Merge pull request #416 from CJackHwang/main

Add Star History section to README

Added a Star History section with a chart to the README.

API.md

@@ -42,7 +42,7 @@
 - Tool Calling 的解析策略在 Go 与 Node Runtime 间保持一致：推荐模型输出 DSML 外壳 `<|DSML|tool_calls>` → `<|DSML|invoke name="...">` → `<|DSML|parameter name="...">`；兼容层也接受 DSML wrapper 别名 `<dsml|tool_calls>`、`<|tool_calls>`、`<｜tool_calls>`、常见 DSML 分隔符漏写形态（如 `<|DSML tool_calls>`）、`DSML` 与工具标签名黏连的常见 typo（如 `<DSMLtool_calls>`），以及旧式 canonical XML `<tool_calls>` → `<invoke name="...">` → `<parameter name="...">`。实现上采用窄容错结构扫描：只有 `tool_calls` wrapper 或可修复的缺失 opening wrapper 会进入工具路径，裸 `<invoke>` 不计为已支持语法；流式场景继续执行防泄漏筛分。若参数体本身是合法 JSON 字面量（如 `123`、`true`、`null`、数组或对象），会按结构化值输出，不再一律当作字符串；若 CDATA 偶发漏闭合，则会在最终 parse / flush 恢复阶段做窄修复，尽量保住已完整包裹的外层工具调用。
 - `Admin API` 将配置与运行时策略分开：`/admin/config*` 管静态配置，`/admin/settings*` 管运行时行为。
 - 当上游返回 thinking-only 响应（模型输出了推理链但无可见文本）时，非流式补全会自动重试一次：以多轮对话 follow-up 方式追加 prompt 后缀 `"Previous reply had no visible output. Please regenerate the visible final answer or tool call now."` 并设置 `parent_message_id` 在同一 DeepSeek session 内让模型重新输出；重试最大 1 次。
-- 引用标记剥离（strip reference markers）当前为固定开启的运行时行为，所有协议适配层统一生效。
+- 引用标记处理边界：流式输出默认隐藏 `[citation:N]` / `[reference:N]` 这类上游内部占位符；非流式输出默认把 DeepSeek 搜索引用标记转换为 Markdown 引用链接。
 
 ---
 
@@ -168,6 +168,8 @@ Gemini 兼容客户端还可以使用 `x-goog-api-key`、`?key=` 或 `?api_key=`
 | GET | `/admin/chat-history/{id}` | Admin | 查看单条服务器端对话记录 |
 | DELETE | `/admin/chat-history/{id}` | Admin | 删除单条服务器端对话记录 |
 | PUT | `/admin/chat-history/settings` | Admin | 更新对话记录保留条数 |
+
+服务器端记录本质上是 DeepSeek 上游响应归档：OpenAI Chat、OpenAI Responses、Claude Messages、Gemini GenerateContent 等直连 DeepSeek 的生成接口，在收到上游响应后会于各协议回译/裁剪前写入记录；列表按请求创建时间倒序展示，流式请求会在生成过程中持续刷新状态与详情。WebUI「API 测试」发出的请求也会进入该记录。
 | GET | `/admin/version` | Admin | 查询当前版本与最新 Release |
 
 OpenAI `/v1/*` 仍是规范路径。对于只配置 DS2API 根地址的客户端，同一套 OpenAI handler 也通过根路径快捷路由暴露：`/models`、`/models/{id}`、`/chat/completions`、`/responses`、`/responses/{response_id}`、`/embeddings`、`/files`、`/files/{file_id}`。

README.MD

@@ -23,6 +23,16 @@
 
 【感谢Linux.do社区及GitHub社区各位开发者对项目的支持与贡献】
 
+## Star History
+
+<a href="https://www.star-history.com/?repos=cjackhwang%2Fds2api&type=date&legend=top-left">
+ <picture>
+   <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/chart?repos=cjackhwang/ds2api&type=date&theme=dark&legend=top-left" />
+   <source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/chart?repos=cjackhwang/ds2api&type=date&legend=top-left" />
+   <img alt="Star History Chart" src="https://api.star-history.com/chart?repos=cjackhwang/ds2api&type=date&legend=top-left" />
+ </picture>
+</a>
+
 > **重要免责声明**
 >
 > 本仓库仅供学习、研究、个人实验和内部验证使用，不提供任何形式的商业授权、适用性保证或结果保证。

README.en.md

@@ -20,6 +20,16 @@ DS2API converts DeepSeek Web chat capability into OpenAI-compatible, Claude-comp
 
 Documentation entry: [Docs Index](docs/README.md) / [Architecture](docs/ARCHITECTURE.en.md) / [API Reference](API.en.md)
 
+## Star History
+
+<a href="https://www.star-history.com/?repos=cjackhwang%2Fds2api&type=date&legend=top-left">
+ <picture>
+   <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/chart?repos=cjackhwang/ds2api&type=date&theme=dark&legend=top-left" />
+   <source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/chart?repos=cjackhwang/ds2api&type=date&legend=top-left" />
+   <img alt="Star History Chart" src="https://api.star-history.com/chart?repos=cjackhwang/ds2api&type=date&legend=top-left" />
+ </picture>
+</a>
+
 > **Important Disclaimer**
 >
 > This repository is provided for learning, research, personal experimentation, and internal validation only. It does not grant any commercial authorization and comes with no warranty of fitness, stability, or results.

VERSION

@@ -1 +1 @@
-4.4.0
+4.4.1

docs/DEVELOPMENT.md

@@ -74,6 +74,7 @@ gofmt -w <changed-go-files>
 
 - Admin API：`/admin/chat-history`、`/admin/chat-history/{id}`。
 - 后端存储：`internal/chathistory/store.go`。
+- 输出归档：`internal/responsehistory` 在协议回译/裁剪前记录 DeepSeek 上游 assistant text / thinking；即使工具调用已被对外响应转成结构化 `tool_calls` 并从可见正文剔除，后台历史仍应保留原始 DSML / XML 片段，方便排查格式漂移。
 - 前端轮询和 ETag：`webui/src/features/chatHistory/ChatHistoryContainer.jsx`。
 
 Tool call 问题优先跑：

docs/prompt-compatibility.md

@@ -114,7 +114,8 @@ DS2API 当前的核心思路，不是把客户端传来的 `messages`、`tools`
 - 对 OpenAI Chat / Responses 的非流式收尾，如果最终可见正文为空，兼容层会优先尝试把思维链中的独立 DSML / XML 工具块当作真实工具调用解析出来。流式链路也会在收尾阶段做同样的 fallback 检测，但不会因为思维链内容去中途拦截或改写流式输出；真正的工具识别始终基于原始上游文本，而不是基于“已经做过可见输出清洗”的版本，因此即使最终可见层会剥离完整 leaked DSML / XML `tool_calls` wrapper、并抑制全空参数或无效 wrapper 块，也不会影响真实工具调用转成结构化 `tool_calls` / `function_call`。补发结果会作为本轮 assistant 的结构化 `tool_calls` / `function_call` 输出返回，而不是塞进 `content` 文本；如果客户端没有开启 thinking / reasoning，思维链只用于检测，不会作为 `reasoning_content` 或可见正文暴露。只有正文为空且思维链里也没有可执行工具调用时，才继续按空回复错误处理。
 - OpenAI Chat / Responses 的空回复错误处理之前会默认做一次内部补偿重试：第一次上游完整结束后，如果最终可见正文为空、没有解析到工具调用、也没有已经向客户端流式发出工具调用，并且终止原因不是 `content_filter`，兼容层会复用同一个 `chat_session_id`、账号、token 与工具策略，把原始 completion `prompt` 追加固定后缀 `Previous reply had no visible output. Please regenerate the visible final answer or tool call now.` 后重新提交一次。重试遵循 DeepSeek 多轮对话协议：从第一次上游 SSE 流中提取 `response_message_id`，并在重试 payload 中设置 `parent_message_id` 为该值，使重试成为同一会话的后续轮次而非断裂的根消息；同时重新获取一次 PoW（若 PoW 获取失败则回退到原始 PoW）。该重试不会重新标准化消息、不会新建 session、不会切换账号，也不会向流式客户端插入重试标记；第二次 thinking / reasoning 会按正常增量直接接到第一次之后，并继续使用 overlap trim 去重。若第二次仍为空，终端错误码仍保持现有 `upstream_empty_output`；若任一尝试触发空 `content_filter`，不做补偿重试并保持 `content_filter` 错误。JS Vercel 运行时同样设置 `parent_message_id`，但因无法直接调用 PoW API 而复用原始 PoW。
 
-- OpenAI Chat / Responses 在最终可见正文渲染阶段，会把 DeepSeek 搜索返回中的 `[citation:N]` / `[reference:N]` 标记替换成对应 Markdown 链接。`citation` 标记按一基序号解析；`reference` 标记只有在同一段正文中出现 `[reference:0]`（允许冒号后有空格）时才按零基序号映射，并且不会影响同段正文里的 `citation` 标记。
+- 非流式 OpenAI Chat / Responses、Claude Messages、Gemini generateContent 在最终可见正文渲染阶段，会把 DeepSeek 搜索返回中的 `[citation:N]` / `[reference:N]` 标记替换成对应 Markdown 链接。`citation` 标记按一基序号解析；`reference` 标记只有在同一段正文中出现 `[reference:0]`（允许冒号后有空格）时才按零基序号映射，并且不会影响同段正文里的 `citation` 标记。
+- 流式输出仍默认隐藏 `[citation:N]` / `[reference:N]` 这类上游内部标记，避免分片输出中泄漏尚未完成映射的引用占位符。
 
 ## 5. prompt 是怎么拼出来的
 

internal/assistantturn/turn_test.go

@@ -11,7 +11,7 @@ func TestBuildTurnFromCollectedTextCitation(t *testing.T) {
 	turn := BuildTurnFromCollected(sse.CollectResult{
 		Text:          "See [citation:1]",
 		CitationLinks: map[int]string{1: "https://example.com"},
-	}, BuildOptions{Model: "deepseek-v4-flash", Prompt: "prompt", SearchEnabled: true, StripReferenceMarkers: true})
+	}, BuildOptions{Model: "deepseek-v4-flash", Prompt: "prompt", SearchEnabled: true})
 	if turn.Text != "See [1](https://example.com)" {
 		t.Fatalf("text mismatch: %q", turn.Text)
 	}
@@ -23,6 +23,20 @@ func TestBuildTurnFromCollectedTextCitation(t *testing.T) {
 	}
 }
 
+func TestBuildTurnFromCollectedKeepsNonStreamReferenceLinks(t *testing.T) {
+	turn := BuildTurnFromCollected(sse.CollectResult{
+		Text: "结论[reference:0]，补充[reference:1]。",
+		CitationLinks: map[int]string{
+			1: "https://example.com/a",
+			2: "https://example.com/b",
+		},
+	}, BuildOptions{Model: "deepseek-v4-flash-search", Prompt: "prompt", SearchEnabled: true})
+	want := "结论[0](https://example.com/a)，补充[1](https://example.com/b)。"
+	if turn.Text != want {
+		t.Fatalf("text mismatch: got %q want %q", turn.Text, want)
+	}
+}
+
 func TestBuildTurnFromCollectedToolCall(t *testing.T) {
 	turn := BuildTurnFromCollected(sse.CollectResult{
 		Text: `<tool_calls><invoke name="Write"><parameter name="content">{"x":1}</parameter></invoke></tool_calls>`,

internal/chathistory/store.go

@@ -43,6 +43,7 @@ type Entry struct {
 	Status           string         `json:"status"`
 	CallerID         string         `json:"caller_id,omitempty"`
 	AccountID        string         `json:"account_id,omitempty"`
+	Surface          string         `json:"surface,omitempty"`
 	Model            string         `json:"model,omitempty"`
 	Stream           bool           `json:"stream"`
 	UserInput        string         `json:"user_input,omitempty"`
@@ -72,6 +73,7 @@ type SummaryEntry struct {
 	Status         string `json:"status"`
 	CallerID       string `json:"caller_id,omitempty"`
 	AccountID      string `json:"account_id,omitempty"`
+	Surface        string `json:"surface,omitempty"`
 	Model          string `json:"model,omitempty"`
 	Stream         bool   `json:"stream"`
 	UserInput      string `json:"user_input,omitempty"`
@@ -92,6 +94,7 @@ type File struct {
 type StartParams struct {
 	CallerID    string
 	AccountID   string
+	Surface     string
 	Model       string
 	Stream      bool
 	UserInput   string
@@ -271,6 +274,7 @@ func (s *Store) Start(params StartParams) (Entry, error) {
 		Status:      "streaming",
 		CallerID:    strings.TrimSpace(params.CallerID),
 		AccountID:   strings.TrimSpace(params.AccountID),
+		Surface:     strings.TrimSpace(params.Surface),
 		Model:       strings.TrimSpace(params.Model),
 		Stream:      params.Stream,
 		UserInput:   strings.TrimSpace(params.UserInput),
@@ -546,10 +550,13 @@ func (s *Store) rebuildIndexLocked() {
 		summaries = append(summaries, summaryFromEntry(item))
 	}
 	sort.Slice(summaries, func(i, j int) bool {
-		if summaries[i].UpdatedAt == summaries[j].UpdatedAt {
-			return summaries[i].CreatedAt > summaries[j].CreatedAt
+		if summaries[i].CreatedAt == summaries[j].CreatedAt {
+			if summaries[i].Revision == summaries[j].Revision {
+				return summaries[i].UpdatedAt > summaries[j].UpdatedAt
+			}
+			return summaries[i].Revision > summaries[j].Revision
 		}
-		return summaries[i].UpdatedAt > summaries[j].UpdatedAt
+		return summaries[i].CreatedAt > summaries[j].CreatedAt
 	})
 	if s.state.Limit < DisabledLimit || !isAllowedLimit(s.state.Limit) {
 		s.state.Limit = DefaultLimit
@@ -593,6 +600,7 @@ func summaryFromEntry(item Entry) SummaryEntry {
 		Status:         item.Status,
 		CallerID:       item.CallerID,
 		AccountID:      item.AccountID,
+		Surface:        item.Surface,
 		Model:          item.Model,
 		Stream:         item.Stream,
 		UserInput:      item.UserInput,

internal/chathistory/store_test.go

@@ -8,6 +8,7 @@ import (
 	"strings"
 	"sync"
 	"testing"
+	"time"
 	"unicode/utf8"
 )
 
@@ -494,6 +495,36 @@ func TestStoreWritesOnlyChangedDetailFiles(t *testing.T) {
 	}
 }
 
+func TestStoreOrdersByCreationTimeNotStreamingUpdates(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "chat_history.json")
+	store := New(path)
+
+	first, err := store.Start(StartParams{UserInput: "first"})
+	if err != nil {
+		t.Fatalf("start first failed: %v", err)
+	}
+	time.Sleep(time.Millisecond)
+	second, err := store.Start(StartParams{UserInput: "second"})
+	if err != nil {
+		t.Fatalf("start second failed: %v", err)
+	}
+	time.Sleep(time.Millisecond)
+	if _, err := store.Update(first.ID, UpdateParams{Status: "streaming", Content: "still running"}); err != nil {
+		t.Fatalf("update first failed: %v", err)
+	}
+
+	snapshot, err := store.Snapshot()
+	if err != nil {
+		t.Fatalf("snapshot failed: %v", err)
+	}
+	if len(snapshot.Items) != 2 {
+		t.Fatalf("expected two items, got %#v", snapshot.Items)
+	}
+	if snapshot.Items[0].ID != second.ID || snapshot.Items[1].ID != first.ID {
+		t.Fatalf("expected creation-time order to stay stable, got %#v", snapshot.Items)
+	}
+}
+
 func TestUpdatePreservesContentWhenNewContentIsEmpty(t *testing.T) {
 	path := filepath.Join(t.TempDir(), "chat_history.json")
 	store := New(path)

internal/completionruntime/nonstream_test.go

@@ -119,6 +119,29 @@ func TestExecuteNonStreamWithRetryUsesParentMessageForEmptyRetry(t *testing.T) {
 	}
 }
 
+func TestExecuteNonStreamWithRetryConvertsReferenceMarkers(t *testing.T) {
+	ds := &fakeDeepSeekCaller{responses: []*http.Response{sseHTTPResponse(
+		http.StatusOK,
+		`data: {"p":"response/content","v":"答案[reference:0]。","citation":{"cite_index":0,"url":"https://example.com/ref"}}`,
+	)}}
+	stdReq := promptcompat.StandardRequest{
+		Surface:         "test",
+		ResponseModel:   "deepseek-v4-flash-search",
+		PromptTokenText: "prompt",
+		FinalPrompt:     "final prompt",
+		Search:          true,
+	}
+
+	result, outErr := ExecuteNonStreamWithRetry(context.Background(), ds, &auth.RequestAuth{}, stdReq, Options{})
+	if outErr != nil {
+		t.Fatalf("unexpected output error: %#v", outErr)
+	}
+	want := "答案[0](https://example.com/ref)。"
+	if result.Turn.Text != want {
+		t.Fatalf("text mismatch: got %q want %q", result.Turn.Text, want)
+	}
+}
+
 func TestStartCompletionAppliesCurrentInputFileGlobally(t *testing.T) {
 	ds := &fakeDeepSeekCaller{responses: []*http.Response{sseHTTPResponse(http.StatusOK, `data: {"p":"response/content","v":"ok"}`)}}
 	stdReq := promptcompat.StandardRequest{

internal/httpapi/claude/current_input_file_test.go

@@ -5,15 +5,25 @@ import (
 	"io"
 	"net/http"
 	"net/http/httptest"
+	"path/filepath"
 	"strings"
 	"testing"
 
 	"ds2api/internal/auth"
+	"ds2api/internal/chathistory"
 	dsclient "ds2api/internal/deepseek/client"
 )
 
 type claudeCurrentInputAuth struct{}
 
+type claudeHistoryConfig struct {
+	aliases map[string]string
+}
+
+func (m claudeHistoryConfig) ModelAliases() map[string]string { return m.aliases }
+func (claudeHistoryConfig) CurrentInputFileEnabled() bool     { return false }
+func (claudeHistoryConfig) CurrentInputFileMinChars() int     { return 0 }
+
 func (claudeCurrentInputAuth) Determine(*http.Request) (*auth.RequestAuth, error) {
 	return &auth.RequestAuth{
 		DeepSeekToken: "direct-token",
@@ -22,6 +32,50 @@ func (claudeCurrentInputAuth) Determine(*http.Request) (*auth.RequestAuth, error
 	}, nil
 }
 
+func TestClaudeDirectRecordsResponseHistory(t *testing.T) {
+	ds := &claudeCurrentInputDS{}
+	historyStore := chathistory.New(filepath.Join(t.TempDir(), "history.json"))
+	h := &Handler{
+		Store:       claudeHistoryConfig{aliases: map[string]string{"claude-sonnet-4-6": "deepseek-v4-flash"}},
+		Auth:        claudeCurrentInputAuth{},
+		DS:          ds,
+		ChatHistory: historyStore,
+	}
+	reqBody := `{"model":"claude-sonnet-4-6","messages":[{"role":"user","content":"hello from claude"}],"max_tokens":1024}`
+	req := httptest.NewRequest(http.MethodPost, "/v1/messages", strings.NewReader(reqBody))
+	req.Header.Set("Content-Type", "application/json")
+	rec := httptest.NewRecorder()
+
+	h.Messages(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("expected 200, got %d body=%s", rec.Code, rec.Body.String())
+	}
+	snapshot, err := historyStore.Snapshot()
+	if err != nil {
+		t.Fatalf("snapshot history: %v", err)
+	}
+	if len(snapshot.Items) != 1 {
+		t.Fatalf("expected one history item, got %d", len(snapshot.Items))
+	}
+	item, err := historyStore.Get(snapshot.Items[0].ID)
+	if err != nil {
+		t.Fatalf("get history item: %v", err)
+	}
+	if item.Surface != "claude.messages" {
+		t.Fatalf("unexpected surface: %q", item.Surface)
+	}
+	if item.Model != "claude-sonnet-4-6" {
+		t.Fatalf("unexpected model: %q", item.Model)
+	}
+	if item.UserInput != "hello from claude" {
+		t.Fatalf("unexpected user input: %q", item.UserInput)
+	}
+	if item.Content != "ok" {
+		t.Fatalf("expected raw upstream content, got %q", item.Content)
+	}
+}
+
 func (claudeCurrentInputAuth) Release(*auth.RequestAuth) {}
 
 type claudeCurrentInputDS struct {
@@ -53,10 +107,12 @@ func (d *claudeCurrentInputDS) CallCompletion(_ context.Context, _ *auth.Request
 
 func TestClaudeDirectAppliesCurrentInputFile(t *testing.T) {
 	ds := &claudeCurrentInputDS{}
+	historyStore := chathistory.New(filepath.Join(t.TempDir(), "history.json"))
 	h := &Handler{
-		Store: mockClaudeConfig{aliases: map[string]string{"claude-sonnet-4-6": "deepseek-v4-flash"}},
-		Auth:  claudeCurrentInputAuth{},
-		DS:    ds,
+		Store:       mockClaudeConfig{aliases: map[string]string{"claude-sonnet-4-6": "deepseek-v4-flash"}},
+		Auth:        claudeCurrentInputAuth{},
+		DS:          ds,
+		ChatHistory: historyStore,
 	}
 	reqBody := `{"model":"claude-sonnet-4-6","messages":[{"role":"user","content":"hello from claude"}],"max_tokens":1024}`
 	req := httptest.NewRequest(http.MethodPost, "/v1/messages", strings.NewReader(reqBody))
@@ -82,4 +138,21 @@ func TestClaudeDirectAppliesCurrentInputFile(t *testing.T) {
 	if !strings.Contains(prompt, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
 		t.Fatalf("expected continuation prompt, got %q", prompt)
 	}
+	snapshot, err := historyStore.Snapshot()
+	if err != nil {
+		t.Fatalf("snapshot history: %v", err)
+	}
+	if len(snapshot.Items) != 1 {
+		t.Fatalf("expected one history item, got %d", len(snapshot.Items))
+	}
+	full, err := historyStore.Get(snapshot.Items[0].ID)
+	if err != nil {
+		t.Fatalf("get history item: %v", err)
+	}
+	if full.HistoryText != string(ds.uploads[0].Data) {
+		t.Fatalf("expected uploaded current input file to be persisted in history text")
+	}
+	if len(full.Messages) != 1 || !strings.Contains(full.Messages[0].Content, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
+		t.Fatalf("expected persisted message to match upstream continuation prompt, got %#v", full.Messages)
+	}
 }

internal/httpapi/claude/handler_messages.go

@@ -2,6 +2,7 @@ package claude
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -15,8 +16,10 @@ import (
 	"ds2api/internal/completionruntime"
 	"ds2api/internal/config"
 	claudefmt "ds2api/internal/format/claude"
+	"ds2api/internal/httpapi/openai/history"
 	"ds2api/internal/httpapi/requestbody"
 	"ds2api/internal/promptcompat"
+	"ds2api/internal/responsehistory"
 	streamengine "ds2api/internal/stream"
 	"ds2api/internal/translatorcliproxy"
 	"ds2api/internal/util"
@@ -79,38 +82,70 @@ func (h *Handler) handleClaudeDirect(w http.ResponseWriter, r *http.Request) boo
 		return true
 	}
 	defer h.Auth.Release(a)
-	if norm.Standard.Stream {
-		h.handleClaudeDirectStream(w, r, a, norm.Standard)
+	stdReq, err := h.applyCurrentInputFile(r.Context(), a, norm.Standard)
+	if err != nil {
+		status, message := mapCurrentInputFileError(err)
+		writeClaudeError(w, status, message)
 		return true
 	}
-	result, outErr := completionruntime.ExecuteNonStreamWithRetry(r.Context(), h.DS, a, norm.Standard, completionruntime.Options{
-		StripReferenceMarkers: stripReferenceMarkersEnabled(),
-		RetryEnabled:          true,
-		CurrentInputFile:      h.Store,
+	historySession := responsehistory.Start(responsehistory.StartParams{
+		Store:    h.ChatHistory,
+		Request:  r,
+		Auth:     a,
+		Surface:  "claude.messages",
+		Standard: stdReq,
+	})
+	if stdReq.Stream {
+		h.handleClaudeDirectStream(w, r, a, stdReq, historySession)
+		return true
+	}
+	result, outErr := completionruntime.ExecuteNonStreamWithRetry(r.Context(), h.DS, a, stdReq, completionruntime.Options{
+		RetryEnabled:     true,
+		CurrentInputFile: h.Store,
 	})
 	if outErr != nil {
+		if historySession != nil {
+			historySession.ErrorTurn(outErr.Status, outErr.Message, outErr.Code, result.Turn)
+		}
 		writeClaudeError(w, outErr.Status, outErr.Message)
 		return true
 	}
+	if historySession != nil {
+		historySession.SuccessTurn(http.StatusOK, result.Turn, responsehistory.GenericUsage(result.Turn))
+	}
 	writeJSON(w, http.StatusOK, claudefmt.BuildMessageResponseFromTurn(
 		fmt.Sprintf("msg_%d", time.Now().UnixNano()),
-		norm.Standard.ResponseModel,
+		stdReq.ResponseModel,
 		result.Turn,
 		exposeThinking,
 	))
 	return true
 }
 
-func (h *Handler) handleClaudeDirectStream(w http.ResponseWriter, r *http.Request, a *auth.RequestAuth, stdReq promptcompat.StandardRequest) {
+func (h *Handler) applyCurrentInputFile(ctx context.Context, a *auth.RequestAuth, stdReq promptcompat.StandardRequest) (promptcompat.StandardRequest, error) {
+	if h == nil {
+		return stdReq, nil
+	}
+	return (history.Service{Store: h.Store, DS: h.DS}).ApplyCurrentInputFile(ctx, a, stdReq)
+}
+
+func mapCurrentInputFileError(err error) (int, string) {
+	return history.MapError(err)
+}
+
+func (h *Handler) handleClaudeDirectStream(w http.ResponseWriter, r *http.Request, a *auth.RequestAuth, stdReq promptcompat.StandardRequest, historySession *responsehistory.Session) {
 	start, outErr := completionruntime.StartCompletion(r.Context(), h.DS, a, stdReq, completionruntime.Options{
 		CurrentInputFile: h.Store,
 	})
 	if outErr != nil {
+		if historySession != nil {
+			historySession.Error(outErr.Status, outErr.Message, outErr.Code, "", "")
+		}
 		writeClaudeError(w, outErr.Status, outErr.Message)
 		return
 	}
 	streamReq := start.Request
-	h.handleClaudeStreamRealtime(w, r, start.Response, streamReq.ResponseModel, streamReq.Messages, streamReq.Thinking, streamReq.Search, streamReq.ToolNames, streamReq.ToolsRaw)
+	h.handleClaudeStreamRealtime(w, r, start.Response, streamReq.ResponseModel, streamReq.Messages, streamReq.Thinking, streamReq.Search, streamReq.ToolNames, streamReq.ToolsRaw, historySession)
 }
 
 func (h *Handler) proxyViaOpenAI(w http.ResponseWriter, r *http.Request, store ConfigReader) bool {
@@ -264,10 +299,17 @@ func stripClaudeThinkingBlocks(raw []byte) []byte {
 	return out
 }
 
-func (h *Handler) handleClaudeStreamRealtime(w http.ResponseWriter, r *http.Request, resp *http.Response, model string, messages []any, thinkingEnabled, searchEnabled bool, toolNames []string, toolsRaw any) {
+func (h *Handler) handleClaudeStreamRealtime(w http.ResponseWriter, r *http.Request, resp *http.Response, model string, messages []any, thinkingEnabled, searchEnabled bool, toolNames []string, toolsRaw any, historySessions ...*responsehistory.Session) {
+	var historySession *responsehistory.Session
+	if len(historySessions) > 0 {
+		historySession = historySessions[0]
+	}
 	defer func() { _ = resp.Body.Close() }()
 	if resp.StatusCode != http.StatusOK {
 		body, _ := io.ReadAll(resp.Body)
+		if historySession != nil {
+			historySession.Error(resp.StatusCode, strings.TrimSpace(string(body)), "error", "", "")
+		}
 		writeClaudeError(w, http.StatusInternalServerError, string(body))
 		return
 	}
@@ -294,6 +336,7 @@ func (h *Handler) handleClaudeStreamRealtime(w http.ResponseWriter, r *http.Requ
 		toolNames,
 		toolsRaw,
 		buildClaudePromptTokenText(messages, thinkingEnabled),
+		historySession,
 	)
 	streamRuntime.sendMessageStart()
 

internal/httpapi/claude/handler_routes.go

@@ -6,6 +6,7 @@ import (
 
 	"github.com/go-chi/chi/v5"
 
+	"ds2api/internal/chathistory"
 	"ds2api/internal/config"
 	dsprotocol "ds2api/internal/deepseek/protocol"
 	"ds2api/internal/textclean"
@@ -16,10 +17,11 @@ import (
 var writeJSON = util.WriteJSON
 
 type Handler struct {
-	Store  ConfigReader
-	Auth   AuthResolver
-	DS     DeepSeekCaller
-	OpenAI OpenAIChatRunner
+	Store       ConfigReader
+	Auth        AuthResolver
+	DS          DeepSeekCaller
+	OpenAI      OpenAIChatRunner
+	ChatHistory *chathistory.Store
 }
 
 func stripReferenceMarkersEnabled() bool {

internal/httpapi/claude/stream_runtime_core.go

@@ -6,6 +6,7 @@ import (
 	"strings"
 	"time"
 
+	"ds2api/internal/responsehistory"
 	"ds2api/internal/sse"
 	streamengine "ds2api/internal/stream"
 	"ds2api/internal/toolcall"
@@ -46,6 +47,7 @@ type claudeStreamRuntime struct {
 	textEmitted        bool
 	ended              bool
 	upstreamErr        string
+	history            *responsehistory.Session
 }
 
 func newClaudeStreamRuntime(
@@ -60,6 +62,7 @@ func newClaudeStreamRuntime(
 	toolNames []string,
 	toolsRaw any,
 	promptTokenText string,
+	history *responsehistory.Session,
 ) *claudeStreamRuntime {
 	return &claudeStreamRuntime{
 		w:                     w,
@@ -74,6 +77,7 @@ func newClaudeStreamRuntime(
 		toolNames:             toolNames,
 		toolsRaw:              toolsRaw,
 		promptTokenText:       promptTokenText,
+		history:               history,
 		messageID:             fmt.Sprintf("msg_%d", time.Now().UnixNano()),
 		thinkingBlockIndex:    -1,
 		textBlockIndex:        -1,
@@ -232,5 +236,11 @@ func (s *claudeStreamRuntime) onParsed(parsed sse.LineResult) streamengine.Parse
 		}
 	}
 
+	if s.history != nil {
+		s.history.Progress(
+			responsehistory.ThinkingForArchive(s.rawThinking.String(), s.toolDetectionThinking.String(), s.thinking.String()),
+			responsehistory.TextForArchive(s.rawText.String(), s.text.String()),
+		)
+	}
 	return streamengine.ParsedDecision{ContentSeen: contentSeen}
 }

internal/httpapi/claude/stream_runtime_finalize.go

@@ -2,6 +2,7 @@ package claude
 
 import (
 	"ds2api/internal/assistantturn"
+	"ds2api/internal/responsehistory"
 	"ds2api/internal/sse"
 	"ds2api/internal/toolcall"
 	"ds2api/internal/toolstream"
@@ -175,6 +176,15 @@ func (s *claudeStreamRuntime) finalize(stopReason string) {
 	if outcome.HasToolCalls {
 		stopReason = "tool_use"
 	}
+	if s.history != nil {
+		s.history.Success(
+			200,
+			responsehistory.ThinkingForArchive(turn.RawThinking, turn.DetectionThinking, turn.Thinking),
+			responsehistory.TextForArchive(turn.RawText, turn.Text),
+			stopReason,
+			responsehistory.GenericUsage(turn),
+		)
+	}
 
 	s.send("message_delta", map[string]any{
 		"type": "message_delta",
@@ -191,10 +201,16 @@ func (s *claudeStreamRuntime) finalize(stopReason string) {
 
 func (s *claudeStreamRuntime) onFinalize(reason streamengine.StopReason, scannerErr error) {
 	if string(reason) == "upstream_error" {
+		if s.history != nil {
+			s.history.Error(500, s.upstreamErr, "upstream_error", responsehistory.ThinkingForArchive(s.rawThinking.String(), s.toolDetectionThinking.String(), s.thinking.String()), responsehistory.TextForArchive(s.rawText.String(), s.text.String()))
+		}
 		s.sendError(s.upstreamErr)
 		return
 	}
 	if scannerErr != nil {
+		if s.history != nil {
+			s.history.Error(500, scannerErr.Error(), "error", responsehistory.ThinkingForArchive(s.rawThinking.String(), s.toolDetectionThinking.String(), s.thinking.String()), responsehistory.TextForArchive(s.rawText.String(), s.text.String()))
+		}
 		s.sendError(scannerErr.Error())
 		return
 	}

internal/httpapi/gemini/handler_generate.go

@@ -2,6 +2,7 @@ package gemini
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"io"
@@ -14,8 +15,10 @@ import (
 	"ds2api/internal/assistantturn"
 	"ds2api/internal/auth"
 	"ds2api/internal/completionruntime"
+	"ds2api/internal/httpapi/openai/history"
 	"ds2api/internal/httpapi/requestbody"
 	"ds2api/internal/promptcompat"
+	"ds2api/internal/responsehistory"
 	"ds2api/internal/sse"
 	"ds2api/internal/toolcall"
 	"ds2api/internal/translatorcliproxy"
@@ -76,33 +79,65 @@ func (h *Handler) handleGeminiDirect(w http.ResponseWriter, r *http.Request, str
 		return true
 	}
 	defer h.Auth.Release(a)
+	stdReq, err = h.applyCurrentInputFile(r.Context(), a, stdReq)
+	if err != nil {
+		status, message := mapCurrentInputFileError(err)
+		writeGeminiError(w, status, message)
+		return true
+	}
+	historySession := responsehistory.Start(responsehistory.StartParams{
+		Store:    h.ChatHistory,
+		Request:  r,
+		Auth:     a,
+		Surface:  "gemini.generate_content",
+		Standard: stdReq,
+	})
 	if stream {
-		h.handleGeminiDirectStream(w, r, a, stdReq)
+		h.handleGeminiDirectStream(w, r, a, stdReq, historySession)
 		return true
 	}
 	result, outErr := completionruntime.ExecuteNonStreamWithRetry(r.Context(), h.DS, a, stdReq, completionruntime.Options{
-		StripReferenceMarkers: stripReferenceMarkersEnabled(),
-		RetryEnabled:          true,
-		CurrentInputFile:      h.Store,
+		RetryEnabled:     true,
+		CurrentInputFile: h.Store,
 	})
 	if outErr != nil {
+		if historySession != nil {
+			historySession.ErrorTurn(outErr.Status, outErr.Message, outErr.Code, result.Turn)
+		}
 		writeGeminiError(w, outErr.Status, outErr.Message)
 		return true
 	}
+	if historySession != nil {
+		historySession.SuccessTurn(http.StatusOK, result.Turn, responsehistory.GenericUsage(result.Turn))
+	}
 	writeJSON(w, http.StatusOK, buildGeminiGenerateContentResponseFromTurn(result.Turn))
 	return true
 }
 
-func (h *Handler) handleGeminiDirectStream(w http.ResponseWriter, r *http.Request, a *auth.RequestAuth, stdReq promptcompat.StandardRequest) {
+func (h *Handler) applyCurrentInputFile(ctx context.Context, a *auth.RequestAuth, stdReq promptcompat.StandardRequest) (promptcompat.StandardRequest, error) {
+	if h == nil {
+		return stdReq, nil
+	}
+	return (history.Service{Store: h.Store, DS: h.DS}).ApplyCurrentInputFile(ctx, a, stdReq)
+}
+
+func mapCurrentInputFileError(err error) (int, string) {
+	return history.MapError(err)
+}
+
+func (h *Handler) handleGeminiDirectStream(w http.ResponseWriter, r *http.Request, a *auth.RequestAuth, stdReq promptcompat.StandardRequest, historySession *responsehistory.Session) {
 	start, outErr := completionruntime.StartCompletion(r.Context(), h.DS, a, stdReq, completionruntime.Options{
 		CurrentInputFile: h.Store,
 	})
 	if outErr != nil {
+		if historySession != nil {
+			historySession.Error(outErr.Status, outErr.Message, outErr.Code, "", "")
+		}
 		writeGeminiError(w, outErr.Status, outErr.Message)
 		return
 	}
 	streamReq := start.Request
-	h.handleStreamGenerateContent(w, r, start.Response, streamReq.ResponseModel, streamReq.PromptTokenText, streamReq.Thinking, streamReq.Search, streamReq.ToolNames, streamReq.ToolsRaw)
+	h.handleStreamGenerateContent(w, r, start.Response, streamReq.ResponseModel, streamReq.PromptTokenText, streamReq.Thinking, streamReq.Search, streamReq.ToolNames, streamReq.ToolsRaw, historySession)
 }
 
 func (h *Handler) proxyViaOpenAI(w http.ResponseWriter, r *http.Request, stream bool) bool {
@@ -294,12 +329,11 @@ func (h *Handler) handleNonStreamGenerateContent(w http.ResponseWriter, resp *ht
 	}
 
 	result := sse.CollectStream(resp, thinkingEnabled, true)
-	stripReferenceMarkers := stripReferenceMarkersEnabled()
 	writeJSON(w, http.StatusOK, buildGeminiGenerateContentResponse(
 		model,
 		finalPrompt,
-		cleanVisibleOutput(result.Thinking, stripReferenceMarkers),
-		cleanVisibleOutput(result.Text, stripReferenceMarkers),
+		cleanVisibleOutput(result.Thinking, false),
+		cleanVisibleOutput(result.Text, false),
 		toolNames,
 	))
 }

internal/httpapi/gemini/handler_routes.go

@@ -5,6 +5,7 @@ import (
 
 	"github.com/go-chi/chi/v5"
 
+	"ds2api/internal/chathistory"
 	"ds2api/internal/textclean"
 	"ds2api/internal/util"
 )
@@ -12,10 +13,11 @@ import (
 var writeJSON = util.WriteJSON
 
 type Handler struct {
-	Store  ConfigReader
-	Auth   AuthResolver
-	DS     DeepSeekCaller
-	OpenAI OpenAIChatRunner
+	Store       ConfigReader
+	Auth        AuthResolver
+	DS          DeepSeekCaller
+	OpenAI      OpenAIChatRunner
+	ChatHistory *chathistory.Store
 }
 
 //nolint:unused // used by native Gemini stream/non-stream runtime helpers.

internal/httpapi/gemini/handler_stream_runtime.go

@@ -9,15 +9,23 @@ import (
 
 	"ds2api/internal/assistantturn"
 	dsprotocol "ds2api/internal/deepseek/protocol"
+	"ds2api/internal/responsehistory"
 	"ds2api/internal/sse"
 	streamengine "ds2api/internal/stream"
 )
 
 //nolint:unused // retained for native Gemini stream handling path.
-func (h *Handler) handleStreamGenerateContent(w http.ResponseWriter, r *http.Request, resp *http.Response, model, finalPrompt string, thinkingEnabled, searchEnabled bool, toolNames []string, toolsRaw any) {
+func (h *Handler) handleStreamGenerateContent(w http.ResponseWriter, r *http.Request, resp *http.Response, model, finalPrompt string, thinkingEnabled, searchEnabled bool, toolNames []string, toolsRaw any, historySessions ...*responsehistory.Session) {
+	var historySession *responsehistory.Session
+	if len(historySessions) > 0 {
+		historySession = historySessions[0]
+	}
 	defer func() { _ = resp.Body.Close() }()
 	if resp.StatusCode != http.StatusOK {
 		body, _ := io.ReadAll(resp.Body)
+		if historySession != nil {
+			historySession.Error(resp.StatusCode, strings.TrimSpace(string(body)), "error", "", "")
+		}
 		writeGeminiError(w, resp.StatusCode, strings.TrimSpace(string(body)))
 		return
 	}
@@ -29,7 +37,7 @@ func (h *Handler) handleStreamGenerateContent(w http.ResponseWriter, r *http.Req
 
 	rc := http.NewResponseController(w)
 	_, canFlush := w.(http.Flusher)
-	runtime := newGeminiStreamRuntime(w, rc, canFlush, model, finalPrompt, thinkingEnabled, searchEnabled, stripReferenceMarkersEnabled(), toolNames, toolsRaw)
+	runtime := newGeminiStreamRuntime(w, rc, canFlush, model, finalPrompt, thinkingEnabled, searchEnabled, stripReferenceMarkersEnabled(), toolNames, toolsRaw, historySession)
 
 	initialType := "text"
 	if thinkingEnabled {
@@ -70,6 +78,7 @@ type geminiStreamRuntime struct {
 	accumulator       *assistantturn.Accumulator
 	contentFilter     bool
 	responseMessageID int
+	history           *responsehistory.Session
 }
 
 //nolint:unused // retained for native Gemini stream handling path.
@@ -84,6 +93,7 @@ func newGeminiStreamRuntime(
 	stripReferenceMarkers bool,
 	toolNames []string,
 	toolsRaw any,
+	history *responsehistory.Session,
 ) *geminiStreamRuntime {
 	return &geminiStreamRuntime{
 		w:                     w,
@@ -97,6 +107,7 @@ func newGeminiStreamRuntime(
 		stripReferenceMarkers: stripReferenceMarkers,
 		toolNames:             toolNames,
 		toolsRaw:              toolsRaw,
+		history:               history,
 		accumulator: assistantturn.NewAccumulator(assistantturn.AccumulatorOptions{
 			ThinkingEnabled:       thinkingEnabled,
 			SearchEnabled:         searchEnabled,
@@ -170,6 +181,13 @@ func (s *geminiStreamRuntime) onParsed(parsed sse.LineResult) streamengine.Parse
 			"modelVersion": s.model,
 		})
 	}
+	if s.history != nil {
+		rawText, text, rawThinking, thinking, detectionThinking := s.accumulator.Snapshot()
+		s.history.Progress(
+			responsehistory.ThinkingForArchive(rawThinking, detectionThinking, thinking),
+			responsehistory.TextForArchive(rawText, text),
+		)
+	}
 	return streamengine.ParsedDecision{ContentSeen: accumulated.ContentSeen}
 }
 
@@ -193,6 +211,15 @@ func (s *geminiStreamRuntime) finalize() {
 		ToolsRaw:              s.toolsRaw,
 	})
 	outcome := assistantturn.FinalizeTurn(turn, assistantturn.FinalizeOptions{})
+	if s.history != nil {
+		s.history.Success(
+			http.StatusOK,
+			responsehistory.ThinkingForArchive(turn.RawThinking, turn.DetectionThinking, turn.Thinking),
+			responsehistory.TextForArchive(turn.RawText, turn.Text),
+			assistantturn.FinishReason(turn),
+			responsehistory.GenericUsage(turn),
+		)
+	}
 
 	if s.bufferContent {
 		parts := buildGeminiPartsFromTurn(turn)

internal/httpapi/gemini/handler_test.go

@@ -7,12 +7,14 @@ import (
 	"io"
 	"net/http"
 	"net/http/httptest"
+	"path/filepath"
 	"strings"
 	"testing"
 
 	"github.com/go-chi/chi/v5"
 
 	"ds2api/internal/auth"
+	"ds2api/internal/chathistory"
 	dsclient "ds2api/internal/deepseek/client"
 )
 
@@ -138,10 +140,12 @@ func TestGeminiDirectAppliesCurrentInputFile(t *testing.T) {
 	ds := &testGeminiDS{
 		resp: makeGeminiUpstreamResponse(`data: {"p":"response/content","v":"ok"}`),
 	}
+	historyStore := chathistory.New(filepath.Join(t.TempDir(), "history.json"))
 	h := &Handler{
-		Store: testGeminiConfig{},
-		Auth:  testGeminiAuth{},
-		DS:    ds,
+		Store:       testGeminiConfig{},
+		Auth:        testGeminiAuth{},
+		DS:          ds,
+		ChatHistory: historyStore,
 	}
 	reqBody := `{"contents":[{"role":"user","parts":[{"text":"hello from gemini"}]}]}`
 	req := httptest.NewRequest(http.MethodPost, "/v1beta/models/gemini-2.5-pro:generateContent", strings.NewReader(reqBody))
@@ -172,6 +176,29 @@ func TestGeminiDirectAppliesCurrentInputFile(t *testing.T) {
 	if !strings.Contains(prompt, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
 		t.Fatalf("expected continuation prompt, got %q", prompt)
 	}
+	snapshot, err := historyStore.Snapshot()
+	if err != nil {
+		t.Fatalf("snapshot history: %v", err)
+	}
+	if len(snapshot.Items) != 1 {
+		t.Fatalf("expected one history item, got %d", len(snapshot.Items))
+	}
+	full, err := historyStore.Get(snapshot.Items[0].ID)
+	if err != nil {
+		t.Fatalf("get history item: %v", err)
+	}
+	if full.Surface != "gemini.generate_content" {
+		t.Fatalf("unexpected surface: %q", full.Surface)
+	}
+	if full.Content != "ok" {
+		t.Fatalf("expected raw upstream content, got %q", full.Content)
+	}
+	if full.HistoryText != string(ds.uploadCalls[0].Data) {
+		t.Fatalf("expected uploaded current input file to be persisted in history text")
+	}
+	if len(full.Messages) != 1 || !strings.Contains(full.Messages[0].Content, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
+		t.Fatalf("expected persisted message to match upstream continuation prompt, got %#v", full.Messages)
+	}
 }
 
 func TestGeminiRoutesRegistered(t *testing.T) {

internal/httpapi/openai/chat/chat_history.go

@@ -14,9 +14,6 @@ import (
 	"ds2api/internal/promptcompat"
 )
 
-const adminWebUISourceHeader = "X-Ds2-Source"
-const adminWebUISourceValue = "admin-webui-api-tester"
-
 type chatHistorySession struct {
 	store       *chathistory.Store
 	entryID     string
@@ -40,6 +37,7 @@ func startChatHistory(store *chathistory.Store, r *http.Request, a *auth.Request
 	entry, err := store.Start(chathistory.StartParams{
 		CallerID:    strings.TrimSpace(a.CallerID),
 		AccountID:   strings.TrimSpace(a.AccountID),
+		Surface:     "openai.chat_completions",
 		Model:       strings.TrimSpace(stdReq.ResponseModel),
 		Stream:      stdReq.Stream,
 		UserInput:   extractSingleUserInput(stdReq.Messages),
@@ -50,6 +48,7 @@ func startChatHistory(store *chathistory.Store, r *http.Request, a *auth.Request
 	startParams := chathistory.StartParams{
 		CallerID:    strings.TrimSpace(a.CallerID),
 		AccountID:   strings.TrimSpace(a.AccountID),
+		Surface:     "openai.chat_completions",
 		Model:       strings.TrimSpace(stdReq.ResponseModel),
 		Stream:      stdReq.Stream,
 		UserInput:   extractSingleUserInput(stdReq.Messages),
@@ -82,7 +81,7 @@ func shouldCaptureChatHistory(r *http.Request) bool {
 	if isVercelStreamPrepareRequest(r) || isVercelStreamReleaseRequest(r) {
 		return false
 	}
-	return strings.TrimSpace(r.Header.Get(adminWebUISourceHeader)) != adminWebUISourceValue
+	return true
 }
 
 func extractSingleUserInput(messages []any) string {
@@ -188,6 +187,23 @@ func (s *chatHistorySession) stopped(thinking, content, finishReason string) {
 	})
 }
 
+func historyTextForArchive(raw, visible string) string {
+	if strings.TrimSpace(raw) != "" {
+		return raw
+	}
+	return visible
+}
+
+func historyThinkingForArchive(raw, detection, visible string) string {
+	if strings.TrimSpace(raw) != "" {
+		return raw
+	}
+	if strings.TrimSpace(detection) != "" {
+		return detection
+	}
+	return visible
+}
+
 func (s *chatHistorySession) retryMissingEntry() bool {
 	if s == nil || s.store == nil || s.disabled {
 		return false

internal/httpapi/openai/chat/chat_history_test.go

@@ -6,6 +6,7 @@ import (
 	"net/http/httptest"
 	"os"
 	"path/filepath"
+	"strconv"
 	"strings"
 	"sync"
 	"testing"
@@ -102,6 +103,86 @@ func TestChatCompletionsNonStreamPersistsHistory(t *testing.T) {
 	}
 }
 
+func TestChatHistoryNonStreamArchivesRawToolCallMarkup(t *testing.T) {
+	historyStore := newTestChatHistoryStore(t)
+	entry, err := historyStore.Start(chathistory.StartParams{
+		CallerID:  "caller:test",
+		Model:     "deepseek-v4-flash",
+		UserInput: "call tool",
+	})
+	if err != nil {
+		t.Fatalf("start history failed: %v", err)
+	}
+	session := &chatHistorySession{
+		store:       historyStore,
+		entryID:     entry.ID,
+		startedAt:   time.Now(),
+		lastPersist: time.Now().Add(-time.Second),
+		finalPrompt: "call tool",
+	}
+	rawToolCall := `<tool_calls><invoke name="search"><parameter name="q">golang</parameter></invoke></tool_calls>`
+
+	h := &Handler{}
+	rec := httptest.NewRecorder()
+	resp := makeOpenAISSEHTTPResponse(`data: {"p":"response/content","v":`+strconv.Quote(rawToolCall)+`}`, `data: [DONE]`)
+	h.handleNonStream(rec, resp, "cid-tool-history", "deepseek-v4-flash", "prompt", 0, false, false, []string{"search"}, nil, session)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("expected 200, got %d body=%s", rec.Code, rec.Body.String())
+	}
+	full, err := historyStore.Get(entry.ID)
+	if err != nil {
+		t.Fatalf("get detail failed: %v", err)
+	}
+	if full.Content != rawToolCall {
+		t.Fatalf("expected raw tool markup archived, got %q", full.Content)
+	}
+	if full.FinishReason != "tool_calls" {
+		t.Fatalf("expected tool_calls finish reason, got %#v", full.FinishReason)
+	}
+}
+
+func TestChatHistoryStreamArchivesRawToolCallMarkup(t *testing.T) {
+	historyStore := newTestChatHistoryStore(t)
+	entry, err := historyStore.Start(chathistory.StartParams{
+		CallerID:  "caller:test",
+		Model:     "deepseek-v4-flash",
+		Stream:    true,
+		UserInput: "call tool",
+	})
+	if err != nil {
+		t.Fatalf("start history failed: %v", err)
+	}
+	session := &chatHistorySession{
+		store:       historyStore,
+		entryID:     entry.ID,
+		startedAt:   time.Now(),
+		lastPersist: time.Now().Add(-time.Second),
+		finalPrompt: "call tool",
+	}
+	rawToolCall := `<tool_calls><invoke name="search"><parameter name="q">golang</parameter></invoke></tool_calls>`
+
+	h := &Handler{}
+	req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", nil)
+	rec := httptest.NewRecorder()
+	resp := makeOpenAISSEHTTPResponse(`data: {"p":"response/content","v":`+strconv.Quote(rawToolCall)+`}`, `data: [DONE]`)
+	h.handleStream(rec, req, resp, "cid-stream-tool-history", "deepseek-v4-flash", "prompt", 0, false, false, []string{"search"}, nil, session)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("expected 200, got %d body=%s", rec.Code, rec.Body.String())
+	}
+	full, err := historyStore.Get(entry.ID)
+	if err != nil {
+		t.Fatalf("get detail failed: %v", err)
+	}
+	if full.Content != rawToolCall {
+		t.Fatalf("expected raw streamed tool markup archived, got %q", full.Content)
+	}
+	if full.FinishReason != "tool_calls" {
+		t.Fatalf("expected tool_calls finish reason, got %#v", full.FinishReason)
+	}
+}
+
 func TestStartChatHistoryRecoversFromTransientWriteFailure(t *testing.T) {
 	historyStore := newTestChatHistoryStore(t)
 	restore := blockChatHistoryDetailDir(t, historyStore.DetailDir())
@@ -213,7 +294,7 @@ func TestHandleStreamContextCancelledMarksHistoryStopped(t *testing.T) {
 	}
 }
 
-func TestChatCompletionsSkipsAdminWebUISource(t *testing.T) {
+func TestChatCompletionsRecordsAdminWebUISource(t *testing.T) {
 	historyStore := newTestChatHistoryStore(t)
 	h := &Handler{
 		Store:       mockOpenAIConfig{},
@@ -226,7 +307,7 @@ func TestChatCompletionsSkipsAdminWebUISource(t *testing.T) {
 	req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", strings.NewReader(reqBody))
 	req.Header.Set("Authorization", "Bearer direct-token")
 	req.Header.Set("Content-Type", "application/json")
-	req.Header.Set(adminWebUISourceHeader, adminWebUISourceValue)
+	req.Header.Set("X-Ds2-Source", "admin-webui-api-tester")
 	rec := httptest.NewRecorder()
 	h.ChatCompletions(rec, req)
 
@@ -237,8 +318,8 @@ func TestChatCompletionsSkipsAdminWebUISource(t *testing.T) {
 	if err != nil {
 		t.Fatalf("snapshot failed: %v", err)
 	}
-	if len(snapshot.Items) != 0 {
-		t.Fatalf("expected admin webui source to be skipped, got %#v", snapshot.Items)
+	if len(snapshot.Items) != 1 {
+		t.Fatalf("expected admin webui source to be recorded, got %#v", snapshot.Items)
 	}
 }
 

internal/httpapi/openai/chat/chat_stream_runtime.go

@@ -195,6 +195,24 @@ func (s *chatStreamRuntime) markContextCancelled() {
 	s.finalFinishReason = string(streamengine.StopReasonContextCancelled)
 }
 
+func (s *chatStreamRuntime) historyText() string {
+	if s == nil {
+		return ""
+	}
+	return historyTextForArchive(s.accumulator.RawText.String(), s.finalText)
+}
+
+func (s *chatStreamRuntime) historyThinking() string {
+	if s == nil {
+		return ""
+	}
+	return historyThinkingForArchive(
+		s.accumulator.RawThinking.String(),
+		s.accumulator.ToolDetectionThinking.String(),
+		s.finalThinking,
+	)
+}
+
 func (s *chatStreamRuntime) resetStreamToolCallState() {
 	s.streamToolCallIDs = map[int]string{}
 	s.streamToolNames = map[int]string{}

internal/httpapi/openai/chat/empty_retry_runtime.go

@@ -31,6 +31,14 @@ type chatNonStreamResult struct {
 	outputError           *assistantturn.OutputError
 }
 
+func (r chatNonStreamResult) historyText() string {
+	return historyTextForArchive(r.rawText, r.text)
+}
+
+func (r chatNonStreamResult) historyThinking() string {
+	return historyThinkingForArchive(r.rawThinking, r.toolDetectionThinking, r.thinking)
+}
+
 func (h *Handler) handleNonStreamWithRetry(w http.ResponseWriter, ctx context.Context, a *auth.RequestAuth, resp *http.Response, payload map[string]any, pow, completionID, model, finalPrompt string, refFileTokens int, thinkingEnabled, searchEnabled bool, toolNames []string, toolsRaw any, historySession *chatHistorySession) {
 	attempts := 0
 	currentResp := resp
@@ -70,7 +78,7 @@ func (h *Handler) handleNonStreamWithRetry(w http.ResponseWriter, ctx context.Co
 		nextResp, err := h.DS.CallCompletion(ctx, a, retryPayload, retryPow, 3)
 		if err != nil {
 			if historySession != nil {
-				historySession.error(http.StatusInternalServerError, "Failed to get completion.", "error", result.thinking, result.text)
+				historySession.error(http.StatusInternalServerError, "Failed to get completion.", "error", result.historyThinking(), result.historyText())
 			}
 			writeOpenAIError(w, http.StatusInternalServerError, "Failed to get completion.")
 			config.Logger.Warn("[openai_empty_retry] retry request failed", "surface", "chat.completions", "stream", false, "retry_attempt", attempts, "error", err)
@@ -90,12 +98,11 @@ func (h *Handler) collectChatNonStreamAttempt(w http.ResponseWriter, resp *http.
 	}
 	result := sse.CollectStream(resp, thinkingEnabled, true)
 	turn := assistantturn.BuildTurnFromCollected(result, assistantturn.BuildOptions{
-		Model:                 model,
-		Prompt:                usagePrompt,
-		SearchEnabled:         searchEnabled,
-		StripReferenceMarkers: stripReferenceMarkersEnabled(),
-		ToolNames:             toolNames,
-		ToolsRaw:              toolsRaw,
+		Model:         model,
+		Prompt:        usagePrompt,
+		SearchEnabled: searchEnabled,
+		ToolNames:     toolNames,
+		ToolsRaw:      toolsRaw,
 	})
 	respBody := openaifmt.BuildChatCompletionWithToolCalls(completionID, model, usagePrompt, turn.Thinking, turn.Text, turn.ToolCalls, toolsRaw)
 	return chatNonStreamResult{
@@ -120,14 +127,14 @@ func (h *Handler) finishChatNonStreamResult(w http.ResponseWriter, result chatNo
 			status, message, code = result.outputError.Status, result.outputError.Message, result.outputError.Code
 		}
 		if historySession != nil {
-			historySession.error(status, message, code, result.thinking, result.text)
+			historySession.error(status, message, code, result.historyThinking(), result.historyText())
 		}
 		writeOpenAIErrorWithCode(w, status, message, code)
 		config.Logger.Info("[openai_empty_retry] terminal empty output", "surface", "chat.completions", "stream", false, "retry_attempts", attempts, "success_source", "none", "content_filter", result.contentFilter)
 		return
 	}
 	if historySession != nil {
-		historySession.success(http.StatusOK, result.thinking, result.text, result.finishReason, openaifmt.BuildChatUsageForModel("", usagePrompt, result.thinking, result.text, refFileTokens))
+		historySession.success(http.StatusOK, result.historyThinking(), result.historyText(), result.finishReason, openaifmt.BuildChatUsageForModel("", usagePrompt, result.thinking, result.text, refFileTokens))
 	}
 	writeJSON(w, http.StatusOK, result.body)
 	source := "first_attempt"
@@ -246,7 +253,7 @@ func (h *Handler) consumeChatStreamAttempt(r *http.Request, resp *http.Response,
 		OnParsed: func(parsed sse.LineResult) streamengine.ParsedDecision {
 			decision := streamRuntime.onParsed(parsed)
 			if historySession != nil {
-				historySession.progress(streamRuntime.accumulator.Thinking.String(), streamRuntime.accumulator.Text.String())
+				historySession.progress(streamRuntime.historyThinking(), streamRuntime.historyText())
 			}
 			return decision
 		},
@@ -258,7 +265,7 @@ func (h *Handler) consumeChatStreamAttempt(r *http.Request, resp *http.Response,
 		OnContextDone: func() {
 			streamRuntime.markContextCancelled()
 			if historySession != nil {
-				historySession.stopped(streamRuntime.accumulator.Thinking.String(), streamRuntime.accumulator.Text.String(), string(streamengine.StopReasonContextCancelled))
+				historySession.stopped(streamRuntime.historyThinking(), streamRuntime.historyText(), string(streamengine.StopReasonContextCancelled))
 			}
 		},
 	})
@@ -278,16 +285,16 @@ func recordChatStreamHistory(streamRuntime *chatStreamRuntime, historySession *c
 		return
 	}
 	if streamRuntime.finalErrorMessage != "" {
-		historySession.error(streamRuntime.finalErrorStatus, streamRuntime.finalErrorMessage, streamRuntime.finalErrorCode, streamRuntime.accumulator.Thinking.String(), streamRuntime.accumulator.Text.String())
+		historySession.error(streamRuntime.finalErrorStatus, streamRuntime.finalErrorMessage, streamRuntime.finalErrorCode, streamRuntime.historyThinking(), streamRuntime.historyText())
 		return
 	}
-	historySession.success(http.StatusOK, streamRuntime.finalThinking, streamRuntime.finalText, streamRuntime.finalFinishReason, streamRuntime.finalUsage)
+	historySession.success(http.StatusOK, streamRuntime.historyThinking(), streamRuntime.historyText(), streamRuntime.finalFinishReason, streamRuntime.finalUsage)
 }
 
 func failChatStreamRetry(streamRuntime *chatStreamRuntime, historySession *chatHistorySession, status int, message, code string) {
 	streamRuntime.sendFailedChunk(status, message, code)
 	if historySession != nil {
-		historySession.error(status, message, code, streamRuntime.accumulator.Thinking.String(), streamRuntime.accumulator.Text.String())
+		historySession.error(status, message, code, streamRuntime.historyThinking(), streamRuntime.historyText())
 	}
 }
 

internal/httpapi/openai/chat/handler_chat.go

@@ -80,14 +80,13 @@ func (h *Handler) ChatCompletions(w http.ResponseWriter, r *http.Request) {
 
 	if !stdReq.Stream {
 		result, outErr := completionruntime.ExecuteNonStreamWithRetry(r.Context(), h.DS, a, stdReq, completionruntime.Options{
-			StripReferenceMarkers: stripReferenceMarkersEnabled(),
-			RetryEnabled:          true,
-			CurrentInputFile:      h.Store,
+			RetryEnabled:     true,
+			CurrentInputFile: h.Store,
 		})
 		sessionID = result.SessionID
 		if outErr != nil {
 			if historySession != nil {
-				historySession.error(outErr.Status, outErr.Message, outErr.Code, result.Turn.Thinking, result.Turn.Text)
+				historySession.error(outErr.Status, outErr.Message, outErr.Code, historyThinkingForArchive(result.Turn.RawThinking, result.Turn.DetectionThinking, result.Turn.Thinking), historyTextForArchive(result.Turn.RawText, result.Turn.Text))
 			}
 			writeOpenAIErrorWithCode(w, outErr.Status, outErr.Message, outErr.Code)
 			return
@@ -96,7 +95,7 @@ func (h *Handler) ChatCompletions(w http.ResponseWriter, r *http.Request) {
 		respBody["usage"] = assistantturn.OpenAIChatUsage(result.Turn)
 		finishReason := assistantturn.FinalizeTurn(result.Turn, assistantturn.FinalizeOptions{}).FinishReason
 		if historySession != nil {
-			historySession.success(http.StatusOK, result.Turn.Thinking, result.Turn.Text, finishReason, assistantturn.OpenAIChatUsage(result.Turn))
+			historySession.success(http.StatusOK, historyThinkingForArchive(result.Turn.RawThinking, result.Turn.DetectionThinking, result.Turn.Thinking), historyTextForArchive(result.Turn.RawText, result.Turn.Text), finishReason, assistantturn.OpenAIChatUsage(result.Turn))
 		}
 		writeJSON(w, http.StatusOK, respBody)
 		return
@@ -164,20 +163,19 @@ func (h *Handler) handleNonStream(w http.ResponseWriter, resp *http.Response, co
 	result := sse.CollectStream(resp, thinkingEnabled, true)
 
 	turn := assistantturn.BuildTurnFromCollected(result, assistantturn.BuildOptions{
-		Model:                 model,
-		Prompt:                finalPrompt,
-		RefFileTokens:         refFileTokens,
-		SearchEnabled:         searchEnabled,
-		StripReferenceMarkers: stripReferenceMarkersEnabled(),
-		ToolNames:             toolNames,
-		ToolsRaw:              toolsRaw,
-		ToolChoice:            promptcompat.DefaultToolChoicePolicy(),
+		Model:         model,
+		Prompt:        finalPrompt,
+		RefFileTokens: refFileTokens,
+		SearchEnabled: searchEnabled,
+		ToolNames:     toolNames,
+		ToolsRaw:      toolsRaw,
+		ToolChoice:    promptcompat.DefaultToolChoicePolicy(),
 	})
 	outcome := assistantturn.FinalizeTurn(turn, assistantturn.FinalizeOptions{})
 	if outcome.ShouldFail {
 		status, message, code := outcome.Error.Status, outcome.Error.Message, outcome.Error.Code
 		if historySession != nil {
-			historySession.error(status, message, code, turn.Thinking, turn.Text)
+			historySession.error(status, message, code, historyThinkingForArchive(turn.RawThinking, turn.DetectionThinking, turn.Thinking), historyTextForArchive(turn.RawText, turn.Text))
 		}
 		writeOpenAIErrorWithCode(w, status, message, code)
 		return
@@ -185,7 +183,7 @@ func (h *Handler) handleNonStream(w http.ResponseWriter, resp *http.Response, co
 	respBody := openaifmt.BuildChatCompletionWithToolCalls(completionID, model, finalPrompt, turn.Thinking, turn.Text, turn.ToolCalls, toolsRaw)
 	respBody["usage"] = assistantturn.OpenAIChatUsage(turn)
 	if historySession != nil {
-		historySession.success(http.StatusOK, turn.Thinking, turn.Text, outcome.FinishReason, assistantturn.OpenAIChatUsage(turn))
+		historySession.success(http.StatusOK, historyThinkingForArchive(turn.RawThinking, turn.DetectionThinking, turn.Thinking), historyTextForArchive(turn.RawText, turn.Text), outcome.FinishReason, assistantturn.OpenAIChatUsage(turn))
 	}
 	writeJSON(w, http.StatusOK, respBody)
 }
@@ -253,7 +251,7 @@ func (h *Handler) handleStream(w http.ResponseWriter, r *http.Request, resp *htt
 		OnParsed: func(parsed sse.LineResult) streamengine.ParsedDecision {
 			decision := streamRuntime.onParsed(parsed)
 			if historySession != nil {
-				historySession.progress(streamRuntime.accumulator.Thinking.String(), streamRuntime.accumulator.Text.String())
+				historySession.progress(streamRuntime.historyThinking(), streamRuntime.historyText())
 			}
 			return decision
 		},
@@ -267,14 +265,15 @@ func (h *Handler) handleStream(w http.ResponseWriter, r *http.Request, resp *htt
 				return
 			}
 			if streamRuntime.finalErrorMessage != "" {
-				historySession.error(streamRuntime.finalErrorStatus, streamRuntime.finalErrorMessage, streamRuntime.finalErrorCode, streamRuntime.accumulator.Thinking.String(), streamRuntime.accumulator.Text.String())
+				historySession.error(streamRuntime.finalErrorStatus, streamRuntime.finalErrorMessage, streamRuntime.finalErrorCode, streamRuntime.historyThinking(), streamRuntime.historyText())
 				return
 			}
-			historySession.success(http.StatusOK, streamRuntime.finalThinking, streamRuntime.finalText, streamRuntime.finalFinishReason, streamRuntime.finalUsage)
+			historySession.success(http.StatusOK, streamRuntime.historyThinking(), streamRuntime.historyText(), streamRuntime.finalFinishReason, streamRuntime.finalUsage)
 		},
 		OnContextDone: func() {
+			streamRuntime.markContextCancelled()
 			if historySession != nil {
-				historySession.stopped(streamRuntime.accumulator.Thinking.String(), streamRuntime.accumulator.Text.String(), string(streamengine.StopReasonContextCancelled))
+				historySession.stopped(streamRuntime.historyThinking(), streamRuntime.historyText(), string(streamengine.StopReasonContextCancelled))
 			}
 		},
 	})

internal/httpapi/openai/responses/empty_retry_runtime.go

@@ -10,11 +10,12 @@ import (
 	"ds2api/internal/config"
 	dsprotocol "ds2api/internal/deepseek/protocol"
 	"ds2api/internal/promptcompat"
+	"ds2api/internal/responsehistory"
 	streamengine "ds2api/internal/stream"
 )
 
-func (h *Handler) handleResponsesStreamWithRetry(w http.ResponseWriter, r *http.Request, a *auth.RequestAuth, resp *http.Response, payload map[string]any, pow, owner, responseID, model, finalPrompt string, refFileTokens int, thinkingEnabled, searchEnabled bool, toolNames []string, toolsRaw any, toolChoice promptcompat.ToolChoicePolicy, traceID string) {
-	streamRuntime, initialType, ok := h.prepareResponsesStreamRuntime(w, resp, owner, responseID, model, finalPrompt, refFileTokens, thinkingEnabled, searchEnabled, toolNames, toolsRaw, toolChoice, traceID)
+func (h *Handler) handleResponsesStreamWithRetry(w http.ResponseWriter, r *http.Request, a *auth.RequestAuth, resp *http.Response, payload map[string]any, pow, owner, responseID, model, finalPrompt string, refFileTokens int, thinkingEnabled, searchEnabled bool, toolNames []string, toolsRaw any, toolChoice promptcompat.ToolChoicePolicy, traceID string, historySession *responsehistory.Session) {
+	streamRuntime, initialType, ok := h.prepareResponsesStreamRuntime(w, resp, owner, responseID, model, finalPrompt, refFileTokens, thinkingEnabled, searchEnabled, toolNames, toolsRaw, toolChoice, traceID, historySession)
 	if !ok {
 		return
 	}
@@ -55,10 +56,13 @@ func (h *Handler) handleResponsesStreamWithRetry(w http.ResponseWriter, r *http.
 	}
 }
 
-func (h *Handler) prepareResponsesStreamRuntime(w http.ResponseWriter, resp *http.Response, owner, responseID, model, finalPrompt string, refFileTokens int, thinkingEnabled, searchEnabled bool, toolNames []string, toolsRaw any, toolChoice promptcompat.ToolChoicePolicy, traceID string) (*responsesStreamRuntime, string, bool) {
+func (h *Handler) prepareResponsesStreamRuntime(w http.ResponseWriter, resp *http.Response, owner, responseID, model, finalPrompt string, refFileTokens int, thinkingEnabled, searchEnabled bool, toolNames []string, toolsRaw any, toolChoice promptcompat.ToolChoicePolicy, traceID string, historySession *responsehistory.Session) (*responsesStreamRuntime, string, bool) {
 	if resp.StatusCode != http.StatusOK {
 		defer func() { _ = resp.Body.Close() }()
 		body, _ := io.ReadAll(resp.Body)
+		if historySession != nil {
+			historySession.Error(resp.StatusCode, strings.TrimSpace(string(body)), "error", "", "")
+		}
 		writeOpenAIError(w, resp.StatusCode, strings.TrimSpace(string(body)))
 		return nil, "", false
 	}
@@ -78,7 +82,7 @@ func (h *Handler) prepareResponsesStreamRuntime(w http.ResponseWriter, resp *htt
 		h.toolcallFeatureMatchEnabled() && h.toolcallEarlyEmitHighConfidence(),
 		toolChoice, traceID, func(obj map[string]any) {
 			h.getResponseStore().put(owner, responseID, obj)
-		},
+		}, historySession,
 	)
 	streamRuntime.refFileTokens = refFileTokens
 	streamRuntime.sendCreated()

internal/httpapi/openai/responses/empty_retry_runtime_test.go

@@ -47,6 +47,7 @@ func TestConsumeResponsesStreamAttemptMarksContextCancelledState(t *testing.T) {
 		promptcompat.DefaultToolChoicePolicy(),
 		"",
 		nil,
+		nil,
 	)
 	resp := makeResponsesOpenAISSEHTTPResponse(
 		`data: {"p":"response/content","v":"hello"}`,

internal/httpapi/openai/responses/responses_handler.go

@@ -18,6 +18,7 @@ import (
 	dsprotocol "ds2api/internal/deepseek/protocol"
 	openaifmt "ds2api/internal/format/openai"
 	"ds2api/internal/promptcompat"
+	"ds2api/internal/responsehistory"
 	"ds2api/internal/sse"
 	streamengine "ds2api/internal/stream"
 )
@@ -95,16 +96,28 @@ func (h *Handler) Responses(w http.ResponseWriter, r *http.Request) {
 	}
 
 	responseID := "resp_" + strings.ReplaceAll(uuid.NewString(), "-", "")
+	historySession := responsehistory.Start(responsehistory.StartParams{
+		Store:    h.ChatHistory,
+		Request:  r,
+		Auth:     a,
+		Surface:  "openai.responses",
+		Standard: stdReq,
+	})
 	if !stdReq.Stream {
 		result, outErr := completionruntime.ExecuteNonStreamWithRetry(r.Context(), h.DS, a, stdReq, completionruntime.Options{
-			StripReferenceMarkers: stripReferenceMarkersEnabled(),
-			RetryEnabled:          true,
-			CurrentInputFile:      h.Store,
+			RetryEnabled:     true,
+			CurrentInputFile: h.Store,
 		})
 		if outErr != nil {
+			if historySession != nil {
+				historySession.ErrorTurn(outErr.Status, outErr.Message, outErr.Code, result.Turn)
+			}
 			writeOpenAIErrorWithCode(w, outErr.Status, outErr.Message, outErr.Code)
 			return
 		}
+		if historySession != nil {
+			historySession.SuccessTurn(http.StatusOK, result.Turn, assistantturn.OpenAIResponsesUsage(result.Turn))
+		}
 		responseObj := openaifmt.BuildResponseObjectWithToolCalls(responseID, stdReq.ResponseModel, result.Turn.Prompt, result.Turn.Thinking, result.Turn.Text, result.Turn.ToolCalls, stdReq.ToolsRaw)
 		responseObj["usage"] = assistantturn.OpenAIResponsesUsage(result.Turn)
 		h.getResponseStore().put(owner, responseID, responseObj)
@@ -116,13 +129,16 @@ func (h *Handler) Responses(w http.ResponseWriter, r *http.Request) {
 		CurrentInputFile: h.Store,
 	})
 	if outErr != nil {
+		if historySession != nil {
+			historySession.Error(outErr.Status, outErr.Message, outErr.Code, "", "")
+		}
 		writeOpenAIErrorWithCode(w, outErr.Status, outErr.Message, outErr.Code)
 		return
 	}
 
 	streamReq := start.Request
 	refFileTokens := streamReq.RefFileTokens
-	h.handleResponsesStreamWithRetry(w, r, a, start.Response, start.Payload, start.Pow, owner, responseID, streamReq.ResponseModel, streamReq.PromptTokenText, refFileTokens, streamReq.Thinking, streamReq.Search, streamReq.ToolNames, streamReq.ToolsRaw, streamReq.ToolChoice, traceID)
+	h.handleResponsesStreamWithRetry(w, r, a, start.Response, start.Payload, start.Pow, owner, responseID, streamReq.ResponseModel, streamReq.PromptTokenText, refFileTokens, streamReq.Thinking, streamReq.Search, streamReq.ToolNames, streamReq.ToolsRaw, streamReq.ToolChoice, traceID, historySession)
 }
 
 func (h *Handler) handleResponsesNonStream(w http.ResponseWriter, resp *http.Response, owner, responseID, model, finalPrompt string, refFileTokens int, thinkingEnabled, searchEnabled bool, toolNames []string, toolsRaw any, toolChoice promptcompat.ToolChoicePolicy, traceID string) {
@@ -135,14 +151,13 @@ func (h *Handler) handleResponsesNonStream(w http.ResponseWriter, resp *http.Res
 	result := sse.CollectStream(resp, thinkingEnabled, true)
 
 	turn := assistantturn.BuildTurnFromCollected(result, assistantturn.BuildOptions{
-		Model:                 model,
-		Prompt:                finalPrompt,
-		RefFileTokens:         refFileTokens,
-		SearchEnabled:         searchEnabled,
-		StripReferenceMarkers: stripReferenceMarkersEnabled(),
-		ToolNames:             toolNames,
-		ToolsRaw:              toolsRaw,
-		ToolChoice:            toolChoice,
+		Model:         model,
+		Prompt:        finalPrompt,
+		RefFileTokens: refFileTokens,
+		SearchEnabled: searchEnabled,
+		ToolNames:     toolNames,
+		ToolsRaw:      toolsRaw,
+		ToolChoice:    toolChoice,
 	})
 	logResponsesToolPolicyRejection(traceID, toolChoice, turn.ParsedToolCalls, "text")
 	outcome := assistantturn.FinalizeTurn(turn, assistantturn.FinalizeOptions{})
@@ -198,6 +213,7 @@ func (h *Handler) handleResponsesStream(w http.ResponseWriter, r *http.Request,
 		func(obj map[string]any) {
 			h.getResponseStore().put(owner, responseID, obj)
 		},
+		nil,
 	)
 	streamRuntime.refFileTokens = refFileTokens
 	streamRuntime.sendCreated()

internal/httpapi/openai/responses/responses_history_test.go

@@ -0,0 +1,100 @@
+package responses
+
+import (
+	"context"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/go-chi/chi/v5"
+
+	"ds2api/internal/auth"
+	"ds2api/internal/chathistory"
+	dsclient "ds2api/internal/deepseek/client"
+)
+
+type responsesHistoryDS struct {
+	payload map[string]any
+}
+
+func (d *responsesHistoryDS) CreateSession(context.Context, *auth.RequestAuth, int) (string, error) {
+	return "session-id", nil
+}
+
+func (d *responsesHistoryDS) GetPow(context.Context, *auth.RequestAuth, int) (string, error) {
+	return "pow", nil
+}
+
+func (d *responsesHistoryDS) UploadFile(context.Context, *auth.RequestAuth, dsclient.UploadFileRequest, int) (*dsclient.UploadFileResult, error) {
+	return &dsclient.UploadFileResult{ID: "file-id"}, nil
+}
+
+func (d *responsesHistoryDS) CallCompletion(_ context.Context, _ *auth.RequestAuth, payload map[string]any, _ string, _ int) (*http.Response, error) {
+	d.payload = payload
+	return &http.Response{
+		StatusCode: http.StatusOK,
+		Header:     make(http.Header),
+		Body:       io.NopCloser(strings.NewReader("data: {\"p\":\"response/content\",\"v\":\"ok\"}\n")),
+	}, nil
+}
+
+func (d *responsesHistoryDS) DeleteSessionForToken(context.Context, string, string) (*dsclient.DeleteSessionResult, error) {
+	return &dsclient.DeleteSessionResult{Success: true}, nil
+}
+
+func (d *responsesHistoryDS) DeleteAllSessionsForToken(context.Context, string) error {
+	return nil
+}
+
+func TestResponsesRecordsResponseHistory(t *testing.T) {
+	store, resolver := newDirectTokenResolver(t)
+	historyStore := chathistory.New(filepath.Join(t.TempDir(), "history.json"))
+	ds := &responsesHistoryDS{}
+	h := &Handler{
+		Store:       store,
+		Auth:        resolver,
+		DS:          ds,
+		ChatHistory: historyStore,
+	}
+	r := chi.NewRouter()
+	RegisterRoutes(r, h)
+
+	req := httptest.NewRequest(http.MethodPost, "/v1/responses", strings.NewReader(`{"model":"deepseek-v4-flash","input":"hello responses"}`))
+	req.Header.Set("Authorization", "Bearer direct-token")
+	req.Header.Set("Content-Type", "application/json")
+	rec := httptest.NewRecorder()
+	r.ServeHTTP(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("expected 200, got %d body=%s", rec.Code, rec.Body.String())
+	}
+	if ds.payload == nil {
+		t.Fatalf("expected upstream payload to be sent")
+	}
+	snapshot, err := historyStore.Snapshot()
+	if err != nil {
+		t.Fatalf("snapshot history: %v", err)
+	}
+	if len(snapshot.Items) != 1 {
+		t.Fatalf("expected one history item, got %d", len(snapshot.Items))
+	}
+	item, err := historyStore.Get(snapshot.Items[0].ID)
+	if err != nil {
+		t.Fatalf("get history item: %v", err)
+	}
+	if item.Surface != "openai.responses" {
+		t.Fatalf("unexpected surface: %q", item.Surface)
+	}
+	if !strings.Contains(item.UserInput, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
+		t.Fatalf("unexpected user input: %q", item.UserInput)
+	}
+	if !strings.Contains(item.HistoryText, "hello responses") {
+		t.Fatalf("expected original input in persisted history text, got %q", item.HistoryText)
+	}
+	if item.Content != "ok" {
+		t.Fatalf("expected raw upstream content, got %q", item.Content)
+	}
+}

internal/httpapi/openai/responses/responses_stream_runtime_core.go

@@ -10,6 +10,7 @@ import (
 	openaifmt "ds2api/internal/format/openai"
 	"ds2api/internal/httpapi/openai/shared"
 	"ds2api/internal/promptcompat"
+	"ds2api/internal/responsehistory"
 	"ds2api/internal/sse"
 	streamengine "ds2api/internal/stream"
 	"ds2api/internal/toolstream"
@@ -61,6 +62,7 @@ type responsesStreamRuntime struct {
 	finalErrorCode    string
 
 	persistResponse func(obj map[string]any)
+	history         *responsehistory.Session
 }
 
 func newResponsesStreamRuntime(
@@ -80,6 +82,7 @@ func newResponsesStreamRuntime(
 	toolChoice promptcompat.ToolChoicePolicy,
 	traceID string,
 	persistResponse func(obj map[string]any),
+	history *responsehistory.Session,
 ) *responsesStreamRuntime {
 	return &responsesStreamRuntime{
 		w:                     w,
@@ -106,6 +109,7 @@ func newResponsesStreamRuntime(
 		toolChoice:            toolChoice,
 		traceID:               traceID,
 		persistResponse:       persistResponse,
+		history:               history,
 		accumulator: shared.StreamAccumulator{
 			ThinkingEnabled:       thinkingEnabled,
 			SearchEnabled:         searchEnabled,
@@ -138,6 +142,9 @@ func (s *responsesStreamRuntime) failResponse(status int, message, code string)
 	if s.persistResponse != nil {
 		s.persistResponse(failedResp)
 	}
+	if s.history != nil {
+		s.history.Error(status, message, code, responsehistory.ThinkingForArchive(s.accumulator.RawThinking.String(), s.accumulator.ToolDetectionThinking.String(), s.accumulator.Thinking.String()), responsehistory.TextForArchive(s.accumulator.RawText.String(), s.accumulator.Text.String()))
+	}
 	s.sendEvent("response.failed", openaifmt.BuildResponsesFailedPayload(s.responseID, s.model, status, message, code))
 	s.sendDone()
 }
@@ -214,6 +221,15 @@ func (s *responsesStreamRuntime) finalize(finishReason string, deferEmptyOutput
 	if s.persistResponse != nil {
 		s.persistResponse(obj)
 	}
+	if s.history != nil {
+		s.history.Success(
+			http.StatusOK,
+			responsehistory.ThinkingForArchive(turn.RawThinking, turn.DetectionThinking, turn.Thinking),
+			responsehistory.TextForArchive(turn.RawText, turn.Text),
+			outcome.FinishReason,
+			assistantturn.OpenAIResponsesUsage(turn),
+		)
+	}
 	s.sendEvent("response.completed", openaifmt.BuildResponsesCompletedPayload(obj))
 	s.sendDone()
 	return true
@@ -272,5 +288,11 @@ func (s *responsesStreamRuntime) onParsed(parsed sse.LineResult) streamengine.Pa
 	}
 
 	batch.flush()
+	if s.history != nil {
+		s.history.Progress(
+			responsehistory.ThinkingForArchive(s.accumulator.RawThinking.String(), s.accumulator.ToolDetectionThinking.String(), s.accumulator.Thinking.String()),
+			responsehistory.TextForArchive(s.accumulator.RawText.String(), s.accumulator.Text.String()),
+		)
+	}
 	return streamengine.ParsedDecision{ContentSeen: accumulated.ContentSeen}
 }

internal/httpapi/openai/shared/stream_accumulator.go

@@ -89,11 +89,11 @@ func (a *StreamAccumulator) applyTextPart(text string) StreamPartDelta {
 	}
 	a.RawText.WriteString(rawTrimmed)
 	delta := StreamPartDelta{Type: "text", RawText: rawTrimmed}
-	cleanedText := CleanVisibleOutput(rawTrimmed, a.StripReferenceMarkers)
-	if a.SearchEnabled && sse.IsCitation(cleanedText) {
+	if a.SearchEnabled && sse.IsCitation(rawTrimmed) {
 		delta.CitationOnly = true
 		return delta
 	}
+	cleanedText := CleanVisibleOutput(rawTrimmed, a.StripReferenceMarkers)
 	trimmed := sse.TrimContinuationOverlapFromBuilder(&a.Text, cleanedText)
 	if trimmed == "" {
 		return delta

internal/httpapi/openai/shared/stream_accumulator_test.go

@@ -95,3 +95,21 @@ func TestStreamAccumulatorSuppressesCitationTextWhenSearchEnabled(t *testing.T)
 		t.Fatalf("visible text = %q", got)
 	}
 }
+
+func TestStreamAccumulatorStripsInlineCitationAndReferenceMarkers(t *testing.T) {
+	acc := StreamAccumulator{SearchEnabled: true, StripReferenceMarkers: true}
+	result := acc.Apply(sse.LineResult{
+		Parsed: true,
+		Parts:  []sse.ContentPart{{Type: "text", Text: "广州天气[citation:1] 多云[reference:0]"}},
+	})
+
+	if !result.ContentSeen {
+		t.Fatalf("expected marker chunk to mark upstream content")
+	}
+	if got := acc.Text.String(); got != "广州天气 多云" {
+		t.Fatalf("visible text = %q", got)
+	}
+	if len(result.Parts) != 1 || result.Parts[0].VisibleText != "广州天气 多云" {
+		t.Fatalf("unexpected parts: %#v", result.Parts)
+	}
+}

internal/js/chat-stream/sse_parse_impl.js

@@ -621,7 +621,7 @@ function stripReferenceMarkersText(text) {
   if (!text) {
     return text;
   }
-  return text.replace(/\[reference:\s*\d+\]/gi, '');
+  return text.replace(/\[(?:citation|reference):\s*\d+\]/gi, '');
 }
 
 function asString(v) {

internal/responsehistory/session.go

@@ -0,0 +1,289 @@
+package responsehistory
+
+import (
+	"errors"
+	"net/http"
+	"strings"
+	"time"
+
+	"ds2api/internal/assistantturn"
+	"ds2api/internal/auth"
+	"ds2api/internal/chathistory"
+	"ds2api/internal/config"
+	"ds2api/internal/prompt"
+	"ds2api/internal/promptcompat"
+)
+
+type Session struct {
+	store       *chathistory.Store
+	entryID     string
+	startedAt   time.Time
+	lastPersist time.Time
+	startParams chathistory.StartParams
+	disabled    bool
+}
+
+type StartParams struct {
+	Store    *chathistory.Store
+	Request  *http.Request
+	Auth     *auth.RequestAuth
+	Surface  string
+	Standard promptcompat.StandardRequest
+}
+
+func Start(params StartParams) *Session {
+	if params.Store == nil || params.Request == nil || params.Auth == nil {
+		return nil
+	}
+	if !params.Store.Enabled() || !shouldCapture(params.Request) {
+		return nil
+	}
+	startParams := chathistory.StartParams{
+		CallerID:    strings.TrimSpace(params.Auth.CallerID),
+		AccountID:   strings.TrimSpace(params.Auth.AccountID),
+		Surface:     strings.TrimSpace(params.Surface),
+		Model:       strings.TrimSpace(params.Standard.ResponseModel),
+		Stream:      params.Standard.Stream,
+		UserInput:   ExtractSingleUserInput(params.Standard.Messages),
+		Messages:    ExtractAllMessages(params.Standard.Messages),
+		HistoryText: params.Standard.HistoryText,
+		FinalPrompt: params.Standard.FinalPrompt,
+	}
+	entry, err := params.Store.Start(startParams)
+	session := &Session{
+		store:       params.Store,
+		entryID:     entry.ID,
+		startedAt:   time.Now(),
+		lastPersist: time.Now(),
+		startParams: startParams,
+	}
+	if err != nil {
+		if entry.ID == "" {
+			config.Logger.Warn("[response_history] start failed", "surface", startParams.Surface, "error", err)
+			return nil
+		}
+		config.Logger.Warn("[response_history] start persisted in memory after write failure", "surface", startParams.Surface, "error", err)
+	}
+	return session
+}
+
+func shouldCapture(r *http.Request) bool {
+	if r == nil || r.URL == nil {
+		return false
+	}
+	if strings.TrimSpace(r.URL.Query().Get("__stream_prepare")) == "1" {
+		return false
+	}
+	if strings.TrimSpace(r.URL.Query().Get("__stream_release")) == "1" {
+		return false
+	}
+	return true
+}
+
+func ExtractSingleUserInput(messages []any) string {
+	for i := len(messages) - 1; i >= 0; i-- {
+		msg, ok := messages[i].(map[string]any)
+		if !ok {
+			continue
+		}
+		role := strings.ToLower(strings.TrimSpace(asString(msg["role"])))
+		if role != "user" {
+			continue
+		}
+		if normalized := strings.TrimSpace(prompt.NormalizeContent(msg["content"])); normalized != "" {
+			return normalized
+		}
+	}
+	return ""
+}
+
+func ExtractAllMessages(messages []any) []chathistory.Message {
+	out := make([]chathistory.Message, 0, len(messages))
+	for _, raw := range messages {
+		msg, ok := raw.(map[string]any)
+		if !ok {
+			continue
+		}
+		role := strings.ToLower(strings.TrimSpace(asString(msg["role"])))
+		content := strings.TrimSpace(prompt.NormalizeContent(msg["content"]))
+		if role == "" || content == "" {
+			continue
+		}
+		out = append(out, chathistory.Message{
+			Role:    role,
+			Content: content,
+		})
+	}
+	return out
+}
+
+func (s *Session) Progress(thinking, content string) {
+	if s == nil || s.store == nil || s.disabled {
+		return
+	}
+	now := time.Now()
+	if now.Sub(s.lastPersist) < 250*time.Millisecond {
+		return
+	}
+	s.lastPersist = now
+	s.persistUpdate(chathistory.UpdateParams{
+		Status:           "streaming",
+		ReasoningContent: thinking,
+		Content:          content,
+		StatusCode:       http.StatusOK,
+		ElapsedMs:        time.Since(s.startedAt).Milliseconds(),
+	})
+}
+
+func (s *Session) Success(statusCode int, thinking, content, finishReason string, usage map[string]any) {
+	if s == nil || s.store == nil || s.disabled {
+		return
+	}
+	s.persistUpdate(chathistory.UpdateParams{
+		Status:           "success",
+		ReasoningContent: thinking,
+		Content:          content,
+		StatusCode:       statusCode,
+		ElapsedMs:        time.Since(s.startedAt).Milliseconds(),
+		FinishReason:     finishReason,
+		Usage:            usage,
+		Completed:        true,
+	})
+}
+
+func (s *Session) Error(statusCode int, message, finishReason, thinking, content string) {
+	if s == nil || s.store == nil || s.disabled {
+		return
+	}
+	s.persistUpdate(chathistory.UpdateParams{
+		Status:           "error",
+		ReasoningContent: thinking,
+		Content:          content,
+		Error:            message,
+		StatusCode:       statusCode,
+		ElapsedMs:        time.Since(s.startedAt).Milliseconds(),
+		FinishReason:     finishReason,
+		Completed:        true,
+	})
+}
+
+func (s *Session) SuccessTurn(statusCode int, turn assistantturn.Turn, usage map[string]any) {
+	outcome := assistantturn.FinalizeTurn(turn, assistantturn.FinalizeOptions{})
+	s.Success(
+		statusCode,
+		ThinkingForArchive(turn.RawThinking, turn.DetectionThinking, turn.Thinking),
+		TextForArchive(turn.RawText, turn.Text),
+		outcome.FinishReason,
+		usage,
+	)
+}
+
+func (s *Session) ErrorTurn(statusCode int, message, finishReason string, turn assistantturn.Turn) {
+	s.Error(
+		statusCode,
+		message,
+		finishReason,
+		ThinkingForArchive(turn.RawThinking, turn.DetectionThinking, turn.Thinking),
+		TextForArchive(turn.RawText, turn.Text),
+	)
+}
+
+func TextForArchive(raw, visible string) string {
+	if strings.TrimSpace(raw) != "" {
+		return raw
+	}
+	return visible
+}
+
+func ThinkingForArchive(raw, detection, visible string) string {
+	if strings.TrimSpace(raw) != "" {
+		return raw
+	}
+	if strings.TrimSpace(detection) != "" {
+		return detection
+	}
+	return visible
+}
+
+func GenericUsage(turn assistantturn.Turn) map[string]any {
+	return map[string]any{
+		"input_tokens":     turn.Usage.InputTokens,
+		"output_tokens":    turn.Usage.OutputTokens,
+		"reasoning_tokens": turn.Usage.ReasoningTokens,
+		"total_tokens":     turn.Usage.TotalTokens,
+	}
+}
+
+func (s *Session) retryMissingEntry() bool {
+	if s == nil || s.store == nil || s.disabled {
+		return false
+	}
+	entry, err := s.store.Start(s.startParams)
+	if errors.Is(err, chathistory.ErrDisabled) {
+		s.disabled = true
+		return false
+	}
+	if entry.ID == "" {
+		if err != nil {
+			config.Logger.Warn("[response_history] recreate missing entry failed", "surface", s.startParams.Surface, "error", err)
+		}
+		return false
+	}
+	s.entryID = entry.ID
+	if err != nil {
+		config.Logger.Warn("[response_history] recreate missing entry persisted in memory after write failure", "surface", s.startParams.Surface, "error", err)
+	}
+	return true
+}
+
+func (s *Session) persistUpdate(params chathistory.UpdateParams) {
+	if s == nil || s.store == nil || s.disabled {
+		return
+	}
+	if _, err := s.store.Update(s.entryID, params); err != nil {
+		s.handlePersistError(params, err)
+	}
+}
+
+func (s *Session) handlePersistError(params chathistory.UpdateParams, err error) {
+	if err == nil || s == nil {
+		return
+	}
+	if errors.Is(err, chathistory.ErrDisabled) {
+		s.disabled = true
+		return
+	}
+	if isMissingError(err) {
+		if s.retryMissingEntry() {
+			if _, retryErr := s.store.Update(s.entryID, params); retryErr != nil {
+				if errors.Is(retryErr, chathistory.ErrDisabled) || isMissingError(retryErr) {
+					s.disabled = true
+					return
+				}
+				config.Logger.Warn("[response_history] retry after missing entry failed", "surface", s.startParams.Surface, "error", retryErr)
+			}
+			return
+		}
+		s.disabled = true
+		return
+	}
+	config.Logger.Warn("[response_history] update failed", "surface", s.startParams.Surface, "error", err)
+}
+
+func isMissingError(err error) bool {
+	if err == nil {
+		return false
+	}
+	return strings.Contains(strings.ToLower(err.Error()), "not found")
+}
+
+func asString(v any) string {
+	switch x := v.(type) {
+	case string:
+		return x
+	case nil:
+		return ""
+	default:
+		return strings.TrimSpace(prompt.NormalizeContent(x))
+	}
+}

internal/server/router.go

@@ -65,8 +65,8 @@ func NewApp() (*App, error) {
 	responsesHandler := &responses.Handler{Store: store, Auth: resolver, DS: dsClient, ChatHistory: chatHistoryStore}
 	filesHandler := &files.Handler{Store: store, Auth: resolver, DS: dsClient, ChatHistory: chatHistoryStore}
 	embeddingsHandler := &embeddings.Handler{Store: store, Auth: resolver, DS: dsClient, ChatHistory: chatHistoryStore}
-	claudeHandler := &claude.Handler{Store: store, Auth: resolver, DS: dsClient, OpenAI: chatHandler}
-	geminiHandler := &gemini.Handler{Store: store, Auth: resolver, DS: dsClient, OpenAI: chatHandler}
+	claudeHandler := &claude.Handler{Store: store, Auth: resolver, DS: dsClient, OpenAI: chatHandler, ChatHistory: chatHistoryStore}
+	geminiHandler := &gemini.Handler{Store: store, Auth: resolver, DS: dsClient, OpenAI: chatHandler, ChatHistory: chatHistoryStore}
 	adminHandler := &admin.Handler{Store: store, Pool: pool, DS: dsClient, OpenAI: chatHandler, ChatHistory: chatHistoryStore}
 	webuiHandler := webui.NewHandler()
 

internal/textclean/reference_markers.go

@@ -2,19 +2,18 @@ package textclean
 
 import "regexp"
 
-var referenceMarkerPattern = regexp.MustCompile(`(?i)\[reference:\s*\d+\]`)
+var citationReferenceMarkerPattern = regexp.MustCompile(`(?i)\[(citation|reference):\s*\d+\]`)
 
 func StripReferenceMarkers(text string) string {
 	if text == "" {
 		return text
 	}
-	return referenceMarkerPattern.ReplaceAllString(text, "")
+	return citationReferenceMarkerPattern.ReplaceAllString(text, "")
 }
 
-// StripReferenceMarkersEnabled returns true while reference-marker
-// stripping remains the fixed runtime default.  When the behaviour is
-// eventually removed this function can be deleted and callers can drop
-// the conditional.
+// StripReferenceMarkersEnabled returns the default for streaming surfaces,
+// where partial citation/reference markers are hidden before the final
+// link metadata is available.
 func StripReferenceMarkersEnabled() bool {
 	return true
 }

tests/node/chat-history-utils.test.js

@@ -58,3 +58,47 @@ test('chat history strict parser inserts history after system messages', async (
     { role: 'user', content: 'latest' },
   ]);
 });
+
+test('chat history transcript parser replaces current input file placeholder', async () => {
+  const {
+    buildListModeMessages,
+  } = await loadUtils();
+  const t = (key) => key;
+  const item = {
+    messages: [{
+      role: 'user',
+      content: 'Continue from the latest state in the attached DS2API_HISTORY.txt context. Treat it as the current working state and answer the latest user request directly.',
+    }],
+    history_text: [
+      '# DS2API_HISTORY.txt',
+      'Prior conversation history and tool progress.',
+      '',
+      '=== 1. SYSTEM ===',
+      'policy',
+      '',
+      '=== 2. USER ===',
+      'hello',
+      '',
+      '=== 3. ASSISTANT ===',
+      'hi',
+      '',
+      '=== 4. TOOL ===',
+      '[name=search_web tool_call_id=call_1]',
+      '{"ok":true}',
+      '',
+      '=== 5. USER ===',
+      'latest',
+      '',
+    ].join('\n'),
+  };
+
+  const result = buildListModeMessages(item, t);
+  assert.equal(result.historyMerged, true);
+  assert.deepEqual(result.messages, [
+    { role: 'system', content: 'policy' },
+    { role: 'user', content: 'hello' },
+    { role: 'assistant', content: 'hi' },
+    { role: 'tool', content: '[name=search_web tool_call_id=call_1]\n{"ok":true}' },
+    { role: 'user', content: 'latest' },
+  ]);
+});

tests/node/chat-stream.test.js

@@ -615,17 +615,17 @@ test('parseChunkForContent preserves space-only content tokens', () => {
   assert.deepEqual(parsed.parts, [{ text: ' ', type: 'text' }]);
 });
 
-test('parseChunkForContent strips reference markers from fragment content', () => {
+test('parseChunkForContent strips citation and reference markers from fragment content', () => {
   const chunk = {
     p: 'response/fragments',
     o: 'APPEND',
     v: [
-      { type: 'RESPONSE', content: '广州天气 [reference:12] 多云' },
+      { type: 'RESPONSE', content: '广州天气 [citation:1] [reference:12] 多云' },
     ],
   };
   const parsed = parseChunkForContent(chunk, false, 'text');
   assert.equal(parsed.finished, false);
-  assert.deepEqual(parsed.parts, [{ text: '广州天气  多云', type: 'text' }]);
+  assert.deepEqual(parsed.parts, [{ text: '广州天气   多云', type: 'text' }]);
 });
 
 test('parseChunkForContent detects content_filter status and ignores upstream output tokens', () => {

webui/src/features/apiTester/useChatStreamClient.js

@@ -123,7 +123,6 @@ export function useChatStreamClient({
             const headers = {
                 'Content-Type': 'application/json',
                 'Authorization': `Bearer ${effectiveKey}`,
-                'X-Ds2-Source': 'admin-webui-api-tester',
             }
             if (requestAccount) {
                 headers['X-Ds2-Target-Account'] = requestAccount

webui/src/features/chatHistory/ChatHistoryContainer.jsx

@@ -10,6 +10,9 @@ import {
     VIEW_MODE_KEY,
 } from './chatHistoryUtils'
 
+const LIST_REFRESH_MS = 1500
+const STREAMING_DETAIL_REFRESH_MS = 750
+
 export default function ChatHistoryContainer({ authFetch, onMessage }) {
     const { t, lang } = useI18n()
     const apiFetch = authFetch || fetch
@@ -136,7 +139,7 @@ export default function ChatHistoryContainer({ authFetch, onMessage }) {
         if (!autoRefreshReady || limit === DISABLED_LIMIT) return undefined
         const timer = window.setInterval(() => {
             loadList({ mode: 'silent', announceError: false })
-        }, 5000)
+        }, LIST_REFRESH_MS)
         return () => window.clearInterval(timer)
     }, [autoRefreshReady, limit])
 
@@ -144,7 +147,7 @@ export default function ChatHistoryContainer({ authFetch, onMessage }) {
         if (!autoRefreshReady || !selectedId || selectedSummary?.status !== 'streaming') return undefined
         const timer = window.setInterval(() => {
             loadDetail(selectedId, { announceError: false })
-        }, 1000)
+        }, STREAMING_DETAIL_REFRESH_MS)
         return () => window.clearInterval(timer)
     }, [autoRefreshReady, selectedId, selectedSummary?.status])
 

webui/src/features/chatHistory/ChatHistoryDetail.jsx

@@ -207,6 +207,10 @@ function MetaGrid({ selectedItem, t }) {
                         {formatElapsed(selectedItem.elapsed_ms, t)}
                     </div>
                 </div>
+                <div className="rounded-lg border border-border bg-card px-3 py-2">
+                    <div className="text-[11px] text-muted-foreground">{t('chatHistory.metaSurface')}</div>
+                    <div className="text-sm font-medium text-foreground break-all">{selectedItem.surface || t('chatHistory.metaUnknown')}</div>
+                </div>
                 <div className="rounded-lg border border-border bg-card px-3 py-2">
                     <div className="text-[11px] text-muted-foreground">{t('chatHistory.metaModel')}</div>
                     <div className="text-sm font-medium text-foreground break-all">{selectedItem.model || t('chatHistory.metaUnknown')}</div>

webui/src/features/chatHistory/ChatHistoryPanels.jsx

@@ -69,7 +69,7 @@ export function ChatHistoryListPane({ items, selectedItem, deletingId, t, lang,
                                     {item.user_input || t('chatHistory.untitled')}
                                 </div>
                                 <div className="text-[11px] text-muted-foreground mt-1 truncate">
-                                    {item.model || '-'}
+                                    {[item.surface, item.model].filter(Boolean).join(' · ') || '-'}
                                 </div>
                             </div>
                             <div className="flex items-center gap-2 shrink-0">

webui/src/features/chatHistory/chatHistoryUtils.js

@@ -15,12 +15,24 @@ const CURRENT_INPUT_FILE_PROMPT = 'Continue from the latest state in the attache
 const LEGACY_CURRENT_INPUT_FILE_PROMPTS = new Set([
     'The current request and prior conversation context have already been provided. Answer the latest user request directly.',
 ])
+const HISTORY_TRANSCRIPT_TITLE = '# DS2API_HISTORY.txt'
+const HISTORY_TRANSCRIPT_ENTRY_RE = /^===\s+\d+\.\s+([A-Z][A-Z_ -]*)\s+===\s*$/gm
 
 function isCurrentInputFilePrompt(value) {
     const text = String(value || '').trim()
     return text === CURRENT_INPUT_FILE_PROMPT || LEGACY_CURRENT_INPUT_FILE_PROMPTS.has(text)
 }
 
+function normalizeHistoryRole(role) {
+    const normalized = String(role || '').trim().toLowerCase()
+    if (normalized === 'function') return 'tool'
+    if (normalized === 'developer') return 'system'
+    if (normalized === 'system' || normalized === 'user' || normalized === 'assistant' || normalized === 'tool') {
+        return normalized
+    }
+    return normalized || 'system'
+}
+
 export function formatDateTime(value, lang) {
     if (!value) return '-'
     try {
@@ -221,11 +233,37 @@ export function parseStrictHistoryMessages(historyText) {
     return parsed
 }
 
+export function parseTranscriptHistoryMessages(historyText) {
+    const rawText = String(historyText || '')
+    const titleIndex = rawText.indexOf(HISTORY_TRANSCRIPT_TITLE)
+    const transcript = titleIndex >= 0 ? rawText.slice(titleIndex) : rawText
+    const matches = [...transcript.matchAll(HISTORY_TRANSCRIPT_ENTRY_RE)]
+    if (!matches.length) return null
+
+    const parsed = []
+    for (let i = 0; i < matches.length; i += 1) {
+        const match = matches[i]
+        const next = matches[i + 1]
+        const role = normalizeHistoryRole(match[1])
+        const start = (match.index || 0) + match[0].length
+        const end = next ? next.index : transcript.length
+        const content = transcript.slice(start, end).replace(/^\r?\n/, '').trim()
+        if (!content) continue
+        parsed.push({ role, content })
+    }
+
+    return parsed.length ? parsed : null
+}
+
+export function parseHistoryMessages(historyText) {
+    return parseStrictHistoryMessages(historyText) || parseTranscriptHistoryMessages(historyText)
+}
+
 export function buildListModeMessages(item, t) {
     const liveMessages = Array.isArray(item?.messages) && item.messages.length > 0
         ? item.messages
         : [{ role: 'user', content: item?.user_input || t('chatHistory.emptyUserInput') }]
-    const historyMessages = parseStrictHistoryMessages(item?.history_text)
+    const historyMessages = parseHistoryMessages(item?.history_text)
 
     if (!historyMessages?.length) {
         return { messages: liveMessages, historyMerged: false }

webui/src/locales/en.json

@@ -18,8 +18,8 @@
             "desc": "Test API connectivity and responses"
         },
         "history": {
-            "label": "Conversations",
-            "desc": "Browse server-side external chat history"
+            "label": "Responses",
+            "desc": "Browse server-side upstream response records"
         },
         "import": {
             "label": "Batch Import",
@@ -261,7 +261,7 @@
         "loading": "Loading conversation history...",
         "loadFailed": "Failed to load conversation history.",
         "retentionTitle": "Retention",
-        "retentionDesc": "The server keeps only the latest N external /v1/chat/completions conversations.",
+        "retentionDesc": "The server keeps only the latest N DeepSeek upstream response records across OpenAI Chat, OpenAI Responses, Claude, and Gemini direct interfaces.",
         "off": "OFF",
         "refresh": "Refresh",
         "clearAll": "Clear all",
@@ -277,7 +277,7 @@
         "viewModeList": "List mode",
         "viewModeMerged": "Merged mode",
         "emptyTitle": "No conversation history yet",
-        "emptyDesc": "When external clients call /v1/chat/completions, the server will save the results here automatically.",
+        "emptyDesc": "When a supported interface talks to DeepSeek upstream and receives a response, the server saves the result here automatically.",
         "untitled": "Untitled conversation",
         "noPreview": "No preview available.",
         "selectPrompt": "Select a record on the left to view details.",
@@ -303,6 +303,7 @@
         "metaTitle": "Metadata",
         "metaAccount": "Account",
         "metaElapsed": "Elapsed",
+        "metaSurface": "Surface",
         "metaModel": "Model",
         "metaStatusCode": "Status code",
         "metaStream": "Output mode",

webui/src/locales/zh.json

@@ -18,8 +18,8 @@
             "desc": "测试 API 连接与响应"
         },
         "history": {
-            "label": "对话记录",
-            "desc": "查看服务器保存的外部对话历史"
+            "label": "响应记录",
+            "desc": "查看服务器保存的上游响应归档"
         },
         "import": {
             "label": "批量导入",
@@ -261,7 +261,7 @@
         "loading": "正在加载对话记录...",
         "loadFailed": "加载对话记录失败",
         "retentionTitle": "保留条数",
-        "retentionDesc": "服务器端只保留最新 N 条外部 /v1/chat/completions 对话记录。",
+        "retentionDesc": "服务器端只保留最新 N 条 DeepSeek 上游响应记录，覆盖 OpenAI Chat、OpenAI Responses、Claude 和 Gemini 直连接口。",
         "off": "OFF",
         "refresh": "刷新",
         "clearAll": "清空全部",
@@ -277,7 +277,7 @@
         "viewModeList": "列表模式",
         "viewModeMerged": "合并模式",
         "emptyTitle": "还没有可用的对话记录",
-        "emptyDesc": "当外部客户端调用 /v1/chat/completions 时，服务端会自动把结果写入这里。",
+        "emptyDesc": "当支持的接口与 DeepSeek 上游交互并收到响应时，服务端会自动把结果写入这里。",
         "untitled": "未命名对话",
         "noPreview": "暂无预览内容",
         "selectPrompt": "从左侧选择一条记录查看详情。",
@@ -303,6 +303,7 @@
         "metaTitle": "元信息",
         "metaAccount": "使用账号",
         "metaElapsed": "耗时",
+        "metaSurface": "接口",
         "metaModel": "模型",
         "metaStatusCode": "状态码",
         "metaStream": "输出模式",