@bitkyc08/opencodex 2.62.0-preview.20260923 → 2.63.0

package/src/oauth/devin.ts

@@ -19,7 +19,7 @@ import { DEFAULT_REGION, type WindsurfRegion } from "./devin/types";
 import { registerUser } from "./devin/register-user";
 import { DEVIN_DEFAULT_API_SERVER, resolveDevinApiBaseUrl, validateDevinApiBaseUrl } from "./devin/api-base";
 import { readDevinCliCredentialOutcome } from "./devin/cli-import";
-import { getCredential } from "./store";
+import { getCredential, listAccounts } from "./store";
 import { DEPRECATED_OAUTH_PROVIDER_ALIASES } from "./index";
 
 export { DEVIN_DEFAULT_API_SERVER } from "./devin/api-base";
@@ -50,46 +50,52 @@ function devinAliasCredentialSlots(providerId: string): string[] {
  * configured provider baseUrl is the fallback, and the US default is the last
  * resort; both are re-validated because neither is trusted more than the
  * network value.
+ *
+ * A stored tenant host is used only for the account that owns `apiKey`, the key
+ * this request will transmit. Devin keeps several accounts per provider id and
+ * the request path injects the admitted account's token, which need not be the
+ * active one, while a provider-configured key, a forwarded bearer, or a test
+ * token is not stored at all. Reading the active slot's host for any of those
+ * would send one account's key to another account's EU or FedStart tenant.
  */
-export function resolveDevinApiServer(configuredBaseUrl?: string, providerId = "devin"): string {
-  // Provider-scoped, keyed by the configured provider id verbatim and consulted
-  // FIRST. `devin-cli` is a deprecated alias for `devin`, but an unmigrated
-  // config row still owns its old credential slot until the startup migration
-  // rekeys the row and the slot together — normalizing the id here would read
-  // the wrong slot for that window. An EU or FedStart tenant is recorded on the
-  // credential rather than in the registry, so a fixed "devin" slot would send
-  // the key to the wrong host either way.
-  const literalCredential = getCredential(providerId);
-  const literal = validateDevinApiBaseUrl(literalCredential?.apiBaseUrl);
-  if (literal !== undefined) return literal;
+export function resolveDevinApiServer(configuredBaseUrl?: string, providerId = "devin", apiKey?: string): string {
+  const owner = apiKey ? findDevinCredentialOwner(providerId, apiKey) : undefined;
+  if (owner !== undefined) {
+    // The owning account decides. An owner whose recorded host is missing or
+    // off-allowlist falls through to the configured base URL rather than
+    // borrowing another slot's tenant: the same key stored twice is the rekey
+    // window, and a second slot's host is not more trustworthy than this one.
+    const host = validateDevinApiBaseUrl(owner.apiBaseUrl);
+    if (host !== undefined) return host;
+  }
+  return validateDevinApiBaseUrl(configuredBaseUrl) ?? DEVIN_DEFAULT_API_SERVER;
+}
 
-  // The startup merge saves providers["devin"] synchronously but fires the
-  // credential rekey detached — runDevinProviderMergeStartupMigration cannot
-  // await inside the synchronous startServer window — so the row can already
-  // say "devin" while the credential still sits in the "devin-cli" slot, and it
-  // stays that way for the whole process when the rekey fails or refuses on an
-  // occupied destination slot. Reading the alias-linked slots in both
-  // directions closes that window: "devin" finds the not-yet-rekeyed
-  // "devin-cli" credential, and a lingering "devin-cli" row finds a credential
-  // already rekeyed to "devin". Every candidate passes the same allowlist — an
-  // alias slot is not trusted more than the literal one.
-  // Only when this id owns no credential at all. A present credential whose
-  // apiBaseUrl is missing or off-allowlist is a different situation: the rekey
-  // refuses an occupied destination slot, so both ids can hold credentials that
-  // belong to two different accounts. Borrowing a tenant across that pair would
-  // send this account's key to the other account's EU or FedStart host, which
-  // is the exact misdirection the provider-scoped lookup exists to prevent. An
-  // unusable host on a credential that does exist falls through to the
-  // configured base URL and then the default, as it did before this window was
-  // closed.
-  if (literalCredential === null || literalCredential === undefined) {
-    for (const slot of devinAliasCredentialSlots(providerId)) {
-      const host = validateDevinApiBaseUrl(getCredential(slot)?.apiBaseUrl);
-      if (host !== undefined) return host;
+/**
+ * The stored credential whose access token is exactly `apiKey`.
+ *
+ * The configured provider id is searched first and verbatim: `devin-cli` is a
+ * deprecated alias for `devin`, but an unmigrated config row still owns its old
+ * slot until the startup migration rekeys the row and the slot together. The
+ * startup merge saves providers["devin"] synchronously but fires the credential
+ * rekey detached (runDevinProviderMergeStartupMigration cannot await inside the
+ * synchronous startServer window), so the alias-linked slots are searched next,
+ * in both directions. Within a slot the active account is read first, then the
+ * rest, because the admitted account can be any of them.
+ *
+ * A key refreshed between token resolution and this lookup would match nothing
+ * and use the configured host for that turn. Devin keys are long-lived
+ * RegisterUser API keys, so this window is not a routine refresh race.
+ */
+function findDevinCredentialOwner(providerId: string, apiKey: string): OAuthCredentials | undefined {
+  for (const slot of [providerId, ...devinAliasCredentialSlots(providerId)]) {
+    const active = getCredential(slot);
+    if (active?.access === apiKey) return active;
+    for (const account of listAccounts(slot)) {
+      if (account.credential.access === apiKey) return account.credential;
     }
   }
-
-  return validateDevinApiBaseUrl(configuredBaseUrl) ?? DEVIN_DEFAULT_API_SERVER;
+  return undefined;
 }
 
 function decodeJwtPayload(token: string): Record<string, unknown> | undefined {

package/src/providers/codebuddy-models.ts

@@ -21,6 +21,9 @@ export const CODEBUDDY_GLOBAL_MODELS = [
   "gpt-5.6-sol",
   "gpt-5.6-terra",
   "gpt-5.6-luna",
+  // 260923 preemptive: GPT-6 Sol and Luna (OpenAI announced 2026-09-22) added ahead of this provider's own catalog; mirrors the GPT-5.6 Sol/Luna rows.
+  "gpt-6-sol",
+  "gpt-6-luna",
   "gpt-5.5",
   "gpt-5.4",
   "gpt-5.3-codex",
@@ -70,6 +73,8 @@ export const CODEBUDDY_GLOBAL_MODEL_CONTEXT_WINDOWS: Record<string, number> = {
   "gpt-5.6-sol": 1_000_000,
   "gpt-5.6-terra": 1_000_000,
   "gpt-5.6-luna": 1_000_000,
+  "gpt-6-sol": 1_000_000,
+  "gpt-6-luna": 1_000_000,
   "gpt-5.5": 1_000_000,
   "gpt-5.4": 272_000,
   "gpt-5.3-codex": 272_000,
@@ -90,6 +95,8 @@ export const CODEBUDDY_GLOBAL_MODEL_MAX_OUTPUT_TOKENS: Record<string, number> =
   "gpt-5.6-sol": 128_000,
   "gpt-5.6-terra": 128_000,
   "gpt-5.6-luna": 128_000,
+  "gpt-6-sol": 128_000,
+  "gpt-6-luna": 128_000,
   "gpt-5.5": 72_000,
   "gpt-5.4": 128_000,
   "gpt-5.3-codex": 128_000,
@@ -106,6 +113,8 @@ export const CODEBUDDY_GLOBAL_MODEL_REASONING_EFFORTS: Record<string, string[]>
   "gpt-5.6-sol": ["low", "medium", "high", "xhigh"],
   "gpt-5.6-terra": ["low", "medium", "high", "xhigh"],
   "gpt-5.6-luna": ["low", "medium", "high", "xhigh"],
+  "gpt-6-sol": ["low", "medium", "high", "xhigh"],
+  "gpt-6-luna": ["low", "medium", "high", "xhigh"],
   "glm-5.3": ["low", "high", "max"],
   "glm-5.2": ["high", "xhigh"],
 };
@@ -114,6 +123,8 @@ export const CODEBUDDY_GLOBAL_MODEL_DEFAULT_REASONING_EFFORTS: Record<string, st
   "gpt-5.6-sol": "high",
   "gpt-5.6-terra": "high",
   "gpt-5.6-luna": "high",
+  "gpt-6-sol": "high",
+  "gpt-6-luna": "high",
   "glm-5.3": "high",
   "glm-5.2": "high",
 };

package/src/providers/kiro-models.ts

@@ -4,7 +4,13 @@ export const KIRO_MODELS = [
   "gpt-5.6-sol",
   "gpt-5.6-terra",
   "gpt-5.6-luna",
+  // 260923 preemptive: GPT-6 Sol and Luna (OpenAI announced 2026-09-22) added ahead of this provider's own catalog; mirrors the GPT-5.6 Sol/Luna rows. Calls fail upstream until Kiro ships the models.
+  "gpt-6-sol",
+  "gpt-6-luna",
   "claude-sonnet-5",
+  // 260923 preemptive: Claude Opus 5.5 added ahead of Kiro's catalog (kiro.dev did not list it on
+  // 2026-09-23). Mirrors claude-opus-5; calls fail upstream until Kiro ships the model.
+  "claude-opus-5.5",
   "claude-opus-5",
   "claude-opus-4.8",
   "claude-opus-4.7",
@@ -28,7 +34,10 @@ export const KIRO_MODEL_CONTEXT_WINDOWS: Record<string, number> = {
   "gpt-5.6-sol": 272_000,
   "gpt-5.6-terra": 272_000,
   "gpt-5.6-luna": 272_000,
+  "gpt-6-sol": 272_000,
+  "gpt-6-luna": 272_000,
   "claude-sonnet-5": 1_000_000,
+  "claude-opus-5.5": 1_000_000,
   "claude-opus-5": 1_000_000,
   "claude-opus-4.8": 1_000_000,
   "claude-opus-4.7": 1_000_000,

package/src/providers/quota/vendor-probes-oauth.ts

@@ -224,6 +224,8 @@ function parseClaudeBucket(value: unknown): { percent?: number; resetAt?: number
   return { percent, resetAt };
 }
 
+const TERMINAL_CONTROL_CHARACTERS = /[\u0000-\u001f\u007f-\u009f]/gu;
+
 function parseClaudeLimit(value: unknown): { label: string; percent: number; resetAt?: number } | null {
   const rec = asRecord(value);
   if (!rec) return null;
@@ -231,13 +233,18 @@ function parseClaudeLimit(value: unknown): { label: string; percent: number; res
   if (percent === undefined) return null;
   const scope = asRecord(rec.scope);
   const model = asRecord(scope?.model);
-  const rawLabel = String(model?.display_name ?? "").trim();
+  const rawLabel = String(model?.display_name ?? "")
+    .replace(TERMINAL_CONTROL_CHARACTERS, "")
+    .trim();
   if (!rawLabel) return null;
   const lowerLabel = rawLabel.toLowerCase();
   const label = lowerLabel.includes("fable") ? "Fable"
     : lowerLabel.includes("opus") ? "Opus"
       : lowerLabel.includes("sonnet") ? "Sonnet"
-        : rawLabel;
+        : null;
+  // An unrecognized display_name is never published as a quota label: stripping
+  // control characters still leaves attacker-chosen residue on the quota line.
+  if (label === null) return null;
   const resetAt = normalizeResetAt(rec.resets_at);
   return { label, percent, ...(resetAt !== undefined ? { resetAt } : {}) };
 }

package/src/providers/registry/entries-core.ts

@@ -23,6 +23,7 @@ import {
   ZAI_GLM_52_REASONING_EFFORTS,
   ZAI_GLM_53_REASONING_EFFORTS,
   OPENAI_GPT56_MODELS,
+  OPENAI_GPT6_MODELS,
   OPENAI_GPT56_PRO_MODELS,
   OPENAI_API_GPT56_CONTEXT_WINDOWS,
   OPENAI_API_GPT56_MAX_INPUT_TOKENS,
@@ -166,7 +167,7 @@ export const PROVIDER_REGISTRY_CORE: readonly ProviderRegistryEntry[] = [
     // (it is the current catalog, so its default ordering wins), then the ids
     // only the old devin entry carried. Degraded-mode seed only either way —
     // `liveModels` discovers the account's real roster.
-    models: ["swe-2", "swe-1-7", "gpt-5-6-sol", "gpt-6-astra", "claude-opus-5", "claude-fable-5-1", "claude-sonnet-5", "glm-5-3", "kimi-k3", "gemini-3-8-flash", "grok-4-6", "swe-1-7-lightning", "gpt-5-6-luna", "gpt-5-6-terra", "claude-opus-4-8", "glm-5-2", "kimi-k2-7", "grok-4-5"],
+    models: ["swe-2", "swe-1-7", "gpt-5-6-sol", "gpt-6-astra", "gpt-6-sol", "gpt-6-luna", "claude-opus-5-5", "claude-opus-5", "claude-fable-5-1", "claude-sonnet-5", "glm-5-3", "kimi-k3", "gemini-3-8-flash", "grok-4-6", "swe-1-7-lightning", "gpt-5-6-luna", "gpt-5-6-terra", "claude-opus-4-8", "glm-5-2", "kimi-k2-7", "grok-4-5"],
     liveModels: true,
     defaultModel: "swe-2",
     modelContextWindows: DEVIN_MODEL_CONTEXT_WINDOWS,
@@ -535,13 +536,19 @@ export const PROVIDER_REGISTRY_CORE: readonly ProviderRegistryEntry[] = [
     featured: true,
     dashboardUrl: "https://platform.openai.com/api-keys",
     defaultModel: "gpt-5.5",
-    models: ["gpt-5.5", ...OPENAI_GPT56_MODELS, ...OPENAI_GPT56_PRO_MODELS, ...OPENAI_DAYBREAK_MODELS, "gpt-6-astra"],
+    models: ["gpt-5.5", ...OPENAI_GPT56_MODELS, ...OPENAI_GPT56_PRO_MODELS, ...OPENAI_DAYBREAK_MODELS, "gpt-6-astra", ...OPENAI_GPT6_MODELS],
     liveModels: true,
-    modelContextWindows: { ...OPENAI_API_GPT56_CONTEXT_WINDOWS, ...OPENAI_DAYBREAK_CONTEXT_WINDOWS, "gpt-6-astra": 1_050_000 },
-    modelMaxInputTokens: { ...OPENAI_API_GPT56_MAX_INPUT_TOKENS, ...OPENAI_DAYBREAK_MAX_INPUT_TOKENS, "gpt-6-astra": 922_000 },
-    modelMaxOutputTokens: { "gpt-6-astra": 128_000 },
+    modelContextWindows: {
+      ...OPENAI_API_GPT56_CONTEXT_WINDOWS, ...OPENAI_DAYBREAK_CONTEXT_WINDOWS, "gpt-6-astra": 1_050_000,
+      ...Object.fromEntries(OPENAI_GPT6_MODELS.map(id => [id, 1_050_000])),
+    },
+    modelMaxInputTokens: {
+      ...OPENAI_API_GPT56_MAX_INPUT_TOKENS, ...OPENAI_DAYBREAK_MAX_INPUT_TOKENS, "gpt-6-astra": 922_000,
+      ...Object.fromEntries(OPENAI_GPT6_MODELS.map(id => [id, 922_000])),
+    },
+    modelMaxOutputTokens: { "gpt-6-astra": 128_000, ...Object.fromEntries(OPENAI_GPT6_MODELS.map(id => [id, 128_000])) },
     modelInputModalities: Object.fromEntries(
-      ["gpt-5.5", ...OPENAI_GPT56_MODELS, ...OPENAI_GPT56_PRO_MODELS, ...OPENAI_DAYBREAK_MODELS, "gpt-6-astra"]
+      ["gpt-5.5", ...OPENAI_GPT56_MODELS, ...OPENAI_GPT56_PRO_MODELS, ...OPENAI_DAYBREAK_MODELS, "gpt-6-astra", ...OPENAI_GPT6_MODELS]
         .map(id => [id, ["text", "image"]]),
     ),
     modelReasoningEfforts: {
@@ -550,6 +557,7 @@ export const PROVIDER_REGISTRY_CORE: readonly ProviderRegistryEntry[] = [
       ),
       ...OPENAI_DAYBREAK_REASONING_EFFORTS,
       "gpt-6-astra": ["low", "medium", "high", "xhigh", "max"],
+      ...Object.fromEntries(OPENAI_GPT6_MODELS.map(id => [id, ["low", "medium", "high", "xhigh", "max"]])),
     },
     virtualModels: OPENAI_API_GPT56_VIRTUAL_MODELS,
   },
@@ -870,7 +878,7 @@ export const PROVIDER_REGISTRY_CORE: readonly ProviderRegistryEntry[] = [
     featured: true,
     dashboardUrl: "https://openrouter.ai/keys",
     jawcodeBundle: "openrouter",
-    models: ["anthropic/claude-sonnet-5", ...OPENROUTER_GPT56_MODELS],
+    models: ["anthropic/claude-sonnet-5", ...OPENROUTER_GPT56_MODELS, ...OPENAI_GPT6_MODELS.map(id => `openai/${id}`)],
     modelContextWindows: {
       "anthropic/claude-sonnet-5": 1_000_000,
       ...OPENROUTER_GPT56_CONTEXT_WINDOWS,
@@ -883,6 +891,9 @@ export const PROVIDER_REGISTRY_CORE: readonly ProviderRegistryEntry[] = [
       "openai/gpt-5.6-sol": true,
       "openai/gpt-5.6-terra": true,
       "openai/gpt-5.6-luna": true,
+      // 260923 preemptive: GPT-6 Sol/Luna are OpenAI-backed routes like the GPT-5.6 rows above.
+      "openai/gpt-6-sol": true,
+      "openai/gpt-6-luna": true,
     },
     // Deliberately no OpenRouter route pin: it bills the endpoint actually used and reports the
     // actual service_tier. B0 confirmation therefore owns downgrade safety. Forcing `only` plus
@@ -985,7 +996,7 @@ export const PROVIDER_REGISTRY_CORE: readonly ProviderRegistryEntry[] = [
     id: "bizrouter", label: "BizRouter", adapter: "openai-chat", baseUrl: "https://api.bizrouter.ai/v1",
     authKind: "key", dashboardUrl: "https://bizrouter.ai/settings/keys",
     defaultModel: "openai/gpt-5.6-sol",
-    models: ["openai/gpt-5.6-sol", "anthropic/claude-sonnet-5", "google/gemini-3.5-flash"],
+    models: ["openai/gpt-5.6-sol", "openai/gpt-6-sol", "openai/gpt-6-luna", "anthropic/claude-sonnet-5", "google/gemini-3.5-flash"],
     note: "Korean enterprise LLM gateway. Per-key allowed models are discovered live from /v1/models. Full catalog: https://bizrouter.ai/models",
   },
   { id: "groq", label: "Groq", adapter: "openai-chat", baseUrl: "https://api.groq.com/openai/v1", authKind: "key", featured: true, dashboardUrl: "https://console.groq.com/keys" },

package/src/providers/registry/entries-extended.ts

@@ -1249,7 +1249,7 @@ export const PROVIDER_REGISTRY_EXTENDED: readonly ProviderRegistryEntry[] = [
     featured: false,
     dashboardUrl: "https://github.com/settings/copilot",
     liveModels: true,
-    models: ["gpt-4o", "gpt-4.1", "gpt-4.1-mini", "claude-sonnet-4", "gemini-2.5-pro", "gpt-5-mini", "gpt-5.3-codex", "gpt-5.4", "gpt-5.4-mini", "gpt-5.5", "gpt-5.6-luna", "gpt-5.6-sol", "gpt-5.6-terra"],
+    models: ["gpt-4o", "gpt-4.1", "gpt-4.1-mini", "claude-sonnet-4", "gemini-2.5-pro", "gpt-5-mini", "gpt-5.3-codex", "gpt-5.4", "gpt-5.4-mini", "gpt-5.5", "gpt-5.6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-6-sol", "gpt-6-luna"],
     defaultModel: "gpt-4o",
     // Copilot fronts a mixed-wire catalog: these models reject /chat/completions for
     // real Codex-agent traffic (function tools + reasoning), so every inbound wire
@@ -1266,6 +1266,9 @@ export const PROVIDER_REGISTRY_EXTENDED: readonly ProviderRegistryEntry[] = [
       "gpt-5.6-sol": "openai-responses",
       "gpt-5.6-terra": "openai-responses",
       "gpt-6-astra": "openai-responses",
+      // 260923 preemptive: GPT-6 Sol/Luna ride Responses like every GPT-5.6/6 row above.
+      "gpt-6-sol": "openai-responses",
+      "gpt-6-luna": "openai-responses",
       "grok-4.5": "openai-responses",
       "grok-4.6": "openai-responses",
       "mai-code-1.1-flash": "openai-responses",

package/src/providers/registry/model-seeds.ts

@@ -6,8 +6,11 @@ import type { ProviderModelDiscoverySpec } from "./types";
 // devlog/_plan/260710_provider_hardening/001_research_frontier.md.
 // 260902 Claude Fable 5.1 (`claude-fable-5-1`): 1M context / 128K output / adaptive thinking
 // always on, per the official models overview and pricing page (platform.claude.com).
-export const ANTHROPIC_MODELS = ["claude-fable-5-1", "claude-fable-5", "claude-sonnet-5", "claude-opus-5", "claude-opus-4-8", "claude-opus-4-7", "claude-opus-4-6", "claude-sonnet-4-6", "claude-haiku-4-5"];
-export const ANTHROPIC_MODEL_CONTEXT_WINDOWS: Record<string, number> = { "claude-fable-5-1": 1_000_000, "claude-sonnet-5": 1_000_000, "claude-fable-5": 1_000_000, "claude-opus-5": 1_000_000, "claude-opus-4-8": 1_000_000, "claude-opus-4-7": 1_000_000, "claude-opus-4-6": 1_000_000, "claude-sonnet-4-6": 1_000_000, "claude-haiku-4-5": 200_000 };
+// 260923 Claude Opus 5.5 (`claude-opus-5-5`, released 2026-09-22): 1M context / 128K output /
+// adaptive thinking always on / effort low..max with a medium default, per the Opus 5.5
+// overview, effort and pricing pages (platform.claude.com).
+export const ANTHROPIC_MODELS = ["claude-fable-5-1", "claude-fable-5", "claude-sonnet-5", "claude-opus-5-5", "claude-opus-5", "claude-opus-4-8", "claude-opus-4-7", "claude-opus-4-6", "claude-sonnet-4-6", "claude-haiku-4-5"];
+export const ANTHROPIC_MODEL_CONTEXT_WINDOWS: Record<string, number> = { "claude-fable-5-1": 1_000_000, "claude-sonnet-5": 1_000_000, "claude-fable-5": 1_000_000, "claude-opus-5-5": 1_000_000, "claude-opus-5": 1_000_000, "claude-opus-4-8": 1_000_000, "claude-opus-4-7": 1_000_000, "claude-opus-4-6": 1_000_000, "claude-sonnet-4-6": 1_000_000, "claude-haiku-4-5": 200_000 };
 // All seeded Claude models support vision: https://platform.claude.com/docs/en/models/overview
 export const ANTHROPIC_MODEL_INPUT_MODALITIES: Record<string, string[]> = Object.fromEntries(
   ANTHROPIC_MODELS.map(id => [id, ["text", "image"]]),
@@ -154,6 +157,12 @@ export const OPENAI_API_GPT56_VIRTUAL_MODELS: Record<string, { wireModelId: stri
   "gpt-5.6-luna-pro": { wireModelId: "gpt-5.6-luna", reasoningMode: "pro" },
 };
 export const OPENAI_API_GPT56_REASONING_EFFORTS = ["low", "medium", "high", "xhigh", "max"];
+/**
+ * GPT-6 Sol and Luna on the OpenAI API (released 2026-09-22,
+ * https://developers.openai.com/api/docs/changelog). Added 2026-09-23 ahead of live discovery; the
+ * API window is not published yet, so the rows mirror gpt-6-astra's 1,050,000 / 922,000 API seed.
+ */
+export const OPENAI_GPT6_MODELS = ["gpt-6-sol", "gpt-6-luna"];
 /*
  * Meta Model API (https://api.meta.ai/v1) — published ladder, deliberately NOT the
  * house set. dev.meta.ai/docs/reasoning lists "none", "minimal", "low", "medium",
@@ -238,6 +247,9 @@ export const OPENROUTER_GPT56_CONTEXT_WINDOWS = {
   "openai/gpt-5.6-sol": OPENROUTER_GPT56_CONTEXT_WINDOW,
   "openai/gpt-5.6-terra": OPENROUTER_GPT56_CONTEXT_WINDOW,
   "openai/gpt-5.6-luna": OPENROUTER_GPT56_CONTEXT_WINDOW,
+  // 260923 preemptive: GPT-6 Sol/Luna ahead of OpenRouter's own listing; same window as GPT-5.6.
+  "openai/gpt-6-sol": OPENROUTER_GPT56_CONTEXT_WINDOW,
+  "openai/gpt-6-luna": OPENROUTER_GPT56_CONTEXT_WINDOW,
 };
 
 /**
@@ -838,6 +850,9 @@ export const DIGITALOCEAN_CHAT_COMPLETION_MODELS = [
   "openai-gpt-5.6-sol",
   "openai-gpt-5.6-terra",
   "openai-gpt-5.6-luna",
+  // 260923 preemptive: GPT-6 Sol/Luna ahead of DigitalOcean's model list.
+  "openai-gpt-6-sol",
+  "openai-gpt-6-luna",
   "qwen3-coder-flash",
   "qwen3.5-397b-a17b",
   "deepseek-4-flash",
@@ -1003,8 +1018,11 @@ export const CLINE_PASS_MODEL_INPUT_MODALITIES: Record<string, string[]> = Objec
 // catalogue snapshot supplied by the original provider author
 // (https://api.opper.ai/v3/models?limit=2000, captured 2026-09-14); `vendor/model` ids
 // (anthropic/claude-sonnet-4-6) pin one route and stay valid, they are just not seeded.
+// 260923: `claude-opus-5-5` pool (anthropic, aws eu, vertex, vertex-eu members; all 1M / 128K,
+// vision) read from the same catalogue endpoint the day after Anthropic's release.
 export const OPPER_MODELS = [
   "claude-sonnet-4-6",
+  "claude-opus-5-5",
   "claude-opus-5",
   "gpt-5.5",
   "gpt-5.4-mini",
@@ -1017,6 +1035,7 @@ export const OPPER_MODELS = [
 // (kimi-k3 output); live discovery owns which models exist.
 export const OPPER_MODEL_CONTEXT_WINDOWS: Record<string, number> = {
   "claude-sonnet-4-6": 1_000_000,
+  "claude-opus-5-5": 1_000_000,
   "claude-opus-5": 1_000_000,
   "gpt-5.5": 1_050_000,
   "gpt-5.4-mini": 400_000,
@@ -1027,6 +1046,7 @@ export const OPPER_MODEL_CONTEXT_WINDOWS: Record<string, number> = {
 };
 export const OPPER_MODEL_MAX_OUTPUT_TOKENS: Record<string, number> = {
   "claude-sonnet-4-6": 64_000,
+  "claude-opus-5-5": 128_000,
   "claude-opus-5": 128_000,
   "gpt-5.5": 128_000,
   "gpt-5.4-mini": 128_000,

package/src/responses/bridge-search-replay-cache.ts

@@ -14,9 +14,9 @@
  * what `appendBridgeSearchTurn` would have written onto a continuation leg, so a replayed turn
  * and a continued turn show the destination the same conversation.
  *
- * Scope. Entries are keyed by the upstream destination in addition to the cell id. The cell id is
- * a v4 UUID minted here, so it cannot collide across conversations, but an unscoped key would let
- * a history replayed against a DIFFERENT provider resurrect a call that provider never made.
+ * Scope. Entries are keyed by the exact conversation and serving identity in addition to the cell
+ * id. The cell id is a v4 UUID minted here, but possession of a client-visible id is not authority
+ * to recover result text under another provider, model, destination, or credential.
  *
  * Bounds and privacy. Result text is web content the caller already received, but it is still
  * request-derived data: it lives in memory only, is never logged, serialized, or exported, and is
@@ -26,7 +26,7 @@
  * alone. Neither re-running the search nor inventing a result is an acceptable recovery.
  */
 
-import { reasoningReplayDestinationIdentity } from "./reasoning-replay-cache";
+import type { OcxReasoningReplayScopeRef } from "../types";
 
 const MAX_ENTRIES = 64;
 const MAX_TOTAL_BYTES = 512 * 1024;
@@ -58,14 +58,24 @@ let clockForTests: (() => number) | null = null;
 const now = (): number => clockForTests?.() ?? Date.now();
 
 /**
- * Identify the upstream destination a bridged search belongs to.
+ * Identify the exact conversation and upstream binding a bridged search belongs to.
  *
- * Reuses the salted process-local destination digest the reasoning replay cache already defines,
- * so both stores agree on what "the same upstream" means and neither invents a second notion of
- * destination identity.
+ * The serving route binds this holder only after provider, model, and physical credential
+ * selection. A missing conversation or binding fails closed: a cell id is client-visible and is
+ * not itself authority to recover another request's retained result.
  */
-export function bridgeSearchReplayScope(baseUrl: string | undefined): string | undefined {
-  return reasoningReplayDestinationIdentity(baseUrl);
+export function bridgeSearchReplayScope(scope: OcxReasoningReplayScopeRef | undefined): string | undefined {
+  const identity = scope?.current;
+  if (!scope?.clientPrincipalId || !scope.clientThreadId || !identity) return undefined;
+  return JSON.stringify([
+    scope.clientPrincipalId,
+    scope.clientThreadId,
+    identity.providerName,
+    identity.providerDestinationIdentity,
+    identity.adapterName,
+    identity.modelId,
+    identity.credentialIdentity,
+  ]);
 }
 
 function keyFor(scope: string, cellItemId: string): string {

package/src/responses/plaintext-v2-agent-messages.ts

@@ -94,7 +94,16 @@ function collaborationCatalogInfo(catalogs: readonly unknown[][]): {
 
 export function hasPlaintextV2CollaborationCatalog(body: unknown): boolean {
   if (!isPlainObject(body)) return false;
-  return Array.isArray(body.tools) && collaborationCatalogInfo([body.tools]).hasV2Catalog;
+  if (Array.isArray(body.tools)) return collaborationCatalogInfo([body.tools]).hasV2Catalog;
+  // Responses Lite carries its default catalog as the first developer input item.
+  // An explicit top-level catalog wins; later historical catalogs are not defaults.
+  if (body.tools !== undefined || !Array.isArray(body.input)) return false;
+  const initial = body.input[0];
+  return isPlainObject(initial)
+    && initial.type === "additional_tools"
+    && initial.role === "developer"
+    && Array.isArray(initial.tools)
+    && collaborationCatalogInfo([initial.tools]).hasV2Catalog;
 }
 
 function hasOptimizedNamespaceConflict(catalogs: readonly unknown[][]): boolean {

package/src/routing/identity-domains.ts

@@ -259,38 +259,45 @@ function memberMatches(member: string, ref: CredentialDomainRef): boolean {
 }
 
 /**
- * Every way a declared grouping can be ambiguous, as operator-readable messages. The
- * config write path rejects on any of these and the load path drops the list, so an
- * ambiguous declaration is reported rather than resolved by whichever group came first.
+ * Every way a declared grouping can be ambiguous, reported by position only. Messages
+ * name group and member indexes, never the operator-supplied strings — a malformed
+ * credential pasted into this list would otherwise be printed verbatim into shared
+ * logs. The config write path rejects on any of these and the load path drops the
+ * list, so an ambiguous declaration is reported rather than resolved by whichever
+ * group came first.
  */
 export function credentialGroupIssues(groups: readonly DeclaredCredentialGroup[]): string[] {
   const issues: string[] = [];
-  const seenIds = new Set<string>();
-  const owner = new Map<string, string>();
-  for (const group of groups) {
+  const seenIds = new Map<string, number>();
+  const owner = new Map<string, { groupIndex: number; memberIndex: number }>();
+  for (const [groupIndex, group] of groups.entries()) {
     // A duplicate id is not cosmetic: both groups key to `declared:<id>`, so the second
     // group's members join the first group's quota domain without anyone saying so.
-    if (seenIds.has(group.id)) issues.push(`duplicate group id ${JSON.stringify(group.id)}`);
-    seenIds.add(group.id);
+    const firstGroupIndex = seenIds.get(group.id);
+    if (firstGroupIndex !== undefined) {
+      issues.push(`duplicate group id at group index ${groupIndex} (first declared at group index ${firstGroupIndex})`);
+    } else {
+      seenIds.set(group.id, groupIndex);
+    }
     if (group.credentials.length === 0) {
-      issues.push(`group ${JSON.stringify(group.id)} lists no credentials`);
+      issues.push(`group at index ${groupIndex} lists no credentials`);
     }
-    for (const member of group.credentials) {
+    for (const [memberIndex, member] of group.credentials.entries()) {
       if (splitMember(member) === undefined) {
         issues.push(
-          `group ${JSON.stringify(group.id)} member ${JSON.stringify(member)} must be provider-qualified as "<provider>:<credential-id>"`,
+          `credential at member index ${memberIndex} in group index ${groupIndex} must be provider-qualified as "<provider>:<credential-id>"`,
         );
         continue;
       }
       const existing = owner.get(canonicalMember(member));
-      if (existing === group.id) {
-        issues.push(`credential ${JSON.stringify(member)} is listed twice in group ${JSON.stringify(group.id)}`);
+      if (existing?.groupIndex === groupIndex) {
+        issues.push(`credential at member index ${memberIndex} is listed twice in group index ${groupIndex}`);
       } else if (existing !== undefined) {
         issues.push(
-          `credential ${JSON.stringify(member)} is declared in more than one group (${existing}, ${group.id})`,
+          `credential at member index ${memberIndex} in group index ${groupIndex} is declared in more than one group (first declared at group index ${existing.groupIndex}, member index ${existing.memberIndex})`,
         );
       } else {
-        owner.set(canonicalMember(member), group.id);
+        owner.set(canonicalMember(member), { groupIndex, memberIndex });
       }
     }
   }

package/src/server/index/websocket-handler.ts

@@ -86,6 +86,7 @@ import {
 } from "../live";
 import type { ServeOptionsContext } from "./serve-options";
 import type { RequestMetricsRecorder } from "../request-metrics";
+import { resolveInboundBodyLimitBytes } from "../request-decompress";
 
 /**
  * The WebSocket half of the Bun.serve options, split out of serve-options.ts to keep that file
@@ -190,12 +191,31 @@ export function createWebsocketHandler(
           ws.close(1009, "message too large");
           return;
         }
+        // An established control connection only ever carries control frames, so the
+        // inbound body limit applies to raw bytes before the parse materializes them.
+        if (ws.data.nativeControl && rawBytes > resolveInboundBodyLimitBytes(config.maxInboundBodyBytes)) {
+          sendJsonFrame(ws, buildWsErrorFrame(413, {
+            type: "invalid_request_error",
+            code: "inbound_body_too_large",
+            message: "Native response control frame exceeds the configured inbound body limit.",
+          }));
+          return;
+        }
         let frame: Record<string, unknown>;
         try {
           frame = JSON.parse(typeof raw === "string" ? raw : raw.toString()) as Record<string, unknown>;
         } catch {
           return; // text-only contract; ignore unparseable frames
         }
+        if ((frame.type === "response.inject" || frame.type === "response.steer")
+          && rawBytes > resolveInboundBodyLimitBytes(config.maxInboundBodyBytes)) {
+          sendJsonFrame(ws, buildWsErrorFrame(413, {
+            type: "invalid_request_error",
+            code: "inbound_body_too_large",
+            message: "Native response control frame exceeds the configured inbound body limit.",
+          }));
+          return;
+        }
         if (frame.type === "response.inject" || frame.type === "response.steer" || (frame.type === "response.create" && ws.data.nativeControl)) {
           try {
             if (frame.type === "response.inject") {
@@ -227,8 +247,8 @@ export function createWebsocketHandler(
           const idleMs = typeof config.stallTimeoutSec === "number" && Number.isFinite(config.stallTimeoutSec)
             ? Math.max(1, config.stallTimeoutSec) * 1000 : 300_000;
           const mode = nativeResponseControlMode(frame, config);
-          nativeControl = mode === "injection" ? new NativeInjectionChannel(frame, idleMs)
-            : mode === "steering" ? new NativeSteeringChannel(frame, idleMs) : undefined;
+          nativeControl = mode === "injection" ? new NativeInjectionChannel(frame, idleMs, config.maxUpstreamBodyBytes)
+            : mode === "steering" ? new NativeSteeringChannel(frame, idleMs, config.maxUpstreamBodyBytes) : undefined;
         } catch {
           sendJsonFrame(ws, buildWsErrorFrame(400, { type: "invalid_request_error", message: "Invalid native steering request settings" }));
           return;