ur-agent 1.81.8 → 1.81.10

package/CHANGELOG.md

@@ -1,5 +1,34 @@
 # Changelog
 
+## 1.81.10
+
+- Reduced provider latency and memory pressure without removing the complete
+  system prompt or tool set from real agent work. Self-hosted Ollama, vLLM,
+  LM Studio, llama.cpp, and OpenAI-compatible requests now reserve a realistic
+  4K output turn, resume automatically if it is exhausted, and keep a stable
+  prompt prefix so local/server prefix caches can be reused across sessions.
+- Added a safe lightweight path for exact greetings on every runnable provider.
+  These turns send no irrelevant repository tools, use a 1K output budget, and
+  use an 8K Ollama context; coding, media, structured-output, custom-prompt,
+  and ongoing agent turns retain their full quality-bearing context.
+- Prevented Ollama helper calls from automatically loading a second installed
+  model and evicting the warm session model. A separate helper model remains
+  available through the explicit `OLLAMA_SMALL_FAST_MODEL` override.
+- Added adaptive recovery for vLLM's OpenAI-compatible context-overflow errors,
+  including wrapped/multiline and comma-formatted token counts, and fixed the
+  retry calculation so thinking budgets can never produce another oversized
+  request.
+- Applied provider-discovered completion limits to request sizing, including
+  OpenRouter's live `max_output_tokens`, so defaults and explicit overrides are
+  validated against the selected model's real limit.
+
+## 1.81.9
+
+- Restored the previous compact six-row UR wordmark as the primary welcome
+  logo while retaining the minimal `UR` fallback for narrow terminals. The
+  original house artwork, theme-aware brand color, screen-reader text, and
+  responsive layout remain intact.
+
 ## 1.81.8
 
 - Made reasoning effort provider- and model-accurate across the model picker,

package/dist/cli.js

@@ -53764,6 +53764,7 @@ __export(exports_providerRegistry, {
   getProviderRuntimeBlockReason: () => getProviderRuntimeBlockReason,
   getProviderRuntimeBackend: () => getProviderRuntimeBackend,
   getProviderReasoningCapabilitiesForModel: () => getProviderReasoningCapabilitiesForModel,
+  getProviderOutputTokenLimitForModel: () => getProviderOutputTokenLimitForModel,
   getProviderFamily: () => getProviderFamily,
   getProviderDefinition: () => getProviderDefinition,
   getProviderContextLengthForModel: () => getProviderContextLengthForModel,
@@ -54913,6 +54914,21 @@ function getProviderContextLengthForModel(model, provider = getInitialSettings()
   const length = match?.contextLength;
   return typeof length === "number" && Number.isFinite(length) && length > 0 ? Math.floor(length) : undefined;
 }
+function getProviderOutputTokenLimitForModel(model, provider = getRuntimeProviderId(), settings = getInitialSettings()) {
+  const providerId = resolveProviderId(provider);
+  if (!providerId)
+    return;
+  const wanted = model.trim().toLowerCase();
+  if (!wanted)
+    return;
+  const known = [
+    ...getCachedProviderModels(providerId, settings),
+    ...PROVIDER_MODELS[providerId] ?? []
+  ];
+  const match = known.find((entry) => entry.id.toLowerCase() === wanted) ?? known.find((entry) => wanted.includes(entry.id.toLowerCase()));
+  const limit = match?.outputTokenLimit;
+  return typeof limit === "number" && Number.isFinite(limit) && limit > 0 ? Math.floor(limit) : undefined;
+}
 function getProviderReasoningCapabilitiesForModel(model, provider = getRuntimeProviderId(), settings = getInitialSettings()) {
   const providerId = resolveProviderId(provider);
   if (!providerId)
@@ -57299,9 +57315,19 @@ function getOllamaKeepAlive(env4 = process.env) {
   const asNumber = Number(trimmed);
   return Number.isFinite(asNumber) ? asNumber : trimmed;
 }
-var MIN_AGENT_NUM_CTX = 32768, DEFAULT_OLLAMA_KEEP_ALIVE = "30m", NUM_CTX_BUCKETS, OUTPUT_HEADROOM_TOKENS = 4096;
+var MIN_AGENT_NUM_CTX = 32768, MIN_CHAT_NUM_CTX = 8192, DEFAULT_OLLAMA_KEEP_ALIVE = "30m", NUM_CTX_BUCKETS, OUTPUT_HEADROOM_TOKENS = 4096;
 var init_ollamaTuning = __esm(() => {
-  NUM_CTX_BUCKETS = [32768, 49152, 65536, 98304, 131072, 196608, 262144];
+  NUM_CTX_BUCKETS = [
+    8192,
+    16384,
+    32768,
+    49152,
+    65536,
+    98304,
+    131072,
+    196608,
+    262144
+  ];
 });
 
 // node_modules/react/cjs/react.development.js
@@ -89922,15 +89948,20 @@ function toOllamaChatRequest(params, stream4, capabilities, baseUrl = getEffecti
   if (typeof params.max_tokens === "number") {
     options.num_predict = params.max_tokens;
   }
+  const estimatedPromptTokens = estimateInputTokens(params);
+  const maxTokens = typeof params.max_tokens === "number" ? params.max_tokens : undefined;
+  const isLightweightChat = tools.length === 0 && (maxTokens ?? Number.POSITIVE_INFINITY) <= 1024 && estimatedPromptTokens <= 4096;
   const numCtx = computeOllamaNumCtx({
     modelContextLength: getOllamaContextLengthForModel(params.model, baseUrl),
-    estimatedPromptTokens: estimateInputTokens(params),
-    maxTokens: typeof params.max_tokens === "number" ? params.max_tokens : undefined,
-    override: getOllamaNumCtxOverride()
+    estimatedPromptTokens,
+    maxTokens,
+    override: getOllamaNumCtxOverride(),
+    ...isLightweightChat ? { minCtx: MIN_CHAT_NUM_CTX } : {}
   });
   if (numCtx !== undefined) {
     options.num_ctx = numCtx;
   }
+  logForDebugging(`Ollama request sizing: ~${estimatedPromptTokens} input tokens, ${maxTokens ?? "default"} output tokens, ${tools.length} tools, num_ctx=${numCtx ?? "server-default"}`);
   if (Object.keys(options).length > 0) {
     request.options = options;
   }
@@ -96080,7 +96111,7 @@ function calculateContextPercentages(currentUsage, contextWindowSize) {
     remaining: 100 - clampedUsed
   };
 }
-function getModelMaxOutputTokens(model) {
+function getModelMaxOutputTokens(model, provider, settings) {
   let defaultTokens = MAX_OUTPUT_TOKENS_DEFAULT;
   let upperLimit = MAX_OUTPUT_TOKENS_UPPER_LIMIT;
   if (process.env.USER_TYPE === "ant") {
@@ -96096,6 +96127,11 @@ function getModelMaxOutputTokens(model) {
     upperLimit = cap.max_tokens;
     defaultTokens = Math.min(defaultTokens, upperLimit);
   }
+  const providerOutputLimit = getProviderOutputTokenLimitForModel(model, provider, settings);
+  if (providerOutputLimit !== undefined) {
+    upperLimit = Math.min(upperLimit, providerOutputLimit);
+    defaultTokens = Math.min(defaultTokens, upperLimit);
+  }
   return { default: defaultTokens, upperLimit };
 }
 function getMaxThinkingTokensForModel(model) {
@@ -96808,7 +96844,6 @@ __export(exports_model, {
 });
 function __resetOllamaRouteMemoForTests() {
   memoizedRoutedDefaultModel = undefined;
-  memoizedRoutedFastModel = undefined;
 }
 function getDefaultOllamaModel() {
   const explicitModel = process.env.OLLAMA_MODEL || process.env.UR_MODEL;
@@ -96837,15 +96872,6 @@ function getSmallFastModel(providerSettings) {
     if (process.env.OLLAMA_SMALL_FAST_MODEL) {
       return process.env.OLLAMA_SMALL_FAST_MODEL;
     }
-    if (isOllamaAutoRouteEnabled()) {
-      if (memoizedRoutedFastModel)
-        return memoizedRoutedFastModel;
-      const routed = pickSmallFastModel(getCachedOllamaModelNames());
-      if (routed) {
-        memoizedRoutedFastModel = routed;
-        return routed;
-      }
-    }
     const sessionModel = getMainLoopModel();
     if (sessionModel)
       return sessionModel;
@@ -97166,7 +97192,7 @@ function getMarketingNameForModel(modelId) {
 function normalizeModelStringForAPI(model) {
   return model.replace(/\[(1|2)m\]/gi, "");
 }
-var DEFAULT_OLLAMA_MODEL2 = "qwen2.5-coder:7b", memoizedRoutedDefaultModel, memoizedRoutedFastModel, LEGACY_MODELO_FIRSTPARTY;
+var DEFAULT_OLLAMA_MODEL2 = "qwen2.5-coder:7b", memoizedRoutedDefaultModel, LEGACY_MODELO_FIRSTPARTY;
 var init_model = __esm(() => {
   init_state();
   init_auth();
@@ -108358,7 +108384,7 @@ var init_auth = __esm(() => {
 
 // src/utils/userAgent.ts
 function getURCodeUserAgent() {
-  return `ur/${"1.81.8"}`;
+  return `ur/${"1.81.10"}`;
 }
 
 // src/utils/workloadContext.ts
@@ -108380,7 +108406,7 @@ function getUserAgent() {
   const clientApp = process.env.UR_AGENT_SDK_CLIENT_APP ? `, client-app/${process.env.UR_AGENT_SDK_CLIENT_APP}` : "";
   const workload = getWorkload();
   const workloadSuffix = workload ? `, workload/${workload}` : "";
-  return `ur-cli/${"1.81.8"} (${process.env.USER_TYPE}, ${process.env.UR_CODE_ENTRYPOINT ?? "cli"}${agentSdkVersion}${clientApp}${workloadSuffix})`;
+  return `ur-cli/${"1.81.10"} (${process.env.USER_TYPE}, ${process.env.UR_CODE_ENTRYPOINT ?? "cli"}${agentSdkVersion}${clientApp}${workloadSuffix})`;
 }
 function getMCPUserAgent() {
   const parts = [];
@@ -108394,7 +108420,7 @@ function getMCPUserAgent() {
     parts.push(`client-app/${process.env.UR_AGENT_SDK_CLIENT_APP}`);
   }
   const suffix = parts.length > 0 ? ` (${parts.join(", ")})` : "";
-  return `ur/${"1.81.8"}${suffix}`;
+  return `ur/${"1.81.10"}${suffix}`;
 }
 function getWebFetchUserAgent() {
   return `UR-User (${getURCodeUserAgent()})`;
@@ -108532,7 +108558,7 @@ var init_user = __esm(() => {
       deviceId,
       sessionId: getSessionId(),
       email: getEmail(),
-      appVersion: "1.81.8",
+      appVersion: "1.81.10",
       platform: getHostPlatformForAnalytics(),
       organizationUuid,
       accountUuid,
@@ -116419,7 +116445,7 @@ var init_metadata = __esm(() => {
   COMPOUND_OPERATOR_REGEX = /\s*(?:&&|\|\||[;|])\s*/;
   WHITESPACE_REGEX = /\s+/;
   getVersionBase = memoize_default(() => {
-    const match = "1.81.8".match(/^\d+\.\d+\.\d+(?:-[a-z]+)?/);
+    const match = "1.81.10".match(/^\d+\.\d+\.\d+(?:-[a-z]+)?/);
     return match ? match[0] : undefined;
   });
   buildEnvContext = memoize_default(async () => {
@@ -116459,7 +116485,7 @@ var init_metadata = __esm(() => {
       isGithubAction: isEnvTruthy(process.env.GITHUB_ACTIONS),
       isURCodeAction: isEnvTruthy(process.env.UR_CODE_ACTION),
       isURAiAuth: isURAISubscriber(),
-      version: "1.81.8",
+      version: "1.81.10",
       versionBase: getVersionBase(),
       buildTime: "",
       deploymentEnvironment: env2.detectDeploymentEnvironment(),
@@ -117129,7 +117155,7 @@ function initialize1PEventLogging() {
   const platform2 = getPlatform();
   const attributes = {
     [import_semantic_conventions4.ATTR_SERVICE_NAME]: "ur",
-    [import_semantic_conventions4.ATTR_SERVICE_VERSION]: "1.81.8"
+    [import_semantic_conventions4.ATTR_SERVICE_VERSION]: "1.81.10"
   };
   if (platform2 === "wsl") {
     const wslVersion = getWslVersion();
@@ -117157,7 +117183,7 @@ function initialize1PEventLogging() {
       })
     ]
   });
-  firstPartyEventLogger = firstPartyEventLoggerProvider.getLogger("com.urhq.ur.events", "1.81.8");
+  firstPartyEventLogger = firstPartyEventLoggerProvider.getLogger("com.urhq.ur.events", "1.81.10");
 }
 async function reinitialize1PEventLoggingIfConfigChanged() {
   if (!is1PEventLoggingEnabled() || !firstPartyEventLoggerProvider) {
@@ -127123,7 +127149,7 @@ function formatA2AAgentCard(options = {}, pretty = true) {
 function formatA2AV1AgentCard(options = {}, pretty = true) {
   return JSON.stringify(buildA2AV1AgentCard(options), null, pretty ? 2 : 0);
 }
-var urVersion = "1.81.8", researchSnapshotDate = "2026-08-10", coverage, priorityRoadmap;
+var urVersion = "1.81.10", researchSnapshotDate = "2026-08-10", coverage, priorityRoadmap;
 var init_trends = __esm(() => {
   init_a2aCardSignature();
   coverage = [
@@ -130013,7 +130039,7 @@ function getAttributionHeader(fingerprint) {
   if (!isAttributionHeaderEnabled()) {
     return "";
   }
-  const version2 = `${"1.81.8"}.${fingerprint}`;
+  const version2 = `${"1.81.10"}.${fingerprint}`;
   const entrypoint = process.env.UR_CODE_ENTRYPOINT ?? "unknown";
   const cch = "";
   const workload = getWorkload();
@@ -182998,7 +183024,11 @@ async function* withRetry(getClient, operation, options2) {
             throw error40;
           }
           const minRequired = (retryContext.thinkingConfig.type === "enabled" ? retryContext.thinkingConfig.budgetTokens : 0) + 1;
-          const adjustedMaxTokens = Math.max(FLOOR_OUTPUT_TOKENS, availableContext, minRequired);
+          if (minRequired > availableContext) {
+            logError2(new Error(`thinking budget requires ${minRequired} tokens but only ${availableContext} output tokens fit in the context window`));
+            throw error40;
+          }
+          const adjustedMaxTokens = availableContext;
           retryContext.maxTokensOverride = adjustedMaxTokens;
           logEvent("tengu_max_tokens_context_overflow_adjustment", {
             inputTokens,
@@ -183079,21 +183109,18 @@ function parseMaxTokensContextOverflowError(error40) {
   if (error40.status !== 400 || !error40.message) {
     return;
   }
-  if (!error40.message.includes("input length and `max_tokens` exceed context limit")) {
-    return;
-  }
-  const regex2 = /input length and `max_tokens` exceed context limit: (\d+) \+ (\d+) > (\d+)/;
-  const match = error40.message.match(regex2);
-  if (!match || match.length !== 4) {
+  const arithmeticMatch = error40.message.match(/input length and `max_tokens` exceed context limit:\s*([\d,]+)\s*\+\s*([\d,]+)\s*>\s*([\d,]+)/i);
+  const vllmMatch = error40.message.match(/maximum context length is\s+([\d,]+)\s+tokens[\s\S]*?requested\s+([\d,]+)\s+output tokens[\s\S]*?prompt contains(?:\s+at least)?\s+([\d,]+)\s+input tokens/i);
+  const rawInputTokens = arithmeticMatch?.[1] ?? vllmMatch?.[3];
+  const rawMaxTokens = arithmeticMatch?.[2] ?? vllmMatch?.[2];
+  const rawContextLimit = arithmeticMatch?.[3] ?? vllmMatch?.[1];
+  if (!rawInputTokens || !rawMaxTokens || !rawContextLimit) {
     return;
   }
-  if (!match[1] || !match[2] || !match[3]) {
-    logError2(new Error("Unable to parse max_tokens from max_tokens exceed context limit error message"));
-    return;
-  }
-  const inputTokens = parseInt(match[1], 10);
-  const maxTokens = parseInt(match[2], 10);
-  const contextLimit = parseInt(match[3], 10);
+  const parseTokenCount = (value) => Number.parseInt(value.replaceAll(",", ""), 10);
+  const inputTokens = parseTokenCount(rawInputTokens);
+  const maxTokens = parseTokenCount(rawMaxTokens);
+  const contextLimit = parseTokenCount(rawContextLimit);
   if (isNaN(inputTokens) || isNaN(maxTokens) || isNaN(contextLimit)) {
     return;
   }
@@ -184934,7 +184961,7 @@ var init_projectSafety = __esm(() => {
 function getInstruments() {
   if (instruments)
     return instruments;
-  const meter = import_api10.metrics.getMeter("ur-agent.gen_ai", "1.81.8");
+  const meter = import_api10.metrics.getMeter("ur-agent.gen_ai", "1.81.10");
   instruments = {
     operationDuration: meter.createHistogram("gen_ai.client.operation.duration", {
       description: "GenAI operation duration.",
@@ -185032,7 +185059,7 @@ function genAiAgentAttributes() {
     "gen_ai.operation.name": GEN_AI_OPERATION_INVOKE_AGENT,
     "gen_ai.provider.name": "ur",
     "gen_ai.agent.name": "UR-Nexus",
-    "gen_ai.agent.version": "1.81.8"
+    "gen_ai.agent.version": "1.81.10"
   };
 }
 function genAiWorkflowAttributes(workflowName, workflowRunId) {
@@ -185053,7 +185080,7 @@ function genAiWorkflowAttributes(workflowName, workflowRunId) {
 function startGenAiWorkflowSpan(workflowName, workflowRunId) {
   const attributes = genAiWorkflowAttributes(workflowName, workflowRunId);
   const name = typeof attributes["gen_ai.workflow.name"] === "string" ? `invoke_workflow ${attributes["gen_ai.workflow.name"]}` : GEN_AI_OPERATION_INVOKE_WORKFLOW;
-  return import_api10.trace.getTracer("ur-agent.gen_ai", "1.81.8").startSpan(name, { kind: import_api10.SpanKind.INTERNAL, attributes });
+  return import_api10.trace.getTracer("ur-agent.gen_ai", "1.81.10").startSpan(name, { kind: import_api10.SpanKind.INTERNAL, attributes });
 }
 function endGenAiWorkflowSpan(span, options2 = {}) {
   try {
@@ -185091,7 +185118,7 @@ function startGenAiMemorySpan(operation, options2 = {}) {
   if (options2.recordCount !== undefined && Number.isInteger(options2.recordCount) && options2.recordCount >= 0) {
     attributes["gen_ai.memory.record.count"] = options2.recordCount;
   }
-  return import_api10.trace.getTracer("ur-agent.gen_ai", "1.81.8").startSpan(operation, { kind: import_api10.SpanKind.INTERNAL, attributes });
+  return import_api10.trace.getTracer("ur-agent.gen_ai", "1.81.10").startSpan(operation, { kind: import_api10.SpanKind.INTERNAL, attributes });
 }
 function endGenAiMemorySpan(span, options2 = {}) {
   try {
@@ -278806,7 +278833,7 @@ function getTelemetryAttributes() {
     attributes["session.id"] = sessionId;
   }
   if (shouldIncludeAttribute("OTEL_METRICS_INCLUDE_VERSION")) {
-    attributes["app.version"] = "1.81.8";
+    attributes["app.version"] = "1.81.10";
   }
   const oauthAccount = getOauthAccountInfo();
   if (oauthAccount) {
@@ -320210,7 +320237,7 @@ function getInstallationEnv() {
   return;
 }
 function getURCodeVersion() {
-  return "1.81.8";
+  return "1.81.10";
 }
 async function getInstalledVSCodeExtensionVersion(command) {
   const { stdout } = await execFileNoThrow(command, ["--list-extensions", "--show-versions"], {
@@ -327580,7 +327607,7 @@ async function setupSdkMcpClients(sdkMcpConfigs, sendMcpMessage) {
     const client2 = new Client({
       name: "ur",
       title: "UR",
-      version: "1.81.8",
+      version: "1.81.10",
       description: "UR-Nexus autonomous engineering workflow engine",
       websiteUrl: PRODUCT_URL
     }, {
@@ -327941,7 +327968,7 @@ var init_client5 = __esm(() => {
       const client2 = new Client({
         name: "ur",
         title: "UR",
-        version: "1.81.8",
+        version: "1.81.10",
         description: "UR-Nexus autonomous engineering workflow engine",
         websiteUrl: PRODUCT_URL
       }, {
@@ -340714,7 +340741,7 @@ async function createRuntime() {
   bootstrapTelemetry();
   const resource = defaultResource().merge(detectResources({ detectors: [envDetector] })).merge(resourceFromAttributes({
     [import_semantic_conventions7.ATTR_SERVICE_NAME]: "ur-agent",
-    [import_semantic_conventions7.ATTR_SERVICE_VERSION]: "1.81.8"
+    [import_semantic_conventions7.ATTR_SERVICE_VERSION]: "1.81.10"
   }));
   const tracerProvider = exporters.traces.length > 0 ? new BasicTracerProvider({
     resource,
@@ -340747,11 +340774,11 @@ async function createRuntime() {
   setMeterProvider(meterProvider);
   setLoggerProvider(loggerProvider);
   if (meterProvider) {
-    const meter = meterProvider.getMeter("ur-agent", "1.81.8");
+    const meter = meterProvider.getMeter("ur-agent", "1.81.10");
     setMeter(meter, (name, options2) => meter.createCounter(name, options2));
   }
   if (loggerProvider) {
-    setEventLogger(loggerProvider.getLogger("ur-agent.events", "1.81.8"));
+    setEventLogger(loggerProvider.getLogger("ur-agent.events", "1.81.10"));
   }
   if (!cleanupRegistered3) {
     cleanupRegistered3 = true;
@@ -341413,9 +341440,9 @@ async function assertMinVersion() {
   if (false) {}
   try {
     const versionConfig = await getDynamicConfig_BLOCKS_ON_INIT("tengu_version_config", { minVersion: "0.0.0" });
-    if (versionConfig.minVersion && lt("1.81.8", versionConfig.minVersion)) {
+    if (versionConfig.minVersion && lt("1.81.10", versionConfig.minVersion)) {
       console.error(`
-It looks like your version of UR (${"1.81.8"}) needs an update.
+It looks like your version of UR (${"1.81.10"}) needs an update.
 A newer version (${versionConfig.minVersion} or higher) is required to continue.
 
 To update, please run:
@@ -341631,7 +341658,7 @@ async function installGlobalPackage(specificVersion) {
     logError2(new AutoUpdaterError("Another process is currently installing an update"));
     logEvent("tengu_auto_updater_lock_contention", {
       pid: process.pid,
-      currentVersion: "1.81.8"
+      currentVersion: "1.81.10"
     });
     return "in_progress";
   }
@@ -341640,7 +341667,7 @@ async function installGlobalPackage(specificVersion) {
     if (!env2.isRunningWithBun() && env2.isNpmFromWindowsPath()) {
       logError2(new Error("Windows NPM detected in WSL environment"));
       logEvent("tengu_auto_updater_windows_npm_in_wsl", {
-        currentVersion: "1.81.8"
+        currentVersion: "1.81.10"
       });
       console.error(`
 Error: Windows NPM detected in WSL
@@ -342175,7 +342202,7 @@ function detectLinuxGlobPatternWarnings() {
 }
 async function getDoctorDiagnostic() {
   const installationType = await getCurrentInstallationType();
-  const version2 = typeof MACRO !== "undefined" ? "1.81.8" : "unknown";
+  const version2 = typeof MACRO !== "undefined" ? "1.81.10" : "unknown";
   const installationPath = await getInstallationPath();
   const invokedBinary = getInvokedBinary();
   const multipleInstallations = await detectMultipleInstallations();
@@ -343110,8 +343137,8 @@ async function updateLatest(channelOrVersion, forceReinstall = false) {
     const maxVersion = await getMaxVersion();
     if (maxVersion && gt(version2, maxVersion)) {
       logForDebugging(`Native installer: maxVersion ${maxVersion} is set, capping update from ${version2} to ${maxVersion}`);
-      if (gte("1.81.8", maxVersion)) {
-        logForDebugging(`Native installer: current version ${"1.81.8"} is already at or above maxVersion ${maxVersion}, skipping update`);
+      if (gte("1.81.10", maxVersion)) {
+        logForDebugging(`Native installer: current version ${"1.81.10"} is already at or above maxVersion ${maxVersion}, skipping update`);
         logEvent("tengu_native_update_skipped_max_version", {
           latency_ms: Date.now() - startTime,
           max_version: maxVersion,
@@ -343122,7 +343149,7 @@ async function updateLatest(channelOrVersion, forceReinstall = false) {
       version2 = maxVersion;
     }
   }
-  if (!forceReinstall && version2 === "1.81.8" && await versionIsAvailable(version2) && await isPossibleURBinary(executablePath)) {
+  if (!forceReinstall && version2 === "1.81.10" && await versionIsAvailable(version2) && await isPossibleURBinary(executablePath)) {
     logForDebugging(`Found ${version2} at ${executablePath}, skipping install`);
     logEvent("tengu_native_update_complete", {
       latency_ms: Date.now() - startTime,
@@ -413812,7 +413839,7 @@ function isAnyTracingEnabled() {
   return isTelemetryEnabled() || isEnhancedTelemetryEnabled() || isBetaTracingEnabled();
 }
 function getTracer() {
-  return import_api39.trace.getTracer("ur-agent.gen_ai", "1.81.8");
+  return import_api39.trace.getTracer("ur-agent.gen_ai", "1.81.10");
 }
 function createSpanAttributes(spanType, customAttributes = {}) {
   const baseAttributes = getTelemetryAttributes();
@@ -443951,7 +443978,7 @@ function Feedback({
       platform: env2.platform,
       gitRepo: envInfo.isGit,
       terminal: env2.terminal,
-      version: "1.81.8",
+      version: "1.81.10",
       transcript: normalizeMessagesForAPI(messages),
       errors: sanitizedErrors,
       lastApiRequest: getLastAPIRequest(),
@@ -444143,7 +444170,7 @@ function Feedback({
                       ", ",
                       env2.terminal,
                       ", v",
-                      "1.81.8"
+                      "1.81.10"
                     ]
                   }, undefined, true, undefined, this)
                 ]
@@ -444249,7 +444276,7 @@ ${sanitizedDescription}
 ` + `**Environment Info**
 ` + `- Platform: ${env2.platform}
 ` + `- Terminal: ${env2.terminal}
-` + `- Version: ${"1.81.8"}
+` + `- Version: ${"1.81.10"}
 ` + `- Feedback ID: ${feedbackId}
 ` + `
 **Errors**
@@ -447359,7 +447386,7 @@ function buildPrimarySection() {
   }, undefined, false, undefined, this);
   return [{
     label: "Version",
-    value: "1.81.8"
+    value: "1.81.10"
   }, {
     label: "Session name",
     value: nameValue
@@ -450862,7 +450889,7 @@ function Config({
         }
       }, undefined, false, undefined, this)
     }, undefined, false, undefined, this) : showSubmenu === "ChannelDowngrade" ? /* @__PURE__ */ jsx_dev_runtime179.jsxDEV(ChannelDowngradeDialog, {
-      currentVersion: "1.81.8",
+      currentVersion: "1.81.10",
       onChoice: (choice) => {
         setShowSubmenu(null);
         setTabsHidden(false);
@@ -450874,7 +450901,7 @@ function Config({
           autoUpdatesChannel: "stable"
         };
         if (choice === "stay") {
-          newSettings.minimumVersion = "1.81.8";
+          newSettings.minimumVersion = "1.81.10";
         }
         updateSettingsForSource("userSettings", newSettings);
         setSettingsData((prev_27) => ({
@@ -459183,7 +459210,7 @@ function HelpV2(t0) {
   let t6;
   if ($2[31] !== tabs) {
     t6 = /* @__PURE__ */ jsx_dev_runtime206.jsxDEV(Tabs, {
-      title: `UR v${"1.81.8"}`,
+      title: `UR v${"1.81.10"}`,
       color: "professionalBlue",
       defaultTab: "general",
       children: tabs
@@ -460116,7 +460143,7 @@ function buildToolUseContext(tools, readFileStateCache, toolPermissionContext, a
 async function handleInitialize(options2) {
   return {
     name: "UR",
-    version: "1.81.8",
+    version: "1.81.10",
     protocolVersion: "0.1.0",
     workspaceRoot: options2.cwd,
     capabilities: {
@@ -477224,7 +477251,7 @@ function getAllReleaseNotes(changelogContent = getStoredChangelogFromMemory()) {
     return [];
   }
 }
-async function checkForReleaseNotes(lastSeenVersion, currentVersion = "1.81.8") {
+async function checkForReleaseNotes(lastSeenVersion, currentVersion = "1.81.10") {
   if (process.env.USER_TYPE === "ant") {
     const changelog = "";
     if (changelog) {
@@ -477251,7 +477278,7 @@ async function checkForReleaseNotes(lastSeenVersion, currentVersion = "1.81.8")
     releaseNotes
   };
 }
-function checkForReleaseNotesSync(lastSeenVersion, currentVersion = "1.81.8") {
+function checkForReleaseNotesSync(lastSeenVersion, currentVersion = "1.81.10") {
   if (process.env.USER_TYPE === "ant") {
     const changelog = "";
     if (changelog) {
@@ -480159,7 +480186,7 @@ function getRecentActivitySync() {
   return cachedActivity;
 }
 function getLogoDisplayData() {
-  const version2 = process.env.DEMO_VERSION ?? "1.81.8";
+  const version2 = process.env.DEMO_VERSION ?? "1.81.10";
   const serverUrl = getDirectConnectServerUrl();
   const displayPath = process.env.DEMO_VERSION ? "/code/ur" : getDisplayPath(getCwd());
   const cwd2 = serverUrl ? `${displayPath} in ${serverUrl.replace(/^https?:\/\//, "")}` : displayPath;
@@ -480234,8 +480261,6 @@ var init_logoV2Utils = __esm(() => {
 
 // src/components/LogoV2/URBanner.tsx
 function getResponsiveURWordmarkVariant(availableWidth = Number.POSITIVE_INFINITY, availableHeight = Number.POSITIVE_INFINITY) {
-  if (availableWidth >= 41 && availableHeight >= 22)
-    return "full";
   if (availableWidth >= 20 && availableHeight >= 8)
     return "compact";
   return "minimal";
@@ -480250,7 +480275,7 @@ function URBanner({
     }, undefined, false, undefined, this);
   }
   const variant = getResponsiveURWordmarkVariant(availableWidth, availableHeight);
-  const rows = variant === "full" ? UR_WORDMARK_ROWS : variant === "compact" ? COMPACT_UR_WORDMARK_ROWS : ["UR"];
+  const rows = variant === "compact" ? UR_WORDMARK_ROWS : ["UR"];
   const showTagline = variant !== "minimal";
   const decorateTagline = (availableWidth ?? Number.POSITIVE_INFINITY) >= 24;
   return /* @__PURE__ */ jsx_dev_runtime233.jsxDEV(ThemedBox_default, {
@@ -480268,30 +480293,12 @@ function URBanner({
     ]
   }, undefined, true, undefined, this);
 }
-var jsx_dev_runtime233, UR_WORDMARK_ROWS, COMPACT_UR_WORDMARK_ROWS, UR_WORDMARK_TAGLINE = "THE AUTONOMOUS AGENT";
+var jsx_dev_runtime233, UR_WORDMARK_ROWS, UR_WORDMARK_TAGLINE = "THE AUTONOMOUS AGENT";
 var init_URBanner = __esm(() => {
   init_ink2();
   init_screenReader();
   jsx_dev_runtime233 = __toESM(require_jsx_dev_runtime(), 1);
   UR_WORDMARK_ROWS = [
-    "UUUUUUUU     UUUUUUUURRRRRRRRRRRRRRRRR   ",
-    "U::::::U     U::::::UR::::::::::::::::R  ",
-    "U::::::U     U::::::UR::::::RRRRRR:::::R ",
-    "UU:::::U     U:::::UURR:::::R     R:::::R",
-    " U:::::U     U:::::U   R::::R     R:::::R",
-    " U:::::D     D:::::U   R::::R     R:::::R",
-    " U:::::D     D:::::U   R::::RRRRRR:::::R ",
-    " U:::::D     D:::::U   R:::::::::::::RR  ",
-    " U:::::D     D:::::U   R::::RRRRRR:::::R ",
-    " U:::::D     D:::::U   R::::R     R:::::R",
-    " U:::::D     D:::::U   R::::R     R:::::R",
-    " U::::::U   U::::::U   R::::R     R:::::R",
-    " U:::::::UUU:::::::U RR:::::R     R:::::R",
-    "  UU:::::::::::::UU  R::::::R     R:::::R",
-    "    UU:::::::::UU    R::::::R     R:::::R",
-    "      UUUUUUUUU      RRRRRRRR     RRRRRRR"
-  ];
-  COMPACT_UR_WORDMARK_ROWS = [
     "\u2588\u2588\u2557   \u2588\u2588\u2557  \u2588\u2588\u2588\u2588\u2588\u2588\u2557 ",
     "\u2588\u2588\u2551   \u2588\u2588\u2551  \u2588\u2588\u2554\u2550\u2550\u2588\u2588\u2557",
     "\u2588\u2588\u2551   \u2588\u2588\u2551  \u2588\u2588\u2588\u2588\u2588\u2588\u2554\u255D",
@@ -481067,7 +481074,7 @@ function LogoV2() {
   if ($2[2] === Symbol.for("react.memo_cache_sentinel")) {
     t2 = () => {
       const currentConfig = getGlobalConfig();
-      if (currentConfig.lastReleaseNotesSeen === "1.81.8") {
+      if (currentConfig.lastReleaseNotesSeen === "1.81.10") {
         return;
       }
       saveGlobalConfig(_temp325);
@@ -481755,12 +481762,12 @@ function LogoV2() {
   return t41;
 }
 function _temp325(current) {
-  if (current.lastReleaseNotesSeen === "1.81.8") {
+  if (current.lastReleaseNotesSeen === "1.81.10") {
     return current;
   }
   return {
     ...current,
-    lastReleaseNotesSeen: "1.81.8"
+    lastReleaseNotesSeen: "1.81.10"
   };
 }
 function _temp241(s_0) {
@@ -497710,7 +497717,7 @@ function compileAgenticCiWorkflow(specName = "default", options2 = {}) {
   if (spec.name !== specName) {
     throw new Error("Agentic CI workflow spec name does not match");
   }
-  const packageVersion = options2.packageVersion ?? (typeof MACRO !== "undefined" ? "1.81.8" : "1.81.8");
+  const packageVersion = options2.packageVersion ?? (typeof MACRO !== "undefined" ? "1.81.10" : "1.81.10");
   if (!/^\d+\.\d+\.\d+(?:-[0-9A-Za-z.-]+)?$/.test(packageVersion)) {
     throw new Error("invalid ur-agent package version");
   }
@@ -498703,7 +498710,7 @@ runner; \`ur trigger\` is the inbound parser that decides what to run.
       path: ".github/workflows/ur.yml",
       root: "project",
       content: compileAgenticCiWorkflow("default", {
-        packageVersion: typeof MACRO !== "undefined" ? "1.81.8" : "1.81.8"
+        packageVersion: typeof MACRO !== "undefined" ? "1.81.10" : "1.81.10"
       })
     },
     {
@@ -498766,7 +498773,7 @@ function value(tokens, flag) {
   return index2 >= 0 ? tokens[index2 + 1] : undefined;
 }
 function cliVersion() {
-  return typeof MACRO !== "undefined" ? "1.81.8" : "1.81.8";
+  return typeof MACRO !== "undefined" ? "1.81.10" : "1.81.10";
 }
 function workflowPath(cwd2) {
   return join168(cwd2, ".github", "workflows", "ur-agentic-ci.yml");
@@ -504622,7 +504629,7 @@ function createAcpStdioApp(deps) {
       }
     },
     authMethods: [],
-    agentInfo: { name: "UR-Nexus", version: "1.81.8" }
+    agentInfo: { name: "UR-Nexus", version: "1.81.10" }
   })).onRequest("authenticate", () => ({})).onRequest("session/new", async (context6) => {
     const result = runtime2.newSession(context6.params.cwd, context6.params.mcpServers, context6.params.additionalDirectories);
     await runtime2.announce({
@@ -504719,7 +504726,7 @@ function createAcpStdioAgent(deps) {
               }
             },
             authMethods: [],
-            agentInfo: { name: "UR-Nexus", version: "1.81.8" }
+            agentInfo: { name: "UR-Nexus", version: "1.81.10" }
           });
           return;
         case "authenticate":
@@ -715971,7 +715978,7 @@ async function captureMemoryDiagnostics(trigger2, dumpNumber = 0) {
     smapsRollup,
     platform: process.platform,
     nodeVersion: process.version,
-    ccVersion: "1.81.8"
+    ccVersion: "1.81.10"
   };
 }
 async function performHeapDump(trigger2 = "manual", dumpNumber = 0) {
@@ -716560,7 +716567,7 @@ var init_bridge_kick = __esm(() => {
 var call154 = async () => {
   return {
     type: "text",
-    value: "1.81.8"
+    value: "1.81.10"
   };
 }, version2, version_default;
 var init_version = __esm(() => {
@@ -728028,7 +728035,7 @@ function generateHtmlReport(data, insights) {
 </html>`;
 }
 function buildExportData(data, insights, facets, remoteStats) {
-  const version3 = typeof MACRO !== "undefined" ? "1.81.8" : "unknown";
+  const version3 = typeof MACRO !== "undefined" ? "1.81.10" : "unknown";
   const remote_hosts_collected = remoteStats?.hosts.filter((h2) => h2.sessionCount > 0).map((h2) => h2.name);
   const facets_summary = {
     total: facets.size,
@@ -732341,7 +732348,7 @@ var init_sessionStorage = __esm(() => {
   init_settings2();
   init_slowOperations();
   init_uuid();
-  VERSION7 = typeof MACRO !== "undefined" ? "1.81.8" : "unknown";
+  VERSION7 = typeof MACRO !== "undefined" ? "1.81.10" : "unknown";
   MAX_TOMBSTONE_REWRITE_BYTES = 50 * 1024 * 1024;
   SKIP_FIRST_PROMPT_PATTERN = /^(?:\s*<[a-z][\w-]*[\s>]|\[Request interrupted by user[^\]]*\])/;
   EPHEMERAL_PROGRESS_TYPES = new Set([
@@ -733556,7 +733563,7 @@ var init_filesystem = __esm(() => {
   });
   getBundledSkillsRoot = memoize_default(function getBundledSkillsRoot2() {
     const nonce = randomBytes24(16).toString("hex");
-    return join243(getURTempDir(), "bundled-skills", "1.81.8", nonce);
+    return join243(getURTempDir(), "bundled-skills", "1.81.10", nonce);
   });
   getResolvedWorkingDirPaths = memoize_default(getPathsForPermissionCheck);
 });
@@ -739956,11 +739963,81 @@ function computeFingerprint(messageText2, version3) {
 }
 function computeFingerprintFromMessages(messages) {
   const firstMessageText = extractFirstMessageText(messages);
-  return computeFingerprint(firstMessageText, "1.81.8");
+  return computeFingerprint(firstMessageText, "1.81.10");
 }
 var FINGERPRINT_SALT = "59cf53e54c78";
 var init_fingerprint = () => {};
 
+// src/utils/model/providerRequestTuning.ts
+function usesConservativeOutputReservation(provider) {
+  return CONSERVATIVE_OUTPUT_PROVIDERS.has(provider);
+}
+function getProviderRequestProfile(input) {
+  if (!LIGHTWEIGHT_CHAT_PROVIDERS.has(input.provider) || !(input.querySource.startsWith("repl_main_thread") || input.querySource === "sdk")) {
+    return { mode: "agent" };
+  }
+  if (input.messages.some((message) => message.type === "assistant")) {
+    return { mode: "agent" };
+  }
+  const visibleUserMessages = input.messages.filter((message) => message.type === "user" && !message.isMeta && !message.isCompactSummary && !message.toolUseResult);
+  if (visibleUserMessages.length !== 1) {
+    return { mode: "agent" };
+  }
+  const prompt = textOnlyContent(visibleUserMessages[0]);
+  if (prompt === undefined || !isTrivialGreeting(prompt)) {
+    return { mode: "agent" };
+  }
+  return {
+    mode: "lightweight-chat",
+    maxOutputTokens: LIGHTWEIGHT_CHAT_MAX_OUTPUT_TOKENS,
+    minContextTokens: LIGHTWEIGHT_CHAT_MIN_CONTEXT_TOKENS,
+    systemPrompt: LIGHTWEIGHT_CHAT_SYSTEM_PROMPT
+  };
+}
+function textOnlyContent(message) {
+  const content = message?.message?.content;
+  if (typeof content === "string")
+    return content.trim();
+  if (!Array.isArray(content) || content.length === 0)
+    return;
+  if (content.some((block2) => !block2 || typeof block2 !== "object" || block2.type !== "text" || typeof block2.text !== "string")) {
+    return;
+  }
+  return content.map((block2) => block2.text).join(`
+`).trim();
+}
+function isTrivialGreeting(prompt) {
+  const normalized = prompt.trim().toLowerCase();
+  return /^(?:hi|hello|hey|hiya|greetings|good (?:morning|afternoon|evening)|thanks|thank you|ok|okay)[!.?,\s]*$/.test(normalized);
+}
+var CONSERVATIVE_OUTPUT_PROVIDERS, LIGHTWEIGHT_CHAT_PROVIDERS, LIGHTWEIGHT_CHAT_MAX_OUTPUT_TOKENS = 1024, LIGHTWEIGHT_CHAT_MIN_CONTEXT_TOKENS, SELF_HOSTED_DEFAULT_MAX_OUTPUT_TOKENS = 4096, LIGHTWEIGHT_CHAT_SYSTEM_PROMPT = "You are UR, a precise and professional AI assistant. Respond directly, clearly, and concisely. Do not claim to have inspected files or used tools in this lightweight chat turn.";
+var init_providerRequestTuning = __esm(() => {
+  init_ollamaTuning();
+  CONSERVATIVE_OUTPUT_PROVIDERS = new Set([
+    "ollama",
+    "lmstudio",
+    "llama.cpp",
+    "vllm",
+    "openai-compatible"
+  ]);
+  LIGHTWEIGHT_CHAT_PROVIDERS = new Set([
+    "ollama",
+    "lmstudio",
+    "llama.cpp",
+    "vllm",
+    "openai-compatible",
+    "openai-api",
+    "anthropic-api",
+    "gemini-api",
+    "openrouter",
+    "codex-cli",
+    "claude-code-cli",
+    "gemini-cli",
+    "antigravity-cli"
+  ]);
+  LIGHTWEIGHT_CHAT_MIN_CONTEXT_TOKENS = MIN_CHAT_NUM_CTX;
+});
+
 // src/services/compact/apiMicrocompact.ts
 function getAPIContextManagement(options5) {
   const {
@@ -740492,6 +740569,15 @@ async function* queryModel(messages, systemPrompt, thinkingConfig, tools, signal
   }
   const previousRequestId = getPreviousRequestIdFromMessages(messages);
   const resolvedModel = isBedrockRuntime() && options5.model.includes("application-inference-profile") ? await getInferenceProfileBackingModel(options5.model) ?? options5.model : options5.model;
+  const requestProvider = options5.providerSettings?.active ?? getRuntimeProvider();
+  const candidateRequestProfile = getProviderRequestProfile({
+    provider: requestProvider,
+    querySource: options5.querySource,
+    messages
+  });
+  const providerRequestProfile = candidateRequestProfile.mode === "lightweight-chat" && !options5.hasAppendSystemPrompt && !options5.outputFormat && !options5.toolChoice && (options5.extraToolSchemas?.length ?? 0) === 0 ? candidateRequestProfile : { mode: "agent" };
+  const isLightweightChat = providerRequestProfile.mode === "lightweight-chat";
+  logForDebugging(`Provider request profile: ${providerRequestProfile.mode} (${requestProvider}, source=${options5.querySource}, messages=${messages.length})`);
   queryCheckpoint("query_tool_schema_build_start");
   const isAgenticQuery = options5.querySource.startsWith("repl_main_thread") || options5.querySource.startsWith("agent:") || options5.querySource === "sdk" || options5.querySource === "hook_agent" || options5.querySource === "verification_agent";
   const betas = getMergedBetas(options5.model, { isAgenticQuery });
@@ -740499,7 +740585,7 @@ async function* queryModel(messages, systemPrompt, thinkingConfig, tools, signal
     betas.push(ADVISOR_BETA_HEADER);
   }
   let advisorModel;
-  if (isAgenticQuery && isAdvisorEnabled()) {
+  if (isAgenticQuery && isAdvisorEnabled() && !isLightweightChat) {
     let advisorOption = options5.advisorModel;
     const advisorExperiment = getExperimentAdvisorModels();
     if (advisorExperiment !== undefined) {
@@ -740519,7 +740605,7 @@ async function* queryModel(messages, systemPrompt, thinkingConfig, tools, signal
       }
     }
   }
-  let useToolSearch = await isToolSearchEnabled(options5.model, tools, options5.getToolPermissionContext, options5.agents, "query");
+  let useToolSearch = isLightweightChat ? false : await isToolSearchEnabled(options5.model, tools, options5.getToolPermissionContext, options5.agents, "query");
   const deferredToolNames = new Set;
   if (useToolSearch) {
     for (const t of tools) {
@@ -740532,7 +740618,9 @@ async function* queryModel(messages, systemPrompt, thinkingConfig, tools, signal
     useToolSearch = false;
   }
   let filteredTools;
-  if (useToolSearch) {
+  if (isLightweightChat) {
+    filteredTools = [];
+  } else if (useToolSearch) {
     const discoveredToolNames = extractDiscoveredToolNames(messages);
     filteredTools = tools.filter((tool) => {
       if (!deferredToolNames.has(tool.name))
@@ -740576,7 +740664,8 @@ async function* queryModel(messages, systemPrompt, thinkingConfig, tools, signal
     preNormalizedMessageCount: messages.length
   });
   queryCheckpoint("query_message_normalization_start");
-  let messagesForAPI = normalizeMessagesForAPI(messages, filteredTools);
+  const requestMessages = isLightweightChat ? messages.filter((message) => message.type === "user" && !message.isMeta && !message.isCompactSummary && !message.toolUseResult) : messages;
+  let messagesForAPI = normalizeMessagesForAPI(requestMessages, filteredTools);
   queryCheckpoint("query_message_normalization_end");
   if (!useToolSearch) {
     messagesForAPI = messagesForAPI.map((msg) => {
@@ -740616,8 +740705,11 @@ ${deferredToolList}
   }
   const hasChromeTools = filteredTools.some((t) => isToolFromMcpServer(t.name, UR_IN_CHROME_MCP_SERVER_NAME));
   const injectChromeHere = useToolSearch && hasChromeTools && !isMcpInstructionsDeltaEnabled();
+  if (providerRequestProfile.mode === "lightweight-chat") {
+    systemPrompt = asSystemPrompt([providerRequestProfile.systemPrompt]);
+  }
   systemPrompt = asSystemPrompt([
-    getAttributionHeader(fingerprint2),
+    usesConservativeOutputReservation(requestProvider) ? "" : getAttributionHeader(fingerprint2),
     getCLISyspromptPrefix({
       isNonInteractive: options5.isNonInteractiveSession,
       hasAppendSystemPrompt: options5.hasAppendSystemPrompt
@@ -740633,7 +740725,7 @@ ${deferredToolList}
     querySource: options5.querySource
   });
   const useBetas = betas.length > 0;
-  const extraToolSchemas = [...options5.extraToolSchemas ?? []];
+  const extraToolSchemas = isLightweightChat ? [] : [...options5.extraToolSchemas ?? []];
   if (advisorModel) {
     extraToolSchemas.push({
       type: "advisor_20260301",
@@ -740702,7 +740794,9 @@ ${deferredToolList}
     const outputConfig = {
       ...extraBodyParams.output_config ?? {}
     };
-    configureEffortParams(effort, outputConfig, extraBodyParams, betasParams, options5.model, options5.providerSettings?.active);
+    if (!isLightweightChat) {
+      configureEffortParams(effort, outputConfig, extraBodyParams, betasParams, options5.model, options5.providerSettings?.active);
+    }
     configureTaskBudgetParams(options5.taskBudget, outputConfig, betasParams);
     if (options5.outputFormat && !("format" in outputConfig)) {
       outputConfig.format = options5.outputFormat;
@@ -740710,8 +740804,9 @@ ${deferredToolList}
         betasParams.push(STRUCTURED_OUTPUTS_BETA_HEADER);
       }
     }
-    const maxOutputTokens2 = retryContext?.maxTokensOverride || options5.maxOutputTokensOverride || getMaxOutputTokensForModel(options5.model);
-    const hasThinking = thinkingConfig.type !== "disabled" && !isEnvTruthy(process.env.UR_CODE_DISABLE_THINKING);
+    const configuredMaxOutputTokens = retryContext?.maxTokensOverride || options5.maxOutputTokensOverride || getMaxOutputTokensForModel(options5.model, options5.providerSettings?.active ?? getRuntimeProvider());
+    const maxOutputTokens2 = providerRequestProfile.mode === "lightweight-chat" ? Math.min(configuredMaxOutputTokens, providerRequestProfile.maxOutputTokens) : configuredMaxOutputTokens;
+    const hasThinking = !isLightweightChat && thinkingConfig.type !== "disabled" && !isEnvTruthy(process.env.UR_CODE_DISABLE_THINKING);
     let thinking = undefined;
     if (hasThinking && modelSupportsThinking(options5.model)) {
       if (!isEnvTruthy(process.env.UR_CODE_DISABLE_ADAPTIVE_THINKING) && modelSupportsAdaptiveThinking(options5.model)) {
@@ -741782,9 +741877,9 @@ function adjustParamsForNonStreaming(params, maxTokensCap) {
 function isMaxTokensCapEnabled() {
   return getFeatureValue_CACHED_MAY_BE_STALE("tengu_otk_slot_v1", false);
 }
-function getMaxOutputTokensForModel(model) {
-  const maxOutputTokens = getModelMaxOutputTokens(model);
-  const defaultTokens = isMaxTokensCapEnabled() ? Math.min(maxOutputTokens.default, CAPPED_DEFAULT_MAX_TOKENS) : maxOutputTokens.default;
+function getMaxOutputTokensForModel(model, provider = getRuntimeProvider()) {
+  const maxOutputTokens = getModelMaxOutputTokens(model, provider);
+  const defaultTokens = usesConservativeOutputReservation(provider) ? Math.min(maxOutputTokens.default, SELF_HOSTED_DEFAULT_MAX_OUTPUT_TOKENS) : isMaxTokensCapEnabled() ? Math.min(maxOutputTokens.default, CAPPED_DEFAULT_MAX_TOKENS) : maxOutputTokens.default;
   const result = validateBoundedIntEnvVar("UR_CODE_MAX_OUTPUT_TOKENS", process.env.UR_CODE_MAX_OUTPUT_TOKENS, defaultTokens, maxOutputTokens.upperLimit);
   return result.effective;
 }
@@ -741805,6 +741900,7 @@ var init_ur2 = __esm(() => {
   init_log2();
   init_messages();
   init_model();
+  init_providerRequestTuning();
   init_tokens();
   init_growthbook();
   init_urAiLimits();
@@ -741891,7 +741987,7 @@ async function sideQuery(opts) {
     betas.push(STRUCTURED_OUTPUTS_BETA_HEADER);
   }
   const messageText2 = extractFirstUserMessageText(messages);
-  const fingerprint2 = computeFingerprint(messageText2, "1.81.8");
+  const fingerprint2 = computeFingerprint(messageText2, "1.81.10");
   const attributionHeader = getAttributionHeader(fingerprint2);
   const systemBlocks = [
     attributionHeader ? { type: "text", text: attributionHeader } : null,
@@ -746725,7 +746821,7 @@ function buildSystemInitMessage(inputs) {
     slash_commands: inputs.commands.filter((c4) => c4.userInvocable !== false).map((c4) => c4.name),
     apiKeySource: getURHQApiKeyWithSource().source,
     betas: getSdkBetas(),
-    ur_version: "1.81.8",
+    ur_version: "1.81.10",
     output_style: outputStyle,
     agents: inputs.agents.map((agent2) => agent2.agentType),
     skills: inputs.skills.filter((s) => s.userInvocable !== false).map((skill2) => skill2.name),
@@ -760561,7 +760657,7 @@ var init_useVoiceEnabled = __esm(() => {
 function getSemverPart(version3) {
   return `${import_semver13.major(version3, { loose: true })}.${import_semver13.minor(version3, { loose: true })}.${import_semver13.patch(version3, { loose: true })}`;
 }
-function useUpdateNotification(updatedVersion, initialVersion = "1.81.8") {
+function useUpdateNotification(updatedVersion, initialVersion = "1.81.10") {
   const [lastNotifiedSemver, setLastNotifiedSemver] = import_react225.useState(() => getSemverPart(initialVersion));
   if (!updatedVersion) {
     return null;
@@ -760610,7 +760706,7 @@ function AutoUpdater({
       return;
     }
     if (false) {}
-    const currentVersion = "1.81.8";
+    const currentVersion = "1.81.10";
     const channel = getInitialSettings()?.autoUpdatesChannel ?? "latest";
     let latestVersion = await getLatestVersion(channel);
     const isDisabled = isAutoUpdaterDisabled();
@@ -760839,12 +760935,12 @@ function NativeAutoUpdater({
     logEvent("tengu_native_auto_updater_start", {});
     try {
       const maxVersion = await getMaxVersion();
-      if (maxVersion && gt("1.81.8", maxVersion)) {
+      if (maxVersion && gt("1.81.10", maxVersion)) {
         const msg = await getMaxVersionMessage();
         setMaxVersionIssue(msg ?? "affects your version");
       }
       const result = await installLatest(channel);
-      const currentVersion = "1.81.8";
+      const currentVersion = "1.81.10";
       const latencyMs = Date.now() - startTime;
       if (result.lockFailed) {
         logEvent("tengu_native_auto_updater_lock_contention", {
@@ -760981,17 +761077,17 @@ function PackageManagerAutoUpdater(t0) {
       const maxVersion = await getMaxVersion();
       if (maxVersion && latest && gt(latest, maxVersion)) {
         logForDebugging(`PackageManagerAutoUpdater: maxVersion ${maxVersion} is set, capping update from ${latest} to ${maxVersion}`);
-        if (gte("1.81.8", maxVersion)) {
-          logForDebugging(`PackageManagerAutoUpdater: current version ${"1.81.8"} is already at or above maxVersion ${maxVersion}, skipping update`);
+        if (gte("1.81.10", maxVersion)) {
+          logForDebugging(`PackageManagerAutoUpdater: current version ${"1.81.10"} is already at or above maxVersion ${maxVersion}, skipping update`);
           setUpdateAvailable(false);
           return;
         }
         latest = maxVersion;
       }
-      const hasUpdate = latest && !gte("1.81.8", latest) && !shouldSkipVersion(latest);
+      const hasUpdate = latest && !gte("1.81.10", latest) && !shouldSkipVersion(latest);
       setUpdateAvailable(!!hasUpdate);
       if (hasUpdate) {
-        logForDebugging(`PackageManagerAutoUpdater: Update available ${"1.81.8"} -> ${latest}`);
+        logForDebugging(`PackageManagerAutoUpdater: Update available ${"1.81.10"} -> ${latest}`);
       }
     };
     $2[0] = t1;
@@ -761025,7 +761121,7 @@ function PackageManagerAutoUpdater(t0) {
       wrap: "truncate",
       children: [
         "currentVersion: ",
-        "1.81.8"
+        "1.81.10"
       ]
     }, undefined, true, undefined, this);
     $2[3] = verbose;
@@ -771878,7 +771974,7 @@ function buildStatusLineCommandInput(permissionMode, exceeds200kTokens, settings
       project_dir: getOriginalCwd(),
       added_dirs: addedDirs
     },
-    version: "1.81.8",
+    version: "1.81.10",
     output_style: {
       name: outputStyleName
     },
@@ -772013,7 +772109,7 @@ function StatusLineInner({
   const attention = customStatusError ?? taskAttention;
   const terminalSize = React133.useContext(TerminalSizeContext);
   const defaultStatusLineText = buildDefaultStatusBar({
-    version: "1.81.8",
+    version: "1.81.10",
     providerLabel: providerRuntime.providerLabel,
     authMode: providerRuntime.authLabel,
     model: renderModelName(mainLoopModel) || providerRuntime.model || "",
@@ -784269,7 +784365,7 @@ async function submitTranscriptShare(messages, trigger2, appearanceId) {
     } catch {}
     const data = {
       trigger: trigger2,
-      version: "1.81.8",
+      version: "1.81.10",
       platform: process.platform,
       transcript,
       subagentTranscripts: Object.keys(subagentTranscripts).length > 0 ? subagentTranscripts : undefined,
@@ -796638,7 +796734,7 @@ function WelcomeV2() {
             dimColor: true,
             children: [
               "v",
-              "1.81.8"
+              "1.81.10"
             ]
           }, undefined, true, undefined, this)
         ]
@@ -797898,7 +797994,7 @@ function completeOnboarding() {
   saveGlobalConfig((current) => ({
     ...current,
     hasCompletedOnboarding: true,
-    lastOnboardingVersion: "1.81.8"
+    lastOnboardingVersion: "1.81.10"
   }));
 }
 function showDialog(root2, renderer) {
@@ -803044,7 +803140,7 @@ function appendToLog(path28, message) {
     cwd: getFsImplementation().cwd(),
     userType: process.env.USER_TYPE,
     sessionId: getSessionId(),
-    version: "1.81.8"
+    version: "1.81.10"
   };
   getLogWriter(path28).write(messageWithTimestamp);
 }
@@ -807208,8 +807304,8 @@ async function getEnvLessBridgeConfig() {
 }
 async function checkEnvLessBridgeMinVersion() {
   const cfg = await getEnvLessBridgeConfig();
-  if (cfg.min_version && lt("1.81.8", cfg.min_version)) {
-    return `Your version of UR (${"1.81.8"}) is too old for Remote Control.
+  if (cfg.min_version && lt("1.81.10", cfg.min_version)) {
+    return `Your version of UR (${"1.81.10"}) is too old for Remote Control.
 Version ${cfg.min_version} or higher is required. Run \`ur update\` to update.`;
   }
   return null;
@@ -807683,7 +807779,7 @@ async function initBridgeCore(params) {
   const rawApi = createBridgeApiClient({
     baseUrl,
     getAccessToken,
-    runnerVersion: "1.81.8",
+    runnerVersion: "1.81.10",
     onDebug: logForDebugging,
     onAuth401,
     getTrustedDeviceToken
@@ -817159,7 +817255,7 @@ function getAgUiCapabilities() {
       name: "UR-Nexus",
       type: "ur-nexus",
       description: "Provider-flexible, local-first autonomous engineering workflow agent.",
-      version: "1.81.8",
+      version: "1.81.10",
       provider: "UR",
       documentationUrl: "https://github.com/Maitham16/UR/blob/master/docs/AG_UI.md"
     },
@@ -818386,7 +818482,7 @@ function createMCPServer(cwd4, debug2, verbose) {
   };
   const server2 = new Server({
     name: "ur-nexus",
-    version: "1.81.8"
+    version: "1.81.10"
   }, {
     capabilities: {
       tools: {}
@@ -819590,7 +819686,7 @@ function thrownResponse(error40) {
 }
 async function createUrMcp2026Runtime(options5) {
   const server2 = createMCPServer(options5.cwd, options5.debug === true, options5.verbose === true);
-  const client2 = new Client({ name: "ur-mcp-2026-adapter", version: "1.81.8" }, { capabilities: {} });
+  const client2 = new Client({ name: "ur-mcp-2026-adapter", version: "1.81.10" }, { capabilities: {} });
   const [clientTransport, serverTransport] = createLinkedTransportPair();
   try {
     await server2.connect(serverTransport);
@@ -819601,7 +819697,7 @@ async function createUrMcp2026Runtime(options5) {
   }
   const runtime2 = new Mcp2026Runtime({
     cwd: options5.cwd,
-    version: "1.81.8",
+    version: "1.81.10",
     backend: {
       listTools: async () => {
         const listed = await client2.listTools();
@@ -822336,7 +822432,7 @@ async function update() {
   logEvent("tengu_update_check", {});
   const diagnostic2 = await getDoctorDiagnostic();
   const result = await checkUpgradeStatus({
-    currentVersion: "1.81.8",
+    currentVersion: "1.81.10",
     packageName: UR_AGENT_PACKAGE_NAME,
     installationType: diagnostic2.installationType,
     latestVersion: () => getLatestNpmPackageVersion(UR_AGENT_PACKAGE_NAME)
@@ -823664,7 +823760,7 @@ ${customInstructions}` : customInstructions;
       }
     }
     logForDiagnosticsNoPII("info", "started", {
-      version: "1.81.8",
+      version: "1.81.10",
       is_native_binary: isInBundledMode()
     });
     registerCleanup(async () => {
@@ -824451,7 +824547,7 @@ Usage: ur --remote "your task description"`, () => gracefulShutdown(1));
         pendingHookMessages
       }, renderAndRun);
     }
-  }).version("1.81.8 (UR-Nexus)", "-v, --version", "Output the version number");
+  }).version("1.81.10 (UR-Nexus)", "-v, --version", "Output the version number");
   program2.option("-w, --worktree [name]", "Create a new git worktree for this session (optionally specify a name)");
   program2.option("--tmux", "Create a tmux session for the worktree (requires --worktree). Uses iTerm2 native panes when available; use --tmux=classic for traditional tmux.");
   if (canUserConfigureAdvisor()) {
@@ -825578,7 +825674,7 @@ if (false) {}
 async function main2() {
   const args = process.argv.slice(2);
   if (args.length === 1 && (args[0] === "--version" || args[0] === "-v" || args[0] === "-V")) {
-    console.log(`${"1.81.8"} (UR-Nexus)`);
+    console.log(`${"1.81.10"} (UR-Nexus)`);
     return;
   }
   if (args[0] === "a2a" && args[1] === "serve" && !args.includes("--help") && !args.includes("-h")) {

package/docs/VALIDATION.md

@@ -19,7 +19,7 @@ You need:
 
 ```sh
 ur --version
-# expected for this release: "1.81.8 (UR-Nexus)"
+# expected for this release: "1.81.10 (UR-Nexus)"
 ```
 
 ### 0.0 Redteam mode and Reverse Skills (1.81.0)

package/documentation/index.html

@@ -45,7 +45,7 @@
     <main id="content" class="content">
       <header class="topbar">
         <div>
-          <p class="eyebrow">Version 1.81.8</p>
+          <p class="eyebrow">Version 1.81.10</p>
           <h1>UR-Nexus Documentation</h1>
           <p class="lead">A practical, tutorial-style reference for installing, configuring, automating, extending, and operating UR-Nexus.</p>
         </div>

package/extensions/jetbrains-ur/build.gradle.kts

@@ -7,7 +7,7 @@ plugins {
 }
 
 group = "dev.urnexus"
-version = "1.81.8"
+version = "1.81.10"
 
 repositories {
     mavenCentral()

package/extensions/vscode-ur-inline-diffs/package.json

@@ -2,7 +2,7 @@
   "name": "ur-inline-diffs",
   "displayName": "UR Inline Diffs",
   "description": "Review, apply, and reject UR inline diff bundles from .ur/ide/diffs inside VS Code.",
-  "version": "1.81.8",
+  "version": "1.81.10",
   "publisher": "ur-nexus",
   "engines": {
     "vscode": "^1.92.0"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ur-agent",
-  "version": "1.81.8",
+  "version": "1.81.10",
   "description": "UR-Nexus — autonomous engineering workflow engine (plan, execute, test, verify, document, benchmark, reproduce)",
   "type": "module",
   "packageManager": "bun@1.3.14",