新增: F-05多模态Phase2a后端(ContentPart+parts字段+provider图片块+truncate防撑爆)

crates/df-ai/src/openai_compat.rs

@@ -16,6 +16,9 @@ use crate::provider::{
     CompletionRequest, CompletionResponse, LlmProvider, StreamChunk, StreamResult,
     TokenUsage, ToolCall, ToolCallDelta,
 };
+// ChatMessage 仅单测构造 CompletionRequest 用，避免非 test 构建的 unused import 警告。
+#[cfg(test)]
+use crate::provider::ChatMessage;
 use crate::retry::{
     retry_with_backoff, AttemptOutcome, is_reqwest_error_retryable, is_status_retryable,
 };
@@ -44,10 +47,13 @@ struct OpenAiRequest {
 }
 
 /// OpenAI 消息格式
+///
+/// `content` 为 `serde_json::Value`：纯文本消息走字符串简写（与老端点兼容），
+/// 含图消息走 `[{type:"text",text},{type:"image_url",image_url:{url}}]` 数组。
 #[derive(Debug, Clone, Serialize, Deserialize)]
 struct OpenAiMessage {
     role: String,
-    content: String,
+    content: serde_json::Value,
     #[serde(skip_serializing_if = "Option::is_none")]
     tool_call_id: Option<String>,
     #[serde(skip_serializing_if = "Option::is_none")]
@@ -308,6 +314,38 @@ impl OpenAICompatProvider {
                     crate::provider::MessageRole::Assistant => "assistant",
                     crate::provider::MessageRole::Tool => "tool",
                 };
+                // F-260614-05 Phase 2a: 多模态 content（须在 move m.tool_calls 之前算，借用 m）。
+                // 含图消息走 content 数组（text/image_url）；纯文本走字符串简写
+                // （保持与现有纯文本端点零回归）。image_url 支持 data URI（base64）与 http(s) URL。
+                let content = if m.has_image() {
+                    let parts: Vec<serde_json::Value> = m
+                        .flattened_parts()
+                        .into_iter()
+                        .map(|p| match p {
+                            crate::provider::ContentPart::Text { text } => serde_json::json!({
+                                "type": "text",
+                                "text": text,
+                            }),
+                            crate::provider::ContentPart::Image { url, base64, media_type, alt: _ } => {
+                                let final_url = match (base64, url, media_type) {
+                                    (Some(b), _, Some(mt)) => {
+                                        format!("data:{};base64,{}", mt, b)
+                                    }
+                                    (None, Some(u), _) => u,
+                                    // 兜底：缺数据时退化为占位，避免发空 url 触发 400
+                                    _ => String::new(),
+                                };
+                                serde_json::json!({
+                                    "type": "image_url",
+                                    "image_url": { "url": final_url },
+                                })
+                            }
+                        })
+                        .collect();
+                    serde_json::Value::Array(parts)
+                } else {
+                    serde_json::Value::String(m.content.clone())
+                };
                 let tool_calls = m.tool_calls.map(|calls| {
                     calls
                         .into_iter()
@@ -325,7 +363,7 @@ impl OpenAICompatProvider {
                 });
                 OpenAiMessage {
                     role: role.to_string(),
-                    content: m.content,
+                    content,
                     tool_call_id: m.tool_call_id,
                     tool_calls,
                 }
@@ -691,4 +729,58 @@ mod tests {
         assert_eq!(tcs[0].function_arguments.as_deref(), Some("{\"q\":"));
         assert!(!c.finished);
     }
+
+    // ---------- F-260614-05 Phase 2a 多模态 convert_request ----------
+
+    /// 含图消息 → content 数组（text + image_url data URI）；纯文本 → 字符串简写
+    #[test]
+    fn openai_convert_multimodal_content() {
+        let provider = OpenAICompatProvider::new("https://api.openai.com", "k", "gpt-4o");
+        let req = CompletionRequest {
+            model: "gpt-4o".into(),
+            messages: vec![ChatMessage::user_parts(
+                "看图",
+                vec![
+                    crate::provider::ContentPart::image_base64("image/png", "iVBOR"),
+                    crate::provider::ContentPart::image_url("https://x/a.png"),
+                ],
+            )],
+            temperature: None,
+            max_tokens: None,
+            stream: false,
+            tools: None,
+            tool_choice: None,
+        };
+        let out = provider.convert_request(req);
+        let msg = &out.messages[0];
+        // content 是数组：[text "看图", image_url(data URI), image_url(http url)]
+        let arr = msg.content.as_array().expect("含图 → content 数组");
+        assert_eq!(arr.len(), 3);
+        assert_eq!(arr[0]["type"], "text");
+        assert_eq!(arr[0]["text"], "看图");
+        assert_eq!(arr[1]["type"], "image_url");
+        assert_eq!(
+            arr[1]["image_url"]["url"],
+            "data:image/png;base64,iVBOR"
+        );
+        assert_eq!(arr[2]["image_url"]["url"], "https://x/a.png");
+    }
+
+    /// 纯文本消息 → content 仍是字符串简写（无图不数组化，对齐纯文本端点兼容）
+    #[test]
+    fn openai_convert_text_only_remains_string() {
+        let provider = OpenAICompatProvider::new("https://api.openai.com", "k", "gpt-4o");
+        let req = CompletionRequest {
+            model: "gpt-4o".into(),
+            messages: vec![ChatMessage::user("hello")],
+            temperature: None,
+            max_tokens: None,
+            stream: false,
+            tools: None,
+            tool_choice: None,
+        };
+        let out = provider.convert_request(req);
+        let msg = &out.messages[0];
+        assert_eq!(msg.content, serde_json::Value::String("hello".into()));
+    }
 }