+
+    /**
+     * 通过Apache POI提取Office文档文本
+     * <p>
+     * 使用 {@link ExtractorFactory} 自动识别文档类型（Word/Excel/PowerPoint，旧版/新版格式），
+     * 无需根据后缀手动分发，POI内部根据文件魔数选择合适的Extractor。
+     * <p>
+     * 注意：POI要求InputStream支持mark/reset，Feign返回的InputStream需先缓存为字节数组再包装。
+     *
+     * @param fileKey  文件key
+     * @param fileName 文件名（仅用于日志）
+     * @return 提取的文本内容；提取失败返回null
+     */
+    private String extractOfficeText(String fileKey, String fileName) {
+        ResponseEntity<Resource> resp = remoteStorageService.fetch(RPCConstants.INNER, fileKey);
+        if (resp == null || !resp.getStatusCode().is2xxSuccessful() || resp.getBody() == null) {
+            log.warn("Office文件资源获取失败 fileKey={} status={}", fileKey,
+                    resp == null ? "null" : resp.getStatusCode());
+            return null;
+        }
+        // POI的ExtractorFactory.createExtractor需要File或mark/reset支持的InputStream
+        // ByteArrayInputStream支持mark/reset，避免POI内部reset失败
+        try (InputStream is = resp.getBody().getInputStream()) {
+            byte[] bytes = is.readAllBytes();
+            try (InputStream poiStream = new java.io.ByteArrayInputStream(bytes);
+                 POITextExtractor extractor = ExtractorFactory.createExtractor(poiStream)) {
+                String text = extractor.getText();
+                log.info("Office文档解析成功 fileName={} chars={}", fileName,
+                        text == null ? 0 : text.length());
+                return text;
+            }
+        } catch (Exception e) {
+            log.warn("Office文档解析失败 fileKey={} fileName={} reason={}", fileKey, fileName, e.getMessage());
+            return null;
+        }
+    }
+
+    /**
+     * 截断过长的内容，避免撑爆LLM上下文
+     */
+    private String truncate(String content) {
+        if (content == null) {
+            return "";
+        }
+        if (content.length() > MAX_CONTENT_CHARS) {
+            return content.substring(0, MAX_CONTENT_CHARS)
+                    + "\n...[文件内容过长，已截断，仅显示前" + MAX_CONTENT_CHARS + "字符]";
+        }
+        return content;
+    }
+
+    /**
+     * 判断文件是否为文本类文件
+     * <p>
+     * 优先按contentType判断，再按文件名后缀兜底
+     *
+     * @param contentType 文件MIME类型
+     * @param fileName    文件名
+     * @return true表示可按文本读取
+     */
+    private boolean isTextFile(String contentType, String fileName) {
+        if (StringUtils.isNotEmpty(contentType)) {
+            String lower = contentType.toLowerCase(Locale.ROOT);
+            if (TEXT_CONTENT_TYPE_PREFIXES.stream().anyMatch(lower::startsWith)) {
+                return true;
+            }
+            if (TEXT_CONTENT_TYPES.contains(lower)) {
+                return true;
+            }
+        }
+        if (StringUtils.isNotEmpty(fileName)) {
+            String lowerName = fileName.toLowerCase(Locale.ROOT);
+            return TEXT_SUFFIXES.stream().anyMatch(lowerName::endsWith);
+        }
+        return false;
+    }
+
+    /**
+     * 判断文件是否为Office文档
+     * <p>
+     * 优先按contentType判断，再按文件名后缀兜底
+     *
+     * @param contentType 文件MIME类型
+     * @param fileName    文件名
+     * @return true表示为Office文档（Word/Excel/PowerPoint）
+     */
+    private boolean isOfficeDocument(String contentType, String fileName) {
+        if (StringUtils.isNotEmpty(contentType)) {
+            String lower = contentType.toLowerCase(Locale.ROOT);
+            if (OFFICE_CONTENT_TYPES.contains(lower)) {
+                return true;
+            }
+        }
+        if (StringUtils.isNotEmpty(fileName)) {
+            String lowerName = fileName.toLowerCase(Locale.ROOT);
+            return OFFICE_SUFFIXES.stream().anyMatch(lowerName::endsWith);
+        }
+        return false;
+    }
+
+    private String buildTextFileSegment(String fileName, String content) {
+        return "=== 文件: " + fileName + " ===\n" + content + "\n=== 文件结束: " + fileName + " ===";
+    }
+
+    private String buildBinaryFileHint(String fileName, String fileType, Integer fileSize) {
+        return "=== 文件: " + fileName + " ===\n" +
+                "[二进制文件，无法直接作为文本解析] 类型: " + (StringUtils.isEmpty(fileType) ? "未知" : fileType) +
+                " 大小: " + (fileSize == null ? "未知" : fileSize + " bytes") +
+                "\n=== 文件结束: " + fileName + " ===";
+    }
+
+    private String buildReadFailureHint(String fileName) {
+        return "=== 文件: " + fileName + " ===\n[文件内容读取失败]\n=== 文件结束: " + fileName + " ===";
+    }
+}
diff --git a/mf-common/mf-common-ai/src/main/java/cn/com/mfish/common/ai/entity/AiRequest.java b/mf-common/mf-common-ai/src/main/java/cn/com/mfish/common/ai/entity/AiRequest.java
index c0f97602..9bc25a65 100644
--- a/mf-common/mf-common-ai/src/main/java/cn/com/mfish/common/ai/entity/AiRequest.java
+++ b/mf-common/mf-common-ai/src/main/java/cn/com/mfish/common/ai/entity/AiRequest.java
@@ -3,6 +3,8 @@ package cn.com.mfish.common.ai.entity;
 import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.Data;
 
+import java.util.List;
+
 /**
  * @description: Ai请求
  * @author: mfish
@@ -17,4 +19,6 @@ public class AiRequest {
     private String sessionId;
     @Schema(description = "消息")
     private AiMessage message;
+    @Schema(description = "文件ID列表(文件fileKey集合，AI将获取文件内容并解析)")
+    private List<String> fileIds;
 }