nolo-cli 0.1.21 → 0.1.23

package/ai/agent/streamAgentChatTurn.ts

@@ -4,7 +4,7 @@ import { createDialogMessageKeyAndId } from "database/keys";
 import { DataType } from "create/types";
 
 import type { RootState } from "app/store";
-import { patch, read } from "database/dbSlice";
+import { patch, read, selectById } from "database/dbSlice";
 import { generateRequestBody } from "ai/llm/generateRequestBody";
 import {
     selectCurrentDialogConfig,
@@ -146,8 +146,23 @@ export interface StreamAgentChatTurnArgs {
     isStreaming?: boolean;
     parentMessageId?: string;
     runtimeOptions?: AgentRuntimeOptions;
+    quickChatPerfStartedAt?: number;
 }
 
+const logQuickChatPerfStage = (
+    startedAt: number | undefined,
+    stage: string,
+    details: Record<string, unknown> = {}
+) => {
+    if (!startedAt) return;
+    console.info("[QuickChatPerf]", {
+        stage,
+        elapsedMs: Date.now() - startedAt,
+        ...(typeof performance !== "undefined" ? { atMs: performance.now() } : {}),
+        ...details,
+    });
+};
+
 const normalizeAgentRunUserInput = (userInput: string | any[]) => {
     if (typeof userInput === "string") {
         return userInput;
@@ -166,6 +181,16 @@ const normalizeAgentRunUserInput = (userInput: string | any[]) => {
     });
 };
 
+const isUsableAgentConfig = (value: unknown): value is Agent =>
+    !!value &&
+    typeof value === "object" &&
+    typeof (value as Agent).dbKey === "string" &&
+    !!(value as Agent).dbKey &&
+    typeof (value as Agent).model === "string" &&
+    !!(value as Agent).model &&
+    typeof (value as Agent).provider === "string" &&
+    !!(value as Agent).provider;
+
 const extractAgentRunUserText = (userInput: string | any[]) => {
     if (typeof userInput === "string") {
         return userInput;
@@ -180,17 +205,6 @@ const extractAgentRunUserText = (userInput: string | any[]) => {
         .trim();
 };
 
-const requiresServerResolvedProviderConfig = (agentConfig: Agent): boolean => {
-    if (agentConfig.useServerProxy !== true) return false;
-
-    const provider = String(agentConfig.provider ?? "").toLowerCase();
-    const apiSource = String((agentConfig as any).apiSource ?? "").toLowerCase();
-    const isCustomProvider = provider === "custom" || apiSource === "custom";
-    if (!isCustomProvider) return false;
-
-    return !agentConfig.customProviderUrl?.trim() || !agentConfig.apiKey?.trim();
-};
-
 const hasAgentRunUserInputContent = (userInput: string | any[]) => {
     if (typeof userInput === "string") {
         return userInput.trim().length > 0;
@@ -209,7 +223,14 @@ export const streamAgentChatTurnHandler = async (
     args: StreamAgentChatTurnArgs,
     thunkApi: any,
 ) => {
-    const { agentKey, userInput, dialogKey: explicitDialogKey, parentMessageId, runtimeOptions } = args;
+    const {
+        agentKey,
+        userInput,
+        dialogKey: explicitDialogKey,
+        parentMessageId,
+        runtimeOptions,
+        quickChatPerfStartedAt,
+    } = args;
     const { getState, dispatch, rejectWithValue } = thunkApi;
     const state = getState() as RootState;
 
@@ -227,11 +248,25 @@ export const streamAgentChatTurnHandler = async (
     try {
         let totalTurnUsage: any = null;
         const agentRunUserInput = normalizeAgentRunUserInput(userInput);
-        // 1. 读取 Agent 配置
-        const agentConfig = (await dispatch(read({ dbKey: agentKey })).unwrap()) as Agent;
+        logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-entered", {
+            agentKey,
+            dialogKey: explicitDialogKey ?? null,
+        });
+        // 1. 读取 Agent 配置。Quick Chat 会提前预热默认 agent；命中 Redux DB 缓存时避免重复读。
+        const cachedAgentConfig = selectById(getState() as RootState, agentKey);
+        const agentConfig = isUsableAgentConfig(cachedAgentConfig)
+            ? cachedAgentConfig
+            : ((await dispatch(read({ dbKey: agentKey })).unwrap()) as Agent);
         if (!agentConfig) {
             return rejectWithValue(`Agent config not found for ID: ${agentKey}`);
         }
+        logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-config-read", {
+            agentKey,
+            model: agentConfig.model,
+            provider: agentConfig.provider,
+            apiSource: agentConfig.apiSource,
+            source: cachedAgentConfig === agentConfig ? "cache" : "read",
+        });
 
         // ── CLI Agent 专用路由 ────────────────────────────────────────────────
         // CLI 共享 prompt / model 这些入口能力，但不复用本地 tool-call 循环。
@@ -330,6 +365,8 @@ export const streamAgentChatTurnHandler = async (
                         userInput: agentRunUserInput,
                         messages: cleanedMessages,
                         stream: true,
+                        persistDialog: false,
+                        clientDialogId: dialogId,
                         runtimeContext: {
                             surface: "web",
                             host: "browser",
@@ -616,6 +653,10 @@ export const streamAgentChatTurnHandler = async (
         }
         runtimeDialogKey = dialogKey;
         const dialogId = extractCustomId(dialogKey);
+        logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-dialog-resolved", {
+            dialogKey,
+            dialogId,
+        });
 
         const userInputText = extractAgentRunUserText(userInput);
 
@@ -625,12 +666,10 @@ export const streamAgentChatTurnHandler = async (
                 : null;
         const currentServer = selectCurrentServer(state);
         const declaredRuntimeServerBase = extractAgentRuntimeServerBase(agentConfig);
-        const needsServerResolvedProviderConfig =
-            requiresServerResolvedProviderConfig(agentConfig);
         const requestedServerBase =
             explicitServerBase ??
             declaredRuntimeServerBase ??
-            (needsServerResolvedProviderConfig ? currentServer : null);
+            null;
         const normalizedRequestedServerBase =
             requestedServerBase && normalizeServerOrigin(requestedServerBase);
         const normalizedCurrentServer = normalizeServerOrigin(
@@ -643,12 +682,13 @@ export const streamAgentChatTurnHandler = async (
             !runtimeOptions?.imageConfigOverride;
         if (requestedServerBase && canAutoRouteRemotely) {
             if (
-                !needsServerResolvedProviderConfig &&
                 normalizedRequestedServerBase &&
                 normalizedCurrentServer &&
                 normalizedRequestedServerBase === normalizedCurrentServer
             ) {
-                // same server as current workspace; keep local flow
+                // Same server as the current workspace; keep the UI-managed
+                // chat/tool loop and let /api/chat hydrate redacted provider
+                // credentials server-side when needed.
             } else {
                 const token = selectCurrentToken(state);
                 const authHeader = token ? `Bearer ${token}` : "";
@@ -711,6 +751,8 @@ export const streamAgentChatTurnHandler = async (
                         userInput: agentRunUserInput,
                         messages: cleanedMessages,
                         stream: true,
+                        persistDialog: false,
+                        clientDialogId: dialogId,
                         runtimeContext: {
                             surface: "web",
                             host: "browser",
@@ -811,6 +853,10 @@ export const streamAgentChatTurnHandler = async (
             recommendedSkillHints: referenceRecommendedSkillHints,
             skillPromptPatches: referenceSkillPromptPatches,
         } = await resolveReferenceAssets(agentConfig.references, dispatch);
+        logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-references-resolved", {
+            referenceCount: normalizedReferences?.length ?? 0,
+            referencedToolCount: referenceTools?.length ?? 0,
+        });
 
         const agentConfigWithReferences = {
             ...agentConfig,
@@ -906,6 +952,10 @@ export const streamAgentChatTurnHandler = async (
                 currentDialog ?? undefined,
                 mergedContentCache,
             );
+            logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-static-context-ready", {
+                model: agentConfigForCall.model,
+                responseApi: true,
+            });
 
             let appendTempUserInput = true;
             let currentParentMessageId = parentMessageId ?? undefined;
@@ -947,6 +997,9 @@ export const streamAgentChatTurnHandler = async (
                     mergedContentCache,
                     dialogKey,
                 );
+                logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-dynamic-context-ready", {
+                    responseApi: true,
+                });
                 const contexts = mergeContexts(staticContexts, dynamicContexts);
 
                 const rawMessages = filterMessagesForParallelBranch(
@@ -1014,6 +1067,11 @@ export const streamAgentChatTurnHandler = async (
                     userInput: userInputText,
                     contexts,
                 });
+                logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-model-request-starting", {
+                    responseApi: true,
+                    dynamicMessageCount: dynamicMessages.length,
+                    stableMessageCount: stableMessages.length,
+                });
 
                 const meta: CompletionMeta = await sendOpenAIResponseRequest({
                     bodyData,
@@ -1025,6 +1083,13 @@ export const streamAgentChatTurnHandler = async (
                         agentConfigForCall,
                         runtimeOptions,
                     ),
+                    quickChatPerfStartedAt,
+                });
+                logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-model-request-finished", {
+                    responseApi: true,
+                    hasToolCalls: meta.hasToolCalls,
+                    hasHandedOff: meta.hasHandedOff,
+                    hasPendingInteraction: meta.hasPendingInteraction,
                 });
 
                 appendTempUserInput = false;
@@ -1093,6 +1158,10 @@ export const streamAgentChatTurnHandler = async (
             currentDialog ?? undefined,
             mergedContentCache,
         );
+        logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-static-context-ready", {
+            model: agentConfigForCall.model,
+            responseApi: false,
+        });
 
         let appendTempUserInput = true;
         let currentParentMessageId = parentMessageId ?? undefined;
@@ -1139,6 +1208,9 @@ export const streamAgentChatTurnHandler = async (
                 mergedContentCache,
                 dialogKey,
             );
+            logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-dynamic-context-ready", {
+                responseApi: false,
+            });
 
             // 合并静态和动态上下文
             const contexts = mergeContexts(staticContexts, dynamicContexts);
@@ -1209,6 +1281,11 @@ export const streamAgentChatTurnHandler = async (
                 userInput: userInputText,
                 contexts,
             });
+            logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-model-request-starting", {
+                responseApi: false,
+                dynamicMessageCount: dynamicMessages.length,
+                stableMessageCount: stableMessages.length,
+            });
 
             const meta: CompletionMeta = await sendOpenAICompletionsRequest({
                 bodyData,
@@ -1220,6 +1297,13 @@ export const streamAgentChatTurnHandler = async (
                     agentConfigForCall,
                     runtimeOptions,
                 ),
+                quickChatPerfStartedAt,
+            });
+            logQuickChatPerfStage(quickChatPerfStartedAt, "stream-agent-model-request-finished", {
+                responseApi: false,
+                hasToolCalls: meta.hasToolCalls,
+                hasHandedOff: meta.hasHandedOff,
+                hasPendingInteraction: meta.hasPendingInteraction,
             });
 
             appendTempUserInput = false;

package/ai/chat/fetchUtils.native.ts

@@ -42,7 +42,9 @@ const buildProxyPayload = (
         ...bodyData,
         url: api,
         provider,
+        agentKey: agentConfig.dbKey,
         ...(apiSource ? { apiSource } : {}),
+        ...((agentConfig as any).apiKeyHeader ? { apiKeyHeader: (agentConfig as any).apiKeyHeader } : {}),
         KEY: apiKey,
     };
 };

package/ai/chat/fetchUtils.ts

@@ -40,7 +40,9 @@ const buildProxyPayload = (
     ...bodyData,
     url: api,
     provider,
+    agentKey: agentConfig.dbKey,
     ...(apiSource ? { apiSource } : {}),
+    ...((agentConfig as any).apiKeyHeader ? { apiKeyHeader: (agentConfig as any).apiKeyHeader } : {}),
     KEY: apiKey,
   };
 };

package/ai/chat/sendOpenAICompletionsRequest.ts

@@ -160,6 +160,20 @@ export type CompletionMeta = {
   usage?: any;
 };
 
+const logQuickChatPerfStage = (
+  startedAt: number | undefined,
+  stage: string,
+  details?: Record<string, unknown>
+) => {
+  if (!startedAt) return;
+  console.info("[QuickChatPerf]", {
+    stage,
+    elapsedMs: Date.now() - startedAt,
+    ...(typeof performance !== "undefined" ? { atMs: performance.now() } : {}),
+    ...(details ?? {}),
+  });
+};
+
 /**
  * 初始化流式状态
  */
@@ -455,6 +469,7 @@ export const sendOpenAICompletionsRequest = async ({
   parentMessageId,
   messageMetadata,
   disableToolsForThisRequest = false,
+  quickChatPerfStartedAt,
 }: {
   bodyData: any;
   agentConfig: any;
@@ -463,6 +478,7 @@ export const sendOpenAICompletionsRequest = async ({
   parentMessageId?: string;
   messageMetadata?: Partial<Message>;
   disableToolsForThisRequest?: boolean;
+  quickChatPerfStartedAt?: number;
 }): Promise<CompletionMeta> => {
   const { dispatch, getState, signal: thunkSignal } = thunkApi;
 
@@ -549,6 +565,10 @@ export const sendOpenAICompletionsRequest = async ({
 
     const api = getApiEndpoint(agentConfig);
     const token = selectCurrentToken(getState() as RootState);
+    logQuickChatPerfStage(quickChatPerfStartedAt, "openai-completions-fetch-starting", {
+      api,
+      dialogKey,
+    });
     const response = await performFetchRequest({
       agentConfig,
       api,
@@ -557,6 +577,11 @@ export const sendOpenAICompletionsRequest = async ({
       signal,
       token,
     });
+    logQuickChatPerfStage(quickChatPerfStartedAt, "openai-completions-fetch-response", {
+      ok: response.ok,
+      status: response.status,
+      dialogKey,
+    });
 
     if (!response.ok) {
       const errorMessage = await parseApiError(response);
@@ -578,6 +603,9 @@ export const sendOpenAICompletionsRequest = async ({
     }
 
     const decoder = new TextDecoder();
+    let loggedFirstStreamChunk = false;
+    let loggedFirstParsedEvent = false;
+    let loggedFirstVisibleDelta = false;
 
     while (true) {
       const { done, value } = await reader.read();
@@ -602,8 +630,25 @@ export const sendOpenAICompletionsRequest = async ({
         break;
       }
 
+      if (!loggedFirstStreamChunk) {
+        loggedFirstStreamChunk = true;
+        logQuickChatPerfStage(
+          quickChatPerfStartedAt,
+          "openai-completions-first-stream-chunk",
+          { dialogKey, byteLength: value.byteLength }
+        );
+      }
+
       const chunk = decoder.decode(value, { stream: true });
       const parsedResults = parseSSE(chunk);
+      if (parsedResults.length > 0 && !loggedFirstParsedEvent) {
+        loggedFirstParsedEvent = true;
+        logQuickChatPerfStage(
+          quickChatPerfStartedAt,
+          "openai-completions-first-sse-event",
+          { dialogKey, eventCount: parsedResults.length }
+        );
+      }
 
       for (const parsedData of parsedResults) {
         const dataList = Array.isArray(parsedData) ? parsedData : [parsedData];
@@ -641,6 +686,14 @@ export const sendOpenAICompletionsRequest = async ({
             delta
           );
           streamState = updatedState;
+          if (hasNewVisibleContent && !loggedFirstVisibleDelta) {
+            loggedFirstVisibleDelta = true;
+            logQuickChatPerfStage(
+              quickChatPerfStartedAt,
+              "openai-completions-first-visible-delta",
+              { dialogKey }
+            );
+          }
 
           emitStreamingUpdate(hasNewVisibleContent, streamState, {
             dispatch,
@@ -712,6 +765,9 @@ export const sendOpenAICompletionsRequest = async ({
     };
     streamState = await finalizeStream(streamState, finalizeCtx);
   } finally {
+    logQuickChatPerfStage(quickChatPerfStartedAt, "openai-completions-stream-finished", {
+      dialogKey,
+    });
     dispatch(removeActiveController({ messageId, dialogKey }));
     try {
       await reader?.cancel();

package/ai/chat/sendOpenAIResponseRequest.ts

@@ -42,6 +42,20 @@ const shouldEnableBuiltInImageGeneration = (agentConfig: any): boolean =>
   !getModelInfo(String(agentConfig?.model || ""))?.hasImageOutput &&
   !!agentConfig?.imageConfig?.enabled;
 
+const logQuickChatPerfStage = (
+  startedAt: number | undefined,
+  stage: string,
+  details?: Record<string, unknown>
+) => {
+  if (!startedAt) return;
+  console.info("[QuickChatPerf]", {
+    stage,
+    elapsedMs: Date.now() - startedAt,
+    ...(typeof performance !== "undefined" ? { atMs: performance.now() } : {}),
+    ...(details ?? {}),
+  });
+};
+
 type StreamState = {
   content: string;
   contentBuffer: Array<
@@ -140,6 +154,7 @@ export const sendOpenAIResponseRequest = async ({
   dialogKey,
   parentMessageId,
   messageMetadata,
+  quickChatPerfStartedAt,
 }: {
   bodyData: any;
   agentConfig: any;
@@ -147,6 +162,7 @@ export const sendOpenAIResponseRequest = async ({
   dialogKey: string;
   parentMessageId?: string;
   messageMetadata?: Partial<Message>;
+  quickChatPerfStartedAt?: number;
 }): Promise<CompletionMeta> => {
   const { dispatch, getState, signal: thunkSignal } = thunkApi;
   const dialogId = extractCustomId(dialogKey);
@@ -354,6 +370,10 @@ export const sendOpenAIResponseRequest = async ({
 
     const api = getApiEndpoint(agentConfig);
     const token = selectCurrentToken(getState() as RootState);
+    logQuickChatPerfStage(quickChatPerfStartedAt, "openai-response-fetch-starting", {
+      api,
+      dialogKey,
+    });
     const response = await performFetchRequest({
       agentConfig,
       api,
@@ -362,6 +382,11 @@ export const sendOpenAIResponseRequest = async ({
       signal,
       token,
     });
+    logQuickChatPerfStage(quickChatPerfStartedAt, "openai-response-fetch-response", {
+      ok: response.ok,
+      status: response.status,
+      dialogKey,
+    });
 
     if (!response.ok) {
       const errorMessage = await parseApiError(response);
@@ -379,13 +404,33 @@ export const sendOpenAIResponseRequest = async ({
     const parseSSE = createSSEParser();
     const decoder = new TextDecoder();
     let finishReason: string | null = null;
+    let loggedFirstStreamChunk = false;
+    let loggedFirstParsedEvent = false;
+    let loggedFirstVisibleDelta = false;
 
     while (true) {
       const { done, value } = await reader.read();
       if (done) break;
 
+      if (!loggedFirstStreamChunk) {
+        loggedFirstStreamChunk = true;
+        logQuickChatPerfStage(
+          quickChatPerfStartedAt,
+          "openai-response-first-stream-chunk",
+          { dialogKey, byteLength: value.byteLength }
+        );
+      }
+
       const chunk = decoder.decode(value, { stream: true });
       const events = parseSSE(chunk);
+      if (events.length > 0 && !loggedFirstParsedEvent) {
+        loggedFirstParsedEvent = true;
+        logQuickChatPerfStage(
+          quickChatPerfStartedAt,
+          "openai-response-first-sse-event",
+          { dialogKey, eventCount: events.length }
+        );
+      }
       const eventList = Array.isArray(events) ? events : [events];
 
       for (const event of eventList) {
@@ -404,6 +449,14 @@ export const sendOpenAIResponseRequest = async ({
             if (event.delta) {
               state.content += event.delta;
               state.contentBuffer = seg(state.content);
+              if (!loggedFirstVisibleDelta) {
+                loggedFirstVisibleDelta = true;
+                logQuickChatPerfStage(
+                  quickChatPerfStartedAt,
+                  "openai-response-first-visible-delta",
+                  { dialogKey }
+                );
+              }
               flush();
             }
             break;
@@ -427,6 +480,14 @@ export const sendOpenAIResponseRequest = async ({
               if (itemText) {
                 state.content = itemText;
                 state.contentBuffer = seg(state.content);
+                if (!loggedFirstVisibleDelta) {
+                  loggedFirstVisibleDelta = true;
+                  logQuickChatPerfStage(
+                    quickChatPerfStartedAt,
+                    "openai-response-first-visible-delta",
+                    { dialogKey }
+                  );
+                }
                 flush();
               }
             }
@@ -506,6 +567,9 @@ export const sendOpenAIResponseRequest = async ({
     await finalize();
     return buildMeta(false, false, "error");
   } finally {
+    logQuickChatPerfStage(quickChatPerfStartedAt, "openai-response-stream-finished", {
+      dialogKey,
+    });
     dispatch(removeActiveController({ messageId, dialogKey }));
     await safeCancel(reader);
   }

package/ai/llm/kimi.ts

@@ -2,7 +2,7 @@ export const FIREWORKS_KIMI_LATEST_MODEL = "accounts/fireworks/models/kimi-lates
 export const FIREWORKS_KIMI_CURRENT_MODEL = "accounts/fireworks/models/kimi-k2p6";
 export const DEEPINFRA_KIMI_FALLBACK_MODEL = "moonshotai/Kimi-K2.6";
 export const OPENROUTER_KIMI_FALLBACK_MODEL = "moonshotai/kimi-k2.6";
-export const KIMI_PLATFORM_FALLBACK_STATUSES = [402, 429, 500, 502, 503, 504];
+export const KIMI_PLATFORM_FALLBACK_STATUSES = [401, 402, 429, 500, 502, 503, 504];
 
 export const isFireworksKimiModel = (model?: string | null): boolean =>
   model === FIREWORKS_KIMI_LATEST_MODEL || model === FIREWORKS_KIMI_CURRENT_MODEL;

package/ai/llm/providers.ts

@@ -276,6 +276,9 @@ export function getApiEndpoint(agent: Agent): string {
     effectiveProvider.toLowerCase() === "custom" ||
     (agent as any).apiSource === "custom"
   ) {
+    if (agent.useServerProxy) {
+      return "";
+    }
     throw new Error(
       "Custom provider URL is required when apiSource is 'custom'."
     );

package/ai/llm/reasoningModels.ts

@@ -4,6 +4,7 @@ const REASONING_MODEL_NAMES = new Set([
   "deepseek-reasoner",
   "gemini-2.5-pro",
   "gemini-2.5-flash",
+  "gemini-3.5-flash",
   "gemini-3-flash-preview",
   "gemini-3-pro-preview",
   "gemini-3.1-pro-preview",