fix(providers): improve error handling and add edge case tests

- fix ToolCalls nil vs empty slice issue to ensure consistent JSON serialization
- add detailed HTTP error handling for common status codes (401, 429, 400, 404, 500, 503)
- add edge case tests for buildRequestBody and parseResponseBody
- clarify anthropic vs anthropic-messages protocol differences in docs

README.md

@@ -1111,9 +1111,11 @@ For direct Anthropic API access or custom endpoints that only support Anthropic'
 ```
 
 > Use `anthropic-messages` protocol when:
-> - Connecting directly to Anthropic's API (fixes 404 errors with `/v1/messages` endpoint)
-> - Using custom endpoints that only support Anthropic's native format
-> - Avoiding OpenAI-compatible wrapper layers
+> - Using third-party proxies that only support Anthropic's native `/v1/messages` endpoint (not OpenAI-compatible `/v1/chat/completions`)
+> - Connecting to services like MiniMax, Synthetic that require Anthropic's native message format
+> - The existing `anthropic` protocol returns 404 errors (indicating the endpoint doesn't support OpenAI-compatible format)
+>
+> **Note:** The `anthropic` protocol uses OpenAI-compatible format (`/v1/chat/completions`), while `anthropic-messages` uses Anthropic's native format (`/v1/messages`). Choose based on your endpoint's supported format.
 
 **Ollama (local)**
 

README.zh.md

@@ -607,9 +607,11 @@ Agent 读取 HEARTBEAT.md
 ```
 
 > 使用 `anthropic-messages` 协议的场景：
-> - 直接连接 Anthropic API（修复 `/v1/messages` 端点的 404 错误）
-> - 使用仅支持 Anthropic 原生格式的自定义端点
-> - 避免 OpenAI 兼容包装层
+> - 使用仅支持 Anthropic 原生 `/v1/messages` 端点的第三方代理（不支持 OpenAI 兼容的 `/v1/chat/completions`）
+> - 连接到 MiniMax、Synthetic 等需要 Anthropic 原生消息格式的服务
+> - 现有的 `anthropic` 协议返回 404 错误（说明端点不支持 OpenAI 兼容格式）
+>
+> **注意：** `anthropic` 协议使用 OpenAI 兼容格式（`/v1/chat/completions`），而 `anthropic-messages` 使用 Anthropic 原生格式（`/v1/messages`）。请根据端点支持的格式选择。
 
 **Ollama (本地)**
 

pkg/providers/anthropic_messages/provider.go

@@ -119,10 +119,25 @@ func (p *Provider) Chat(
 		return nil, fmt.Errorf("reading response body: %w", err)
 	}
 
-	// Check for HTTP errors
+	// Check for HTTP errors with detailed messages
+	switch resp.StatusCode {
+	case http.StatusUnauthorized:
+		return nil, fmt.Errorf("authentication failed (401): check your API key")
+	case http.StatusTooManyRequests:
+		return nil, fmt.Errorf("rate limited (429): %s", string(body))
+	case http.StatusBadRequest:
+		return nil, fmt.Errorf("bad request (400): %s", string(body))
+	case http.StatusNotFound:
+		return nil, fmt.Errorf("endpoint not found (404): %s", string(body))
+	case http.StatusInternalServerError:
+		return nil, fmt.Errorf("internal server error (500): %s", string(body))
+	case http.StatusServiceUnavailable:
+		return nil, fmt.Errorf("service unavailable (503): %s", string(body))
+	default:
 		if resp.StatusCode != http.StatusOK {
 			return nil, fmt.Errorf("API request failed with status %d: %s", resp.StatusCode, string(body))
 		}
+	}
 
 	// Parse response
 	return parseResponseBody(body)
@@ -274,7 +289,7 @@ func parseResponseBody(body []byte) (*LLMResponse, error) {
 
 	// Extract content and tool calls
 	var content strings.Builder
-	var toolCalls []ToolCall
+	toolCalls := make([]ToolCall, 0) // Initialize as empty slice (not nil) for consistent JSON serialization
 
 	for _, block := range resp.Content {
 		switch block.Type {

pkg/providers/anthropic_messages/provider_test.go

@@ -9,6 +9,7 @@ import (
 	"context"
 	"encoding/json"
 	"reflect"
+	"strings"
 	"testing"
 )
 
@@ -431,6 +432,162 @@ func TestGetDefaultModel(t *testing.T) {
 	}
 }
 
+// TestBuildRequestBodyEdgeCases tests edge cases for buildRequestBody.
+func TestBuildRequestBodyEdgeCases(t *testing.T) {
+	tests := []struct {
+		name     string
+		messages []Message
+		tools    []ToolDefinition
+		model    string
+		options  map[string]any
+		wantErr  bool
+	}{
+		{
+			name:     "empty message list",
+			messages: []Message{},
+			model:    "test-model",
+			options: map[string]any{
+				"max_tokens": 8192,
+			},
+			wantErr: false,
+		},
+		{
+			name: "very long system message",
+			messages: []Message{
+				{Role: "system", Content: strings.Repeat("This is a very long system prompt. ", 1000)},
+				{Role: "user", Content: "Hello"},
+			},
+			model: "test-model",
+			options: map[string]any{
+				"max_tokens": 8192,
+			},
+			wantErr: false,
+		},
+		{
+			name: "multiple consecutive system messages",
+			messages: []Message{
+				{Role: "system", Content: "First system message"},
+				{Role: "system", Content: "Second system message"},
+				{Role: "system", Content: "Third system message"},
+				{Role: "user", Content: "Hello"},
+			},
+			model: "test-model",
+			options: map[string]any{
+				"max_tokens": 8192,
+			},
+			wantErr: false,
+		},
+		{
+			name: "tool result without tool call",
+			messages: []Message{
+				{Role: "user", Content: "Use a tool"},
+				{Role: "assistant", Content: "", ToolCalls: []ToolCall{
+					{ID: "tool-1", Name: "test_tool", Arguments: map[string]any{"arg": "value"}},
+				}},
+				{Role: "user", ToolCallID: "tool-1", Content: "Tool result"},
+			},
+			model: "test-model",
+			options: map[string]any{
+				"max_tokens": 8192,
+			},
+			wantErr: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got, err := buildRequestBody(tt.messages, tt.tools, tt.model, tt.options)
+			if (err != nil) != tt.wantErr {
+				t.Errorf("buildRequestBody() error = %v, wantErr %v", err, tt.wantErr)
+				return
+			}
+			if err != nil {
+				return
+			}
+
+			// Verify basic structure
+			if got == nil {
+				t.Error("buildRequestBody() returned nil")
+				return
+			}
+			if got["model"] != tt.model {
+				t.Errorf("model = %v, want %v", got["model"], tt.model)
+			}
+		})
+	}
+}
+
+// TestParseResponseBodyEdgeCases tests edge cases for parseResponseBody.
+func TestParseResponseBodyEdgeCases(t *testing.T) {
+	tests := []struct {
+		name    string
+		body    []byte
+		wantErr bool
+		check   func(*testing.T, *LLMResponse)
+	}{
+		{
+			name: "empty content blocks",
+			body: []byte(`{
+				"id": "msg-empty",
+				"type": "message",
+				"role": "assistant",
+				"content": [],
+				"stop_reason": "end_turn",
+				"model": "test-model",
+				"usage": {"input_tokens": 5, "output_tokens": 0}
+			}`),
+			wantErr: false,
+			check: func(t *testing.T, resp *LLMResponse) {
+				if resp.Content != "" {
+					t.Errorf("Content = %q, want empty string", resp.Content)
+				}
+				if len(resp.ToolCalls) != 0 {
+					t.Errorf("ToolCalls length = %d, want 0", len(resp.ToolCalls))
+				}
+			},
+		},
+		{
+			name: "multiple tool use blocks",
+			body: []byte(`{
+				"id": "msg-multi",
+				"type": "message",
+				"role": "assistant",
+				"content": [
+					{"type": "tool_use", "id": "tool-1", "name": "func1", "input": {"arg": "val1"}},
+					{"type": "tool_use", "id": "tool-2", "name": "func2", "input": {"arg": "val2"}}
+				],
+				"stop_reason": "tool_use",
+				"model": "test-model",
+				"usage": {"input_tokens": 10, "output_tokens": 20}
+			}`),
+			wantErr: false,
+			check: func(t *testing.T, resp *LLMResponse) {
+				if len(resp.ToolCalls) != 2 {
+					t.Errorf("ToolCalls length = %d, want 2", len(resp.ToolCalls))
+				}
+			},
+		},
+		{
+			name:    "malformed JSON response",
+			body:    []byte(`{invalid json`),
+			wantErr: true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got, err := parseResponseBody(tt.body)
+			if (err != nil) != tt.wantErr {
+				t.Errorf("parseResponseBody() error = %v, wantErr %v", err, tt.wantErr)
+				return
+			}
+			if tt.check != nil && err == nil {
+				tt.check(t, got)
+			}
+		})
+	}
+}
+
 // TestProviderChatErrors tests error handling in Chat.
 // Note: apiBase check removed as it's dead code - normalizeBaseURL() always provides a default.
 func TestProviderChatErrors(t *testing.T) {