feat(llm): 支持 OpenAI Responses API 格式的 LLM 配置

依赖升级到 norma v0.3.1(新增 FormatOpenAIResponses + reasoning_content 回传
修复)。每个 LLM profile 现在可选第三种格式 openai-responses,打
POST /v1/responses,与 Chat Completions / Anthropic 并列。

- agent.Config:ConfigFrom / FromEnv / Provider() 支持 openai-responses
  (BaseURL 归一化去 /responses 尾巴,默认 model gpt-5)
- server:saveLLMConfig 放开二元 format 归一;模型列表端点复用 OpenAI
  /models(Responses 模型列表仍在 /v1/models)
- db: llm_profiles 的 format CHECK 约束加入 openai-responses,并幂等迁移
  (DROP + ADD CONSTRAINT)补旧库
- web: 格式下拉新增「OpenAI (Responses API)」,类型与回显同步

流式/非流式两条路都走新 adapter;输出、token 统计、操作统计、LLM 录制
四条链路复用中性层,无需改动。

agent/provider.go

@@ -132,6 +132,12 @@ func FromEnv() (Config, bool) {
 		if c.Model == "" {
 			c.Model = "gpt-4o"
 		}
+	case "openai-responses":
+		c.Format = llm.FormatOpenAIResponses
+		c.APIKey = oaiKey
+		if c.Model == "" {
+			c.Model = "gpt-5"
+		}
 	default:
 		c.Format = llm.FormatAnthropic
 		c.APIKey = anthKey
@@ -165,6 +171,13 @@ func ConfigFrom(provider, model, baseURL, apiKey, proxy string) Config {
 		if c.Model == "" {
 			c.Model = "gpt-4o"
 		}
+	case "openai-responses":
+		c.Format = llm.FormatOpenAIResponses
+		// provider appends "/responses"; tolerate a full endpoint URL.
+		c.BaseURL = strings.TrimRight(strings.TrimSuffix(c.BaseURL, "/responses"), "/")
+		if c.Model == "" {
+			c.Model = "gpt-5"
+		}
 	default:
 		c.Format = llm.FormatAnthropic
 		// provider appends "/v1/messages".
@@ -186,10 +199,13 @@ func isFalsy(s string) bool {
 	return false
 }
 
-// Provider returns the short provider name ("anthropic"/"openai").
+// Provider returns the short provider name ("anthropic"/"openai"/"openai-responses").
 func (c Config) Provider() string {
-	if c.Format == llm.FormatOpenAI {
+	switch c.Format {
+	case llm.FormatOpenAI:
 		return "openai"
+	case llm.FormatOpenAIResponses:
+		return "openai-responses"
 	}
 	return "anthropic"
 }

agent/provider_responses_test.go

@@ -0,0 +1,30 @@
+package agent
+
+import (
+	"testing"
+
+	"github.com/Autumn-27/norma/llm"
+)
+
+// TestConfigFromOpenAIResponses locks the openai-responses format wiring:
+// ConfigFrom resolves the Responses format, strips a full /responses endpoint
+// back to the API base, and Provider() round-trips the short name. NewProvider
+// must build a working provider for it.
+func TestConfigFromOpenAIResponses(t *testing.T) {
+	c := ConfigFrom("openai-responses", "gpt-5", "https://gw.example/v1/responses", "sk-x", "")
+	if c.Format != llm.FormatOpenAIResponses {
+		t.Fatalf("format=%v, want FormatOpenAIResponses", c.Format)
+	}
+	if c.BaseURL != "https://gw.example/v1" {
+		t.Fatalf("base_url=%q, want the /responses suffix stripped", c.BaseURL)
+	}
+	if c.Provider() != "openai-responses" {
+		t.Fatalf("Provider()=%q", c.Provider())
+	}
+	if !c.Stream { // default streaming preserved
+		t.Fatal("Stream should default true")
+	}
+	if _, err := c.NewProvider(); err != nil {
+		t.Fatalf("NewProvider: %v", err)
+	}
+}