@bman654/clodex 2.10.0 → 2.11.1

package/README.md

@@ -241,11 +241,13 @@ Manage favorite models (max 20) and short aliases. Favorites feed the endpoint-m
 
 #### Context stops and the pricing boundary
 
-A context window is a cost dial as much as a capacity number. OpenAI prices GPT-5.6
-prompts above **272,000 input tokens at 2x input and 1.5x output for the full
-request**, which is why the Codex catalog reports a 272,000 window rather than the
-model's ceiling. Clodex follows that: the default `standard` stop stays under the
-line, and a larger window is something you ask for.
+A context window is a cost dial as much as a capacity number. OpenAI prices GPT-5.5
+and later prompts above **272,000 input tokens at 2x input and 1.5x output for the
+full request**, which is why the Codex catalog reports a 272,000 window rather than
+the model's ceiling. Newer families inherit the same boundary, so a model released
+after this was written is covered without a clodex update. Clodex follows that: the
+default `standard` stop stays under the line, and a larger window is something you
+ask for.
 
 ```sh
 clodex models --context sol=max --save     # this model's default, with a cost warning
@@ -253,19 +255,21 @@ clodex claude --context sol=max            # this launch only, nothing saved
 clodex models --context sol=default --save # back to the provider's tuned window
 ```
 
-Each stop is reported with the numbers behind it: the raw window, the headroom
-percentage the Codex catalog uses, the effective window a client should fill, and the
-account ceiling a larger stop can reach. A stop above the ceiling is clamped and says
-so. When a request's own reported token count crosses the boundary, clodex warns once
-per model for the life of the process, because the client's token count and the
-provider's differ after translation and only the provider's settles it.
+Each stop is reported with the numbers behind it: the raw window, the effective
+window a client should fill, and the account ceiling a larger stop can reach. A stop
+above the ceiling is clamped and says so. When a request's own reported token count
+crosses the boundary, clodex warns once per model for the life of the process,
+because the client's token count and the provider's differ after translation and only
+the provider's settles it.
 
 Two things worth knowing about the numbers:
 
-- **ChatGPT/Codex OAuth models carry a 95% headroom convention**, matching the Codex
-  client. Their reported window is 5% below the raw catalog value: `gpt-5.6-sol`
-  reports 258,400 rather than 272,000. This applies to that provider only; API-key
-  and OpenCode Go models keep their full window.
+- **Clodex reports the window the provider actually gives, and holds nothing back.**
+  Deciding how much of a window to leave free is the client's job — Claude Code
+  already reserves a fixed amount below whatever window it is told, and shrinking the
+  number first only costs usable context. A provider that declares a share of its own
+  is still honoured; clodex just never invents one. Use `--context` if you want a
+  smaller window than the provider offers.
 - **The account ceiling moves.** It is server-side and per-account, and it has
   changed by more than 2x within a single day in the past. `max` reads whatever the
   catalog reports now and clamps to it, so a stale ceiling shrinks the stop rather

package/dist/cli.js

@@ -382,7 +382,7 @@ import { join } from "path";
 // package.json
 var package_default = {
   name: "@bman654/clodex",
-  version: "2.10.0",
+  version: "2.11.1",
   publishConfig: {
     access: "public"
   },
@@ -469,7 +469,7 @@ var package_default = {
 
 // src/constants.ts
 var CODEX_RESPONSES_LITE_WS_URL = "wss://chatgpt.com/backend-api/codex/responses";
-var CODEX_RESPONSES_LITE_VERSION = "0.144.1";
+var CODEX_RESPONSES_LITE_VERSION = "0.153.3";
 var CODEX_RESPONSES_WEBSOCKETS_BETA = "responses_websockets=2026-02-06";
 var TEST_TIMEOUT_MS = 1e4;
 var CONFLICTING_ENV_VARS = [
@@ -3121,7 +3121,6 @@ async function deleteProviderCredential(authRef, diag) {
 }
 
 // src/context-modes.ts
-var DEFAULT_EFFECTIVE_CONTEXT_PERCENT = 95;
 function positiveInteger(value) {
   return typeof value === "number" && Number.isSafeInteger(value) && value > 0 ? value : void 0;
 }
@@ -5873,6 +5872,18 @@ var CHATGPT_CODEX_UNSUPPORTED_MODELS = /* @__PURE__ */ new Set([
   // confirmed: rejected by chatgpt.com/backend-api/codex
 ]);
 var OPENAI_OAUTH_MODEL_SEEDS = [
+  // GPT-6 family. The window and ceiling are what the live Codex catalog returned on
+  // 2026-09-04 and are deliberately NOT the published API numbers: the model card
+  // lists a 1,050,000 context window, but the Codex client is served a smaller one,
+  // and this path is Codex-only. Output limit and the pricing band come from the
+  // card (https://developers.openai.com/api/docs/models/gpt-6-astra), which states
+  // "Prompts with more than 272K input tokens are priced at 2x input and cache rates
+  // and 1.5x output for the full request" — the same boundary the GPT-5.6 family has.
+  { id: "gpt-6-astra", name: "GPT-6 Astra", contextWindow: 272e3, maxContextWindow: 872e3, maxOutputTokens: 128e3, reasoning: true, useResponsesLite: true, preferWebSockets: true },
+  // "An alias for our flagship general-purpose models, with safeguards calibrated
+  // for defensive cybersecurity work" — access is gated on a separate opt-in
+  // program, so most installs will never see this id in their catalog.
+  { id: "gpt-daybreak-blue-latest", name: "GPT Daybreak Blue", contextWindow: 272e3, maxContextWindow: 872e3, maxOutputTokens: 128e3, reasoning: true, useResponsesLite: true, preferWebSockets: true },
   // GPT-5.6 family (Sol / Terra / Luna)
   { id: "gpt-5.6-sol", name: "GPT-5.6 Sol", contextWindow: 272e3, maxContextWindow: 872e3, maxOutputTokens: 128e3, reasoning: true },
   { id: "gpt-5.6-terra", name: "GPT-5.6 Terra", contextWindow: 272e3, maxContextWindow: 872e3, maxOutputTokens: 128e3, reasoning: true },
@@ -5891,14 +5902,26 @@ var OPENAI_OAUTH_MODEL_SEEDS = [
   { id: "o1", name: "o1", reasoning: true },
   { id: "o1-mini", name: "o1 Mini", reasoning: true }
 ];
-var PRICING_BOUNDARY_FAMILIES = /^gpt-5\.[56]/;
+function hasPricingBoundary(id) {
+  const version = /^gpt-(\d+)(?:\.(\d+))?(?:-|$)/i.exec(id);
+  if (version) {
+    const major = Number(version[1]);
+    const minor = version[2] ? Number(version[2]) : 0;
+    if (major > 5 || major === 5 && minor >= 5) return true;
+  }
+  return /^gpt-daybreak(?:-|$)/i.test(id);
+}
 function openAiPricingMetadata(id) {
-  if (!PRICING_BOUNDARY_FAMILIES.test(id)) return {};
+  if (!hasPricingBoundary(id)) return {};
   return {
     pricingBoundary: GPT_5_6_PRICING_BOUNDARY,
     pricingBoundaryNote: GPT_5_6_PRICING_NOTE
   };
 }
+var LEGACY_IMPOSED_CONTEXT_PERCENT = 95;
+function migratedEffectiveContextPercent(cached) {
+  return cached === LEGACY_IMPOSED_CONTEXT_PERCENT ? void 0 : cached;
+}
 function applyOAuthSeedContextMetadata(models) {
   const seedById = new Map(buildOpenAiOAuthModels().map((model) => [model.id, model]));
   return models.map((model) => {
@@ -5907,10 +5930,11 @@ function applyOAuthSeedContextMetadata(models) {
     return {
       ...model,
       maxContextWindow: model.maxContextWindow ?? seed?.maxContextWindow,
-      effectiveContextPercent: model.effectiveContextPercent ?? seed?.effectiveContextPercent ?? DEFAULT_EFFECTIVE_CONTEXT_PERCENT,
+      effectiveContextPercent: migratedEffectiveContextPercent(model.effectiveContextPercent),
       pricingBoundary: model.pricingBoundary ?? seed?.pricingBoundary ?? pricing.pricingBoundary,
       pricingBoundaryNote: model.pricingBoundaryNote ?? seed?.pricingBoundaryNote ?? pricing.pricingBoundaryNote,
-      maxOutputTokens: model.maxOutputTokens ?? seed?.maxOutputTokens
+      maxOutputTokens: model.maxOutputTokens ?? seed?.maxOutputTokens,
+      reasoning: seed?.reasoning ?? model.reasoning
     };
   });
 }
@@ -5926,7 +5950,7 @@ function buildOpenAiOAuthModels() {
       brand: deriveBrand(prefix),
       contextWindow: resolveContextWindow(seed.id, seed.contextWindow),
       maxContextWindow: seed.maxContextWindow,
-      effectiveContextPercent: seed.effectiveContextPercent ?? DEFAULT_EFFECTIVE_CONTEXT_PERCENT,
+      effectiveContextPercent: seed.effectiveContextPercent,
       pricingBoundary: seed.pricingBoundary ?? pricing.pricingBoundary,
       pricingBoundaryNote: seed.pricingBoundaryNote ?? pricing.pricingBoundaryNote,
       maxOutputTokens: seed.maxOutputTokens,
@@ -10253,7 +10277,7 @@ async function createLanguageModel(spec) {
 }
 var ANTHROPIC_EFFORT_LEVELS = ["low", "medium", "high"];
 var OPENAI_EFFORT_LEVELS = ["low", "medium", "high", "xhigh"];
-var GPT_56_EFFORT_LEVELS = ["none", "low", "medium", "high", "xhigh", "max"];
+var CODEX_EXTENDED_EFFORT_LEVELS = ["none", "low", "medium", "high", "xhigh", "max"];
 var GEMINI_EFFORT_LEVELS = ["low", "medium", "high"];
 var MISTRAL_EFFORT_LEVELS = ["high", "off"];
 var XAI_EFFORT_LEVELS = ["none", "low", "medium", "high"];
@@ -10417,14 +10441,32 @@ function mapCodexEffortToAnthropic(effort) {
       return void 0;
   }
 }
-function isGpt56Model(modelId) {
-  return /^gpt-5\.6(?:-|$)/i.test(modelId);
+function isCodexReasoningFamily(modelId) {
+  return !isChatVariant(modelId) && supportsExtendedCodexEffort(modelId);
+}
+function isChatVariant(modelId) {
+  return /-chat(?:-|$)/i.test(modelId);
+}
+function supportsExtendedCodexEffort(modelId) {
+  const version = /^gpt-(\d+)(?:\.(\d+))?(?:-|$)/i.exec(modelId);
+  if (version) {
+    const major = Number(version[1]);
+    const minor = version[2] ? Number(version[2]) : 0;
+    if (major > 5 || major === 5 && minor >= 6) return true;
+  }
+  return /^gpt-daybreak(?:-|$)/i.test(modelId);
+}
+function supportsNoneEffort(modelId) {
+  return /^gpt-5\.6(?:-|$)/i.test(modelId) || /^gpt-daybreak-blue(?:-|$)/i.test(modelId);
 }
 function isReasoningSummaryUnsupportedModel(modelId) {
   return /codex-spark(?:-|$)/i.test(modelId);
 }
 function mapCodexEffortToOpenAI(effort, modelId) {
-  if (modelId && isGpt56Model(modelId) && GPT_56_EFFORT_LEVELS.includes(effort)) {
+  if (effort === "none") {
+    return modelId && supportsNoneEffort(modelId) ? "none" : void 0;
+  }
+  if (modelId && supportsExtendedCodexEffort(modelId) && CODEX_EXTENDED_EFFORT_LEVELS.includes(effort)) {
     return effort;
   }
   if (effort === "xhigh") return "high";
@@ -10565,9 +10607,9 @@ function getReasoningCapabilities(npm, modelId, metadata) {
   }
   if (npm === "@ai-sdk/openai" || npm === "@ai-sdk/azure") {
     const prefersResponses = modelPrefersResponsesApi(modelId);
-    if (prefersResponses || metadata?.reasoning) {
+    if (prefersResponses || isCodexReasoningFamily(modelId) || metadata?.reasoning) {
       return {
-        levels: isGpt56Model(modelId) ? [...GPT_56_EFFORT_LEVELS] : [...OPENAI_EFFORT_LEVELS],
+        levels: supportsExtendedCodexEffort(modelId) ? CODEX_EXTENDED_EFFORT_LEVELS.filter((level) => level !== "none" || supportsNoneEffort(modelId)) : [...OPENAI_EFFORT_LEVELS],
         defaultLevel: "medium",
         supportsSummaries: true,
         mode: "controllable",
@@ -10731,10 +10773,14 @@ function effortProviderOptions(npm, effort, modelId, metadata) {
     return mapped ? { openrouter: { reasoning: { effort: mapped, exclude: false } } } : void 0;
   }
   if (npm === "@ai-sdk/openai" || npm === "@ai-sdk/azure") {
-    if (!modelId || !modelPrefersResponsesApi(modelId)) return void 0;
+    if (!modelId || isChatVariant(modelId)) return void 0;
+    if (!(modelPrefersResponsesApi(modelId) || isCodexReasoningFamily(modelId))) {
+      return void 0;
+    }
     const reasoningEffort = mapCodexEffortToOpenAI(effort, modelId);
     if (!reasoningEffort) return void 0;
-    return isReasoningSummaryUnsupportedModel(modelId) ? { openai: { reasoningEffort, reasoningSummary: null } } : { openai: { reasoningEffort } };
+    const openaiOptions = { reasoningEffort, forceReasoning: true };
+    return isReasoningSummaryUnsupportedModel(modelId) ? { openai: { ...openaiOptions, reasoningSummary: null } } : { openai: openaiOptions };
   }
   if (npm === "@ai-sdk/xai") {
     if (!modelId || !isXaiReasoningEffortModel(modelId)) return void 0;
@@ -15021,7 +15067,7 @@ function parseOpenAiModelEntries(body) {
   }
   return [];
 }
-function buildDynamicOAuthModel(entry, seedById) {
+function buildDynamicOAuthModel(entry, seedById, codexCatalog) {
   const seed = seedById.get(entry.id);
   if (seed) {
     return {
@@ -15047,12 +15093,27 @@ function buildDynamicOAuthModel(entry, seedById) {
     brand: deriveBrand(prefix),
     contextWindow: entry.context_window ?? resolveContextWindow(id),
     maxContextWindow: entry.max_context_window,
-    effectiveContextPercent: entry.effective_context_window_percent ?? DEFAULT_EFFECTIVE_CONTEXT_PERCENT,
+    // Absent means no reduction. clodex reports the window the provider actually
+    // gives; deciding how much of it to leave free is the client's job, and Claude
+    // Code already reserves a flat 33,000 tokens below whatever it is told.
+    effectiveContextPercent: entry.effective_context_window_percent,
     maxOutputTokens: entry.max_output_tokens,
     ...openAiPricingMetadata(id),
     modelFormat: "openai",
     npm: "@ai-sdk/openai",
-    reasoning: modelPrefersResponsesApi(id),
+    // Assume a model from the Codex listing reasons. That endpoint reports no
+    // reasoning field of its own, so the old `modelPrefersResponsesApi(id)` was an
+    // id-pattern GUESS that silently said "no" to every family it had not been
+    // taught yet — gpt-6-astra and gpt-daybreak-blue-latest both landed as
+    // non-reasoning that way, which dropped the user's chosen effort and removed
+    // the effort selector from the patched binary (getPatchReasoningCapabilities
+    // early-returns on a `false`). Verified against all 11 models in the live
+    // catalog on 2026-09-04.
+    //
+    // This only decides what the effort UI offers. It is NOT on its own enough to
+    // put reasoning.effort on the wire — effortProviderOptions admits by family —
+    // so a wrong `true` here costs an unusable menu entry, not a 400.
+    reasoning: codexCatalog ? true : modelPrefersResponsesApi(id),
     useResponsesLite: entry.useResponsesLite,
     preferWebSockets: entry.preferWebSockets
   };
@@ -15081,7 +15142,7 @@ async function fetchJsonWithAuth(url, accessToken, timeoutMs) {
 async function refreshOpenAiOAuthModels(accessToken) {
   const TIMEOUT_MS = 1e4;
   const seedById = new Map(buildOpenAiOAuthModels().map((m) => [m.id, m]));
-  const toModels = (entries) => entries.map((entry) => buildDynamicOAuthModel(entry, seedById));
+  const toModels = (entries, codexCatalog) => entries.map((entry) => buildDynamicOAuthModel(entry, seedById, codexCatalog));
   const claudeVersion = getInstalledClaudeVersion();
   const codexResult = await fetchJsonWithAuth(
     `https://chatgpt.com/backend-api/codex/models?client_version=${claudeVersion}`,
@@ -15090,7 +15151,7 @@ async function refreshOpenAiOAuthModels(accessToken) {
   );
   const codexEntries = parseOpenAiModelEntries(codexResult.body);
   if (codexEntries.length > 0) {
-    return { models: toModels(codexEntries), source: "live" };
+    return { models: toModels(codexEntries, true), source: "live" };
   }
   const chatGptResult = await fetchJsonWithAuth(
     "https://chatgpt.com/backend-api/models",
@@ -15099,7 +15160,7 @@ async function refreshOpenAiOAuthModels(accessToken) {
   );
   const chatGptEntries = parseOpenAiModelEntries(chatGptResult.body).filter(({ id }) => !CHATGPT_CODEX_UNSUPPORTED_MODELS.has(id));
   if (chatGptEntries.length > 0) {
-    return { models: toModels(chatGptEntries), source: "live" };
+    return { models: toModels(chatGptEntries, false), source: "live" };
   }
   const failures = [codexResult.error, chatGptResult.error].filter((error) => error !== void 0);
   const credentialFailure = failures.find((error) => /(?:\brejected\b|\b401\b|\b403\b)/i.test(error));