nolo-cli 0.97.6 → 0.97.8

package/{chunk-QNTL2KFU.js → chunk-XEALKKFF.js}

@@ -45,7 +45,7 @@ import {
 import {
   inlineImageUrlsForCustomProvider,
   prepareTools
-} from "./chunk-LYKHXIQC.js";
+} from "./chunk-X5YVT5HS.js";
 import {
   getServerProviderSecret
 } from "./chunk-HS7MLR74.js";
@@ -63,7 +63,7 @@ import {
   rememberMemoryFunctionSchema,
   resolveAgentRequiredPackIds,
   setTodoListFunc
-} from "./chunk-ZGNDLCYH.js";
+} from "./chunk-TQE6BVFT.js";
 import {
   deriveCredentialGroup,
   deriveCustomProviderCredentialGroup
@@ -107,7 +107,7 @@ import {
 } from "./chunk-E3YWHWCR.js";
 import {
   getDefaultCliLocalRuntimeDb
-} from "./chunk-74D5HEDI.js";
+} from "./chunk-FOXQ7BAB.js";
 import {
   applyBuiltinAgentRuntimeOverride,
   applyTokenUsageToDayStats,
@@ -115,14 +115,14 @@ import {
   prepareTokenUsageData,
   resolveBuiltinPlatformAgentConfig,
   runKeyed
-} from "./chunk-BS7YHAQP.js";
+} from "./chunk-UWVGKNAL.js";
 import {
   BUILTIN_DIALOG_LLM_CHAT_FALLBACK_MODEL,
   BUILTIN_TITLE_LLM_CONFIG
 } from "./chunk-523DITKR.js";
 import {
   fetchServerSyncedCredential
-} from "./chunk-2LM7XUCX.js";
+} from "./chunk-SPCBJGFO.js";
 import {
   FORCED_TOOLS,
   addDefaultLightWebToolsForConfiguredAgents,
@@ -149,6 +149,7 @@ import {
   createDevinProvider,
   createHybridRecordStore,
   createLocalWorkspaceToolExecutors,
+  deriveStableUuid,
   executeLocalToolWithPolicy,
   executeOpenAiCompatibleChatCompletion,
   filterRetainedLedgerRecords,
@@ -156,11 +157,13 @@ import {
   internalListWorkspaceEntries,
   internalSearchWorkspace,
   isLineRangeCovered,
+  isStableSessionIdentityEnabled,
   localDialogMessageRecordToRuntimeMessage,
   parseSyncServersEnv,
   readDevinUpstreamFailure,
-  resolveOpenAiCompatibleProviderConfig
-} from "./chunk-ANRWQGOV.js";
+  resolveOpenAiCompatibleProviderConfig,
+  resolveSessionKey
+} from "./chunk-GDFA73QZ.js";
 import {
   accumulateGeminiChunks,
   buildGeminiGenerateContentRequest,
@@ -182,14 +185,14 @@ import {
   readPlatformChatSseCompletion,
   resolvePlatformChatProviderConfig,
   shouldUsePlatformChatProvider
-} from "./chunk-CV366NXM.js";
+} from "./chunk-V5OWQOJZ.js";
 import {
   readSseDataValues
 } from "./chunk-GMZLP5DV.js";
 import {
   resolveAgentRuntimeConfigFromRecord,
   resolveHistoricalToolContentCap
-} from "./chunk-4XYBLM4D.js";
+} from "./chunk-LQ4PDQCI.js";
 import {
   isDevinOAuthAgent
 } from "./chunk-5ZUL4CUO.js";
@@ -198,7 +201,7 @@ import {
   convertMessagesToResponsesInput,
   stablePromptCacheKey,
   toResponsesTools
-} from "./chunk-PUOC4YRO.js";
+} from "./chunk-M7VLQGQP.js";
 import {
   buildNoloWorkspaceOpenAiTools,
   parseNoloWorkspaceToolArguments,
@@ -216,14 +219,14 @@ import {
 } from "./chunk-2MBTWINN.js";
 import {
   normalizeUsage
-} from "./chunk-JNXHAQPJ.js";
+} from "./chunk-LDCAFLXD.js";
 import {
   createProviderCallTimingTracker,
   finalizeProviderCallTiming,
   observeMeaningfulProviderResponse,
   observeMeaningfulProviderResponseAt,
   withProviderCallTimingFields
-} from "./chunk-WS2VEVHS.js";
+} from "./chunk-7MMSLEPA.js";
 import {
   clipCompactText
 } from "./chunk-5MMAK75Q.js";
@@ -1712,6 +1715,44 @@ function resolveThinkingSpec(input) {
 function stringValue(value) {
   return typeof value === "string" && value.length > 0 ? value : void 0;
 }
+function ephemeralCache(slot) {
+  const mode = (process.env.NOLO_ANTHROPIC_CACHE_TTL ?? "turn").trim().toLowerCase();
+  if (mode === "5m") return { type: "ephemeral" };
+  if (mode === "1h") return { type: "ephemeral", ttl: "1h" };
+  return slot === "tail" ? { type: "ephemeral" } : { type: "ephemeral", ttl: "1h" };
+}
+function findTurnStartIndex(messages) {
+  for (let i = messages.length - 1; i >= 0; i--) {
+    const m = messages[i];
+    if (m.role !== "user") continue;
+    const content = Array.isArray(m.content) ? m.content : [];
+    if (content.length === 0) continue;
+    if (content.some((c) => c && c.type === "tool_result")) continue;
+    return i;
+  }
+  return -1;
+}
+function enforceCacheTtlOrder(...groups) {
+  let seenShort = false;
+  const visit = (cc, holder) => {
+    if (!isCacheControl(cc)) return;
+    const rec = cc;
+    if (rec.ttl === "1h" && seenShort) {
+      const { ttl: _ttl, ...rest } = rec;
+      holder.cache_control = rest;
+    } else if (rec.ttl !== "1h") seenShort = true;
+  };
+  for (const group of groups) {
+    for (const item of group) {
+      visit(item.cache_control, item);
+      if (Array.isArray(item.content)) {
+        for (const part of item.content) {
+          if (part && typeof part === "object") visit(part.cache_control, part);
+        }
+      }
+    }
+  }
+}
 function isCacheControl(value) {
   return Boolean(
     value && typeof value === "object" && !Array.isArray(value) && typeof value.type === "string"
@@ -1798,7 +1839,7 @@ function buildAnthropicMessagesBody(args) {
       if (typeof message.content === "string" && Number.isFinite(stablePrefixChars) && stablePrefixChars > 0 && stablePrefixChars <= message.content.length) {
         const stable = message.content.slice(0, stablePrefixChars);
         const dynamic = message.content.slice(stablePrefixChars);
-        system.push({ type: "text", text: stable, cache_control: { type: "ephemeral" } });
+        system.push({ type: "text", text: stable, cache_control: ephemeralCache("prefix") });
         if (dynamic) system.push({ type: "text", text: dynamic });
       } else {
         system.push(...toAnthropicContent(message.content));
@@ -1845,7 +1886,7 @@ function buildAnthropicMessagesBody(args) {
     system.unshift({ type: "text", text: CLAUDE_CODE_SYSTEM_INSTRUCTION });
   }
   if (system.length > 0 && !system.some((block) => isCacheControl(block.cache_control))) {
-    system[system.length - 1].cache_control = { type: "ephemeral" };
+    system[system.length - 1].cache_control = ephemeralCache("prefix");
   }
   const tools = Array.isArray(args.openAiBody.tools) ? args.openAiBody.tools.flatMap((raw) => {
     if (!raw || typeof raw !== "object") return [];
@@ -1859,7 +1900,7 @@ function buildAnthropicMessagesBody(args) {
     }];
   }) : [];
   if (tools.length > 0 && !tools.some((tool) => isCacheControl(tool.cache_control))) {
-    tools[tools.length - 1].cache_control = { type: "ephemeral" };
+    tools[tools.length - 1].cache_control = ephemeralCache("prefix");
   }
   const model = stringValue(args.openAiBody.model) ?? args.agentConfig.model ?? "claude-sonnet-5";
   const maxTokensRaw = args.openAiBody.max_completion_tokens ?? args.openAiBody.max_tokens ?? args.agentConfig.max_tokens;
@@ -1881,15 +1922,22 @@ function buildAnthropicMessagesBody(args) {
   const finalMaxTokens = spec.maxTokens;
   const finalTemperature = spec.temperature;
   if (messages.length > 0) {
-    const lastMessage = messages[messages.length - 1];
-    const content = Array.isArray(lastMessage.content) ? lastMessage.content : [];
-    if (content.length > 0) {
-      const lastBlock = content[content.length - 1];
-      if (!isCacheControl(lastBlock.cache_control)) {
-        lastBlock.cache_control = { type: "ephemeral" };
-      }
+    const mode = (process.env.NOLO_ANTHROPIC_CACHE_TTL ?? "turn").trim().toLowerCase();
+    const lastIdx = messages.length - 1;
+    const startIdx = mode === "5m" ? -1 : findTurnStartIndex(messages);
+    const mark = (idx, slot) => {
+      const content = Array.isArray(messages[idx].content) ? messages[idx].content : [];
+      if (content.length === 0) return;
+      const block = content[content.length - 1];
+      if (!isCacheControl(block.cache_control)) block.cache_control = ephemeralCache(slot);
+    };
+    if (startIdx === lastIdx) mark(lastIdx, "turnStart");
+    else {
+      if (startIdx >= 0) mark(startIdx, "turnStart");
+      mark(lastIdx, "tail");
     }
   }
+  enforceCacheTtlOrder(tools, system, messages);
   return {
     model: normalizeAnthropicWireModel(model),
     messages,
@@ -3189,6 +3237,7 @@ var resolveCliProviderTransport = async (ctx) => {
 };
 
 // packages/agent-runtime/codexResponsesProvider.ts
+import { createHash } from "node:crypto";
 import { randomUUID as randomUUID2 } from "node:crypto";
 var CODEX_RESPONSES_URL = "https://chatgpt.com/backend-api/codex/responses";
 var CODEX_CLIENT_VERSION = "0.159.2";
@@ -3210,11 +3259,25 @@ function decodeCodexAccountId(accessToken) {
     return void 0;
   }
 }
-function createCodexRequestIdentity(installationId = CODEX_INSTALLATION_ID) {
-  const sessionId = randomUUID2();
-  const threadId = randomUUID2();
-  const windowId = randomUUID2();
-  const turnId = randomUUID2();
+var turnStartedAtByTurn = /* @__PURE__ */ new Map();
+function createCodexRequestIdentity(installationId = CODEX_INSTALLATION_ID, stable) {
+  const key = stable?.sessionKey && isStableSessionIdentityEnabled() ? stable.sessionKey : void 0;
+  const sessionId = key ? deriveStableUuid(key, "session") : randomUUID2();
+  const threadId = key ? sessionId : randomUUID2();
+  const windowId = key ? deriveStableUuid(key, "window") : randomUUID2();
+  const turnId = key ? deriveStableUuid(key, `turn:${stable?.turnToken ?? ""}`) : randomUUID2();
+  let turnStartedAt = Date.now();
+  if (key) {
+    const tk = `${key}\0${stable?.turnToken ?? ""}`;
+    const prev = turnStartedAtByTurn.get(tk);
+    if (prev !== void 0) turnStartedAt = prev;
+    else {
+      if (turnStartedAtByTurn.size >= 500) {
+        turnStartedAtByTurn.delete(turnStartedAtByTurn.keys().next().value);
+      }
+      turnStartedAtByTurn.set(tk, turnStartedAt);
+    }
+  }
   const turnMetadata = {
     installation_id: installationId,
     session_id: sessionId,
@@ -3222,7 +3285,7 @@ function createCodexRequestIdentity(installationId = CODEX_INSTALLATION_ID) {
     turn_id: turnId,
     window_id: windowId,
     request_kind: "turn",
-    turn_started_at_unix_ms: Date.now()
+    turn_started_at_unix_ms: turnStartedAt
   };
   const turnMetadataJson = JSON.stringify(turnMetadata);
   return {
@@ -3293,7 +3356,7 @@ function convertMessagesToCodexInput(messages) {
     };
   });
 }
-function buildCodexRequestBody(args, identity = createCodexRequestIdentity()) {
+function buildCodexRequestBody(args, identity = createCodexRequestIdentity(), sessionKey) {
   const rawMessages = Array.isArray(args.openAiBody.messages) ? args.openAiBody.messages : [];
   const nonSystem = rawMessages.filter((m) => {
     const role = m && typeof m === "object" ? String(m.role) : "";
@@ -3313,7 +3376,10 @@ function buildCodexRequestBody(args, identity = createCodexRequestIdentity()) {
     store: false,
     // Keep routing stable across turns and request UUIDs. Growing input/history
     // is intentionally excluded; only stable request-prefix material belongs.
-    prompt_cache_key: stablePromptCacheKey([model, stableInstructions, tools ?? []], "nolo-codex"),
+    // 真实 Codex CLI 用 session_id 作 prompt_cache_key（codex-rs/core/src/client.rs
+    // prompt_cache_key()）；OpenAI 文档：同 key 超过 ~15 req/min 会溢出到其他机器。
+    // 有会话键时按会话分片，无则回退到前缀派生键。
+    prompt_cache_key: sessionKey && isStableSessionIdentityEnabled() ? identity.sessionId : stablePromptCacheKey([model, stableInstructions, tools ?? []], "nolo-codex"),
     client_metadata: identity.clientMetadata
   };
   if (instructions) body.instructions = instructions;
@@ -3336,8 +3402,14 @@ async function readSseEvents(response) {
 var defaultSleep = (ms) => new Promise((resolve) => setTimeout(resolve, ms));
 async function callCodexResponsesOnce(args, accountId, timingTracker) {
   const fetchImpl = args.fetchImpl ?? fetch;
-  const identity = createCodexRequestIdentity();
-  const body = buildCodexRequestBody(args, identity);
+  const msgs = Array.isArray(args.openAiBody.messages) ? args.openAiBody.messages : [];
+  const lastUser = [...msgs].reverse().find((m) => m && m.role === "user");
+  const turnToken = createHash("sha256").update(JSON.stringify(lastUser?.content ?? "")).digest("hex").slice(0, 24);
+  const identity = createCodexRequestIdentity(
+    void 0,
+    args.sessionKey ? { sessionKey: args.sessionKey, turnToken } : void 0
+  );
+  const body = buildCodexRequestBody(args, identity, args.sessionKey);
   const headers = buildCodexCompatibilityHeaders(identity, accountId);
   headers.Authorization = `Bearer ${args.accessToken}`;
   const response = await fetchImpl(CODEX_RESPONSES_URL, {
@@ -3465,6 +3537,7 @@ var resolveCodexOAuthTransport = async (ctx) => {
             stream: false,
             ...tools.length > 0 ? { tools } : {}
           },
+          sessionKey: resolveSessionKey({ dialogId: options?.dialogId }),
           fetchImpl: (url, init) => fetchWithTransientRetry(fetchImpl, url, init, {
             sleep: deps.sleep,
             loopbackRequest

package/{chunk-WGPKQLZO.js → chunk-ZOWXSMIG.js}

@@ -5,14 +5,14 @@ import {
   inlineImageUrlsForCustomProvider,
   prepareTools,
   shouldInlineImageUrlsForAgent
-} from "./chunk-LYKHXIQC.js";
+} from "./chunk-X5YVT5HS.js";
 import {
   INVALID_TOOL_ARGS_REPLACEMENT,
   buildInvalidToolCallSelfHealResult,
   parseToolCallArguments,
   performFetchRequest,
   sanitizeOutboundMessages
-} from "./chunk-KFJ7ZWBF.js";
+} from "./chunk-7JIOW5N3.js";
 import {
   createSSEParser,
   createToolRunId,
@@ -23,7 +23,7 @@ import {
   toolRunSetPending,
   toolRunStarted,
   toolRunSucceeded
-} from "./chunk-ZGNDLCYH.js";
+} from "./chunk-TQE6BVFT.js";
 import {
   getPublicImageAgentMode
 } from "./chunk-VIDPVFP3.js";
@@ -38,10 +38,10 @@ import {
   messageStreamEnd,
   messageStreaming,
   updateToolMessage
-} from "./chunk-BS7YHAQP.js";
+} from "./chunk-UWVGKNAL.js";
 import {
   streamAgentChatTurn
-} from "./chunk-2LM7XUCX.js";
+} from "./chunk-SPCBJGFO.js";
 import {
   addActiveController,
   removeActiveController,
@@ -59,10 +59,10 @@ import {
 import {
   EMPTY_ASSISTANT_REPAIR_PROMPT,
   getModelInfo
-} from "./chunk-JNXHAQPJ.js";
+} from "./chunk-LDCAFLXD.js";
 import {
   write
-} from "./chunk-NVRC52JS.js";
+} from "./chunk-4FFUDT6D.js";
 import {
   getApiEndpoint
 } from "./chunk-3BXNTXYI.js";

package/{chunk-OXEHLZNH.js → chunk-ZTZCZV42.js}

@@ -7,7 +7,7 @@ import {
   listUserRecordsFromServers,
   readDbRecordFromServers,
   readLiveDbRecordAfterTombstoneMerge
-} from "./chunk-RJYTVNWD.js";
+} from "./chunk-RS4ENQ4O.js";
 import {
   fetchWithTransientRetry,
   parsePositiveFiniteNumberOrFallback
@@ -723,7 +723,7 @@ function isExplicitDialogIdMismatch(meta, dialogId) {
   return meta && typeof meta === "object" && "id" in meta && meta.id != null && String(meta.id) !== dialogId;
 }
 async function readDialogFromLocalDb(dialogKey, dialogId, limit, readLocalDialog) {
-  const reader = readLocalDialog ?? (await import("./localDialogRead-3FHUPYAR.js")).readDialogFromLocalDb;
+  const reader = readLocalDialog ?? (await import("./localDialogRead-ZHZ3KXDM.js")).readDialogFromLocalDb;
   const result = await reader({ dialogKey, dialogId, limit });
   return {
     ...result,
@@ -740,7 +740,7 @@ async function readLocalDialogAtDbPathDefault(args) {
   const [{ Level }, { safeJsonEncoding }, { fetchMessages }] = await Promise.all([
     import("./levelLazyShim-GZBKODGI.js"),
     import("./serverStoreFactory-YMXDF4NI.js"),
-    import("./fetchMessages-N2BW2L62.js")
+    import("./fetchMessages-I5ICDDHM.js")
   ]);
   const db = new Level(args.dbPath, {
     valueEncoding: safeJsonEncoding,