fix(model): anthropic默认max_tokens 1024→8192 — thinking+text共享预算,长思考不再截断正文为空

- 根因: deepseek-v4-flash-vision-exp 为 thinking 模型, content=[thinking,text] 共享 max_tokens,默认1024被长思考吃光→仅thinking块无text块, stop_reason=max_tokens → resp.Content="" → no reply produced
- 仅当 req.MaxTokens==0 套用默认(显式值不受影响); loop→agent→anthropic 传递链无截断
- 新增 anthropic_budget_test(默认发8192 + 长思考短text断言Content=text+thinking不泄漏)
- build/vet/test绿+交叉编译

internal/runtime/model/anthropic.go

@@ -20,7 +20,8 @@ const (
 	// anthropicDefaultMaxTokens is the default max_tokens sent when
 	// Request.MaxTokens is zero. Anthropic requires max_tokens to be present and
 	// positive, so a sensible default is applied rather than omitting it.
-	anthropicDefaultMaxTokens = 1024
+	// max_tokens is shared by thinking + text, so it must leave room for both.
+	anthropicDefaultMaxTokens = 8192
 	// anthropicDataPrefix marks a payload embedded as a data: URI.
 	anthropicDataPrefix = "data:"
 )

internal/runtime/model/anthropic_budget_test.go

@@ -0,0 +1,69 @@
+package model
+
+import (
+	"context"
+	"fmt"
+	"net/http"
+	"strings"
+	"testing"
+
+	"git.yeij.top/AskaEth/Cyrene_v2/internal/config"
+)
+
+// TestAnthropicChat_Chat_DefaultMaxTokensLargeEnoughForThinking guards the fix
+// for thinking models whose reasoning shares the max_tokens budget with the
+// visible answer: the zero-value default must leave room for both. It also
+// asserts an explicit Request.MaxTokens is never overridden by that default
+// (the existing 64-token case in anthropic_test.go is the compatibility proof).
+func TestAnthropicChat_Chat_DefaultMaxTokensLargeEnoughForThinking(t *testing.T) {
+	t.Parallel()
+	ch := make(chan anthropicCaptured, 1)
+	srv := newAnthropicServer(t, http.StatusOK, "application/json", anthropicTextResponse, ch)
+	ad := NewAnthropicChat(config.LLMConfig{AnthropicBaseURL: srv.URL, AnthropicModel: "claude-test"})
+
+	if _, err := ad.Chat(context.Background(),
+		Request{Messages: []Message{TextMessage(RoleUser, "hi")}}, Options{}); err != nil {
+		t.Fatalf("Chat error: %v", err)
+	}
+	if got := (<-ch).body["max_tokens"]; got != float64(anthropicDefaultMaxTokens) {
+		t.Errorf("default body.max_tokens = %v, want %d", got, anthropicDefaultMaxTokens)
+	}
+	if anthropicDefaultMaxTokens < 4096 {
+		t.Errorf("anthropicDefaultMaxTokens = %d, too small for a long thinking turn plus its answer", anthropicDefaultMaxTokens)
+	}
+}
+
+// TestAnthropicChat_Chat_LongThinkingThenShortText asserts a response whose
+// thinking block is far longer than its text block still surfaces the text and
+// never leaks the thinking into Content (the thinking-is-not-the-answer red
+// line). This is the shape a thinking model produces when the budget is large
+// enough, versus the truncation the old 1024 cap caused.
+func TestAnthropicChat_Chat_LongThinkingThenShortText(t *testing.T) {
+	t.Parallel()
+	thinking := strings.Repeat("推理", 1500) // 3000 runes, far longer than the answer
+	body := fmt.Sprintf(`{
+	  "id":"msg_think","model":"claude-test",
+	  "content":[
+	    {"type":"thinking","thinking":%q},
+	    {"type":"text","text":"现在是下午三点。"}
+	  ],
+	  "usage":{"input_tokens":20,"output_tokens":900}
+	}`, thinking)
+	srv := newAnthropicServer(t, http.StatusOK, "application/json", body, nil)
+	ad := NewAnthropicChat(config.LLMConfig{AnthropicBaseURL: srv.URL, AnthropicModel: "claude-test"})
+
+	out, err := ad.Chat(context.Background(),
+		Request{Messages: []Message{TextMessage(RoleUser, "几点了？")}}, Options{})
+	if err != nil {
+		t.Fatalf("Chat error: %v", err)
+	}
+	if out.Content != "现在是下午三点。" {
+		t.Errorf("Content = %q, want the text block", out.Content)
+	}
+	if strings.Contains(out.Content, "推理") {
+		t.Errorf("Content leaked thinking: %q", out.Content)
+	}
+	if out.Message.Role != RoleAssistant || len(out.Message.Content) != 1 || out.Message.Content[0].Text != "现在是下午三点。" {
+		t.Errorf("Message = %+v, want a single assistant text part", out.Message)
+	}
+}