npm - @coseung2/opencodex - Versions diffs - 2.8.0-cs.14 → 2.8.0-cs.16 - Mend

@coseung2/opencodex 2.8.0-cs.14 → 2.8.0-cs.16

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (20) hide show

package/gui/dist/assets/{index-MUpaVatk.js → index-BZGMtkmp.js} +2 -2
package/gui/dist/index.html +1 -1
package/package.json +1 -1
package/packages/ocx-notch/README.md +3 -1
package/src/adapters/identity.ts +1 -1
package/src/adapters/openai-responses.ts +39 -1
package/src/cli/observe.ts +18 -3
package/src/codex/app-server-processes.ts +3 -5
package/src/codex/catalog/metadata.ts +4 -0
package/src/codex/catalog/sync.ts +3 -0
package/src/codex/routing.ts +7 -1
package/src/lib/errors.ts +18 -0
package/src/providers/opencode-go-transport.ts +41 -0
package/src/providers/quota.ts +68 -60
package/src/providers/registry.ts +7 -0
package/src/server/request-log-conversation.ts +8 -0
package/src/server/request-log.ts +5 -4
package/src/server/responses/core.ts +61 -4
package/src/usage/log.ts +1 -2
package/vendor/ocx-notch/win32-x64/ocx-notch.exe +0 -0

package/gui/dist/index.html CHANGED Viewed

@@ -16,7 +16,7 @@
         } catch (e) {}
       })();
     </script>
-    <script type="module" crossorigin src="/assets/index-MUpaVatk.js"></script>
+    <script type="module" crossorigin src="/assets/index-BZGMtkmp.js"></script>
     <link rel="stylesheet" crossorigin href="/assets/index-CxisOo-q.css">
   </head>
   <body>

package/package.json CHANGED Viewed

@@ -1,6 +1,6 @@
 {
   "name": "@coseung2/opencodex",
-  "version": "2.8.0-cs.14",
+  "version": "2.8.0-cs.16",
   "description": "Universal provider proxy for OpenAI Codex & Claude Code — use any LLM with Codex CLI/App/SDK and Claude Code",
   "type": "module",
   "main": "./bin/package-main.mjs",

package/packages/ocx-notch/README.md CHANGED Viewed

@@ -45,7 +45,7 @@ Startup failures and Rust panics are written to the bounded diagnostic log `%LOC
 - Online shutdown uses OCX's authenticated `POST /api/stop` graceful-stop endpoint directly, with CLI fallback for older OCX versions. Power transitions probe health immediately and then every ~75ms with a short timeout so the control reflects the real listener state quickly.
 - The header keeps Private and WS on separate rows, with fixed 0-to-Max segmented gauges beside them. The unboxed power and minus controls retain generous invisible hit areas and show hover/pressed feedback.
 - Click the notch to expand provider details.
-- Use the inline Providers and Logs tabs below the memory header to switch content. Logs show only the latest 10 requests with status, duration, relative time, reasoning effort, Fast state, and token usage.
+- Use the inline Providers and Logs tabs below the memory header to switch content. Logs show only the latest 10 requests with status, output tok/s, relative time, reasoning effort, Fast state, and token usage. Estimated rates are prefixed with `~`; unavailable rates use an em dash.
 - When expanded, click the top-right minus control to collapse back to the 58px notch.
 - Drag either side edge to resize the notch width. Position and width are restored on the next launch.
 - Drag anywhere on the notch to move it; its chosen position is preserved while it expands, collapses, or refreshes.
@@ -58,3 +58,5 @@ Startup failures and Rust panics are written to the bounded diagnostic log `%LOC
 - Right-click to add a provider, set the real OCX account rotation threshold, fine-tune it by 1%, **Refresh**, or **Exit**. The opaque provider modal groups supported presets into Account, Free, and Paid tabs. Canonical OpenAI adds another Codex account, OAuth presets use browser/device authorization, and required-key presets use masked API-key entry. Fixed-endpoint key-optional presets (including OpenCode Free and MiMo Free) accept an empty key; entering a key instead adds a distinct account slot to OCX's existing key pool, which can be switched from the provider row. Ollama, vLLM, and LM Studio use local auth without a key. Cloudflare Workers AI also appears in Free and asks for its Account ID before creating the provider. Endpoint-choice and other unresolved placeholder-URL presets remain omitted because this compact modal cannot preserve those setup contracts safely. `Off` writes threshold `0`.
 Quota percentages are shown as used percentages with 5-hour/weekly/monthly/custom-window columns in one compact row, reset countdowns, 5px progress bars, green fill, and the green-to-amber threshold warning used by the OCX dashboard. Columns are derived from whichever quota windows the management API returns, so provider-specific windows appear without a plan-name list in the notch. Provider usage is merged by exact provider name, limited to the newest day, and formatted with Korean `만/억/조` units. OpenAI account rows show their own 5-hour/weekly/monthly quotas; OAuth and key-pool rows show their masked identity and active/health state, while provider-level quota remains associated with the active account. The native window uses a subtle 238/255 global alpha.
+For OpenCode Go, the collapsed header keeps the provider report's 5-hour/weekly/monthly windows rather than copying a monthly-only key row, and each expanded key row shows that key's own 5-hour/weekly/monthly allocation.

package/src/adapters/identity.ts CHANGED Viewed

@@ -25,7 +25,7 @@ export const CODEX_GPT5_IDENTITY_LINE_AGENT = "You are Codex, an agent based on
  * Avoid a broad `You are Codex.*` rewrite that could touch unrelated content.
  */
 const CODEX_GPT5_IDENTITY_RE =
-  /You are Codex, (?:a coding agent|an agent) based on GPT-5(?:\.[0-9]+)*\./g;
+  /You are Codex, (?:a coding agent|an agent) based on GPT-[0-9]+(?:\.[0-9]+)*\./g;
 /** Proxy-neutral replacement: no "opencodex proxy" mention, just the GPT-5/OpenAI disclaimer. */
 export const NEUTRAL_IDENTITY_LINE = "You are a coding agent. Do not claim to be GPT-5 or to be made by OpenAI.";

package/src/adapters/openai-responses.ts CHANGED Viewed

@@ -255,6 +255,44 @@ function normalizeConfiguredReasoningSummaryDelivery(
  * - Drops tool_search_call/tool_search_output input items
  * - Sets parallel_tool_calls to false
  */
+const MUSE_SPARK_WEB_SEARCH_STRICT_MODELS = new Set(["muse-spark-1.3-contributor", "muse-spark-1.2-contributor"]);
+function stripMuseSparkUnsupportedWebSearchFields(body: unknown, modelId: unknown): unknown {
+  if (!isPlainObject(body) || typeof modelId !== "string"
+    || !MUSE_SPARK_WEB_SEARCH_STRICT_MODELS.has(modelId.trim().toLowerCase())) return body;
+  const rewrite = (tools: unknown[]) => {
+    let changed = false;
+    const next = tools.map(tool => {
+      if (!isPlainObject(tool) || tool.type !== "web_search") return tool;
+      const copy = { ...tool };
+      let toolChanged = false;
+      for (const field of ["search_content_types", "indexed_web_access"]) {
+        if (Object.hasOwn(copy, field)) { delete copy[field]; toolChanged = true; }
+      }
+      if (toolChanged) changed = true;
+      return toolChanged ? copy : tool;
+    });
+    return changed ? next : tools;
+  };
+  let changed = false;
+  const next: Record<string, unknown> = { ...body };
+  if (Array.isArray(body.tools)) {
+    const tools = rewrite(body.tools);
+    if (tools !== body.tools) { next.tools = tools; changed = true; }
+  }
+  if (Array.isArray(body.input)) {
+    const input = body.input.map(item => {
+      if (!isPlainObject(item) || item.type !== "additional_tools" || !Array.isArray(item.tools)) return item;
+      const tools = rewrite(item.tools);
+      if (tools === item.tools) return item;
+      changed = true;
+      return { ...item, tools };
+    });
+    if (changed) next.input = input;
+  }
+  return changed ? next : body;
+}
 function stripSparkCompatibility(body: unknown): unknown {
   if (!isPlainObject(body)) return body;
   const model = typeof body.model === "string" ? body.model : "";
@@ -1024,7 +1062,7 @@ export function createResponsesPassthroughAdapter(provider: OcxProviderConfig):
       if (parsed._compactionRequest === true && !isCanonicalOpenAiForwardProvider(provider)) {
         outBody = buildRoutedCompactionBody(outBody);
       }
-      const sanitizedBody = normalizeToolSchemas(stripSparkCompatibility(stripUnsupportedReasoningParams(stripItemIdsWhenUnstored(stripInvalidItemIds(stripUnsupportedHostedTools(sanitizeReasoningInputContent(scrubOcxCompactionItems(outBody))))))));
+      const sanitizedBody = normalizeToolSchemas(stripMuseSparkUnsupportedWebSearchFields(stripSparkCompatibility(stripUnsupportedReasoningParams(stripItemIdsWhenUnstored(stripInvalidItemIds(stripUnsupportedHostedTools(sanitizeReasoningInputContent(scrubOcxCompactionItems(outBody))))))), parsed.modelId));
       const body = JSON.stringify(stripDisabledReasoningSummaries(
         normalizeConfiguredReasoningSummaryDelivery(sanitizedBody, provider, parsed.modelId),
         provider,

package/src/cli/observe.ts CHANGED Viewed

@@ -21,7 +21,16 @@ const USAGE = `Usage:
   ocx observe claude-inbound [--limit <n>] [--json]
   ocx observe injection [--limit <n>] [--json]`;
-type LogEntry = Record<string, unknown> & { id?: string | number; timestamp?: string; provider?: string; model?: string; status?: number };
+type LogEntry = Record<string, unknown> & {
+  id?: string | number;
+  timestamp?: string;
+  provider?: string;
+  model?: string;
+  status?: number;
+  displayMetrics?: {
+    tokPerSecond?: { kind?: string; value?: number; estimated?: boolean };
+  };
+};
 function query(params: Record<string, string | number | undefined>): string {
   const search = new URLSearchParams();
@@ -43,8 +52,14 @@ function formatLog(row: LogEntry): string {
   const time = String(row.timestamp ?? row.createdAt ?? "");
   const route = [row.provider, row.model].filter(Boolean).join("/");
   const status = row.status ?? row.statusCode ?? "?";
-  const duration = row.durationMs !== undefined ? `${String(row.durationMs)}ms` : "";
-  return [time, String(status), route, duration].filter(Boolean).join("  ");
+  const metric = row.displayMetrics?.tokPerSecond;
+  const rate = metric?.kind === "value"
+    && typeof metric.value === "number"
+    && Number.isFinite(metric.value)
+    && metric.value > 0
+    ? `${metric.estimated ? "~" : ""}${metric.value.toLocaleString("en-US", { maximumFractionDigits: 1 })} tok/s`
+    : "— tok/s";
+  return [time, String(status), route, rate].filter(Boolean).join("  ");
 }
 async function logs(argv: string[], deps: RuntimeApiDeps): Promise<void> {

package/src/codex/app-server-processes.ts CHANGED Viewed

@@ -329,11 +329,9 @@ export function listWindowsSnapshots(): ProcessSnapshot[] {
   const psCommand = [
     "$ErrorActionPreference='SilentlyContinue'",
     "$me=[System.Security.Principal.WindowsIdentity]::GetCurrent().Name",
-    "Get-CimInstance Win32_Process | Where-Object {",
-    "  -not [string]::IsNullOrWhiteSpace($_.CommandLine) -and (",
-    `    $_.CommandLine -match ${basenameMatch} -or`,
-    `    $_.CommandLine -match ${codeModeMatch}`,
-    "  )",
+    "Get-CimInstance Win32_Process -Filter \"CommandLine LIKE '%codex%' OR CommandLine LIKE '%code-mode-host%'\" -Property Handle,ProcessId,CommandLine | Where-Object {",
+    `  $_.CommandLine -match ${basenameMatch} -or`,
+    `  $_.CommandLine -match ${codeModeMatch}`,
     "} | ForEach-Object {",
     "  try {",
     "    $o=Invoke-CimMethod -InputObject $_ -MethodName GetOwner -ErrorAction Stop",

package/src/codex/catalog/metadata.ts CHANGED Viewed

@@ -36,17 +36,20 @@ import type { RawEntry } from "./parsing";
 import { readCurrentCatalogOrCache, unique } from "./bundled";
 export const NATIVE_DAYBREAK_BLUE_MODEL = "gpt-daybreak-blue-latest";
+export const NATIVE_GPT6_ASTRA_MODEL = "gpt-6-astra";
 export const NATIVE_OPENAI_MODELS = [
   "gpt-5.5", "gpt-5.4", "gpt-5.4-mini", "gpt-5.3-codex-spark",
   "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna",
   NATIVE_DAYBREAK_BLUE_MODEL,
+  NATIVE_GPT6_ASTRA_MODEL,
 ];
 export const DOCUMENTED_NATIVE_OPENAI_ADDITIONS = [
   "gpt-5.3-codex-spark",
   "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna",
   NATIVE_DAYBREAK_BLUE_MODEL,
+  NATIVE_GPT6_ASTRA_MODEL,
 ];
 export const SUPPORTED_NATIVE_OPENAI_SLUGS = new Set(NATIVE_OPENAI_MODELS);
@@ -69,6 +72,7 @@ export const NATIVE_OPENAI_CONTEXT_OVERRIDES: Record<string, { contextWindow?: n
   "gpt-5.6-terra": { contextWindow: NATIVE_GPT56_CONTEXT_WINDOW, maxContextWindow: NATIVE_GPT56_CONTEXT_WINDOW, maxInputTokens: NATIVE_GPT56_MAX_INPUT_TOKENS },
   "gpt-5.6-luna": { contextWindow: NATIVE_GPT56_CONTEXT_WINDOW, maxContextWindow: NATIVE_GPT56_CONTEXT_WINDOW, maxInputTokens: NATIVE_GPT56_MAX_INPUT_TOKENS },
   [NATIVE_DAYBREAK_BLUE_MODEL]: { contextWindow: NATIVE_GPT56_CONTEXT_WINDOW, maxContextWindow: NATIVE_GPT56_CONTEXT_WINDOW, maxInputTokens: NATIVE_GPT56_MAX_INPUT_TOKENS },
+  [NATIVE_GPT6_ASTRA_MODEL]: { contextWindow: 272_000, maxContextWindow: 872_000, maxInputTokens: 872_000 },
 };
 export function nativeOpenAiContextWindow(slug: string): number | undefined {

package/src/codex/catalog/sync.ts CHANGED Viewed

@@ -233,6 +233,9 @@ export function deriveEntry(
   };
   if (isRouted) {
     applyReasoningLevels(entry, model?.reasoningEfforts, model?.defaultReasoningEffort, preserveExact);
+    // The no-template fallback does not pass through normalizeRoutedCatalogEntry, so pin the
+    // capability before strict defaults would otherwise advertise parallel calls unconditionally.
+    entry.supports_parallel_tool_calls = model?.provider === "cursor" || model?.parallelToolCalls === true;
   }
   else {
     applyReasoningLevels(entry, isGpt56NativeSlug(slug) ? undefined : ["low", "medium", "high", "xhigh"]);

package/src/codex/routing.ts CHANGED Viewed

@@ -298,8 +298,14 @@ export function computeCodexUsageScore(quota: {
       ? quota.monthlyPercent
       : CODEX_UNKNOWN_USAGE_SCORE;
   }
+  if (isCodexFiveHourQuotaPlan(plan)
+    && typeof quota.fiveHourPercent === "number"
+    && Number.isFinite(quota.fiveHourPercent)
+    && quota.fiveHourPercent >= 0
+    && quota.fiveHourPercent <= 100) {
+    return quota.fiveHourPercent;
+  }
   const values = [
-    ...(isCodexFiveHourQuotaPlan(plan) ? [quota.fiveHourPercent] : []),
     quota.weeklyPercent,
     quota.monthlyPercent,
   ]

package/src/lib/errors.ts CHANGED Viewed

@@ -4,6 +4,24 @@ export interface OcxErrorPayload {
   code: string | null;
 }
+/** Canonical human-readable message paths used by Responses upstream failures. */
+export function upstreamErrorMessageFromPayload(payload: unknown): string | undefined {
+  if (!payload || typeof payload !== "object" || Array.isArray(payload)) return undefined;
+  const json = payload as {
+    error?: { message?: unknown };
+    last_error?: { message?: unknown };
+    response?: {
+      error?: { message?: unknown };
+      incomplete_details?: { message?: unknown };
+    };
+  };
+  const message = json.error?.message
+    ?? json.last_error?.message
+    ?? json.response?.error?.message
+    ?? json.response?.incomplete_details?.message;
+  return typeof message === "string" ? message : undefined;
+}
 /** OpenAI / Codex hard block for high-risk cybersecurity activity (HTTP 400 or mid-stream). */
 export const CYBER_POLICY_ERROR_CODE = "cyber_policy";

package/src/providers/opencode-go-transport.ts ADDED Viewed

@@ -0,0 +1,41 @@
+import { createHash } from "node:crypto";
+import type { OcxProviderConfig } from "../types";
+import { registryEntryForProviderDestination } from "./registry";
+export const OPENCODE_GO_SESSION_HEADER = "x-opencode-session";
+function hasHeaderCaseInsensitive(
+  headers: Record<string, string> | undefined,
+  name: string,
+): boolean {
+  const target = name.toLowerCase();
+  return Object.keys(headers ?? {}).some(key => key.toLowerCase() === target);
+}
+/** Derive a provider-scoped opaque value without exposing Codex task or subagent ids. */
+export function deriveOpenCodeGoSessionId(sessionLane: string): string {
+  const digest = createHash("sha256")
+    .update("opencodex/opencode-go/session/v1\0")
+    .update(sessionLane)
+    .digest("hex")
+    .slice(0, 32);
+  return `ocx_${digest}`;
+}
+/** Add per-conversation Go affinity only to the canonical fixed-key destination. */
+export function resolveOpenCodeGoTransport<T extends OcxProviderConfig>(
+  provider: T,
+  sessionLane: string | undefined,
+): T {
+  if (registryEntryForProviderDestination(provider)?.id !== "opencode-go") return provider;
+  if (!sessionLane) return provider;
+  if (hasHeaderCaseInsensitive(provider.headers, OPENCODE_GO_SESSION_HEADER)) return provider;
+  return {
+    ...provider,
+    headers: {
+      ...(provider.headers ?? {}),
+      [OPENCODE_GO_SESSION_HEADER]: deriveOpenCodeGoSessionId(sessionLane),
+    },
+  };
+}

package/src/providers/quota.ts CHANGED Viewed

@@ -992,6 +992,22 @@ function opencodeGoSegments(api: NonNullable<Awaited<ReturnType<typeof fetchOpen
   return segments;
 }
+function opencodeGoQuotaFromSegments(
+  segments: NonNullable<ProviderQuotaWindow["segments"]>,
+  now: number,
+): ProviderQuota | null {
+  if (segments.length === 0) return null;
+  return {
+    customWindows: [{
+      // Row label intentionally empty: the segments carry their own labels.
+      label: "",
+      percent: 0,
+      segments,
+    }],
+    updatedAt: now,
+  };
+}
 /**
  * opencode.go allocation: prefer the key-scoped usage endpoint (exact console
  * percents + real reset times); fall back to the local request-count estimate
@@ -1001,18 +1017,8 @@ async function fetchOpencodeGoQuota(name: string, config: OcxProviderConfig): Pr
   const activeKey = resolveEnvValue(config.apiKey)?.trim() ?? config.apiKey;
   const api = await fetchOpencodeGoUsageApi(activeKey).catch(() => null);
   if (api) {
-    const segments = opencodeGoSegments(api);
-    if (segments.length > 0) {
-      return report(name, "opencode-go:usage-api", {
-        customWindows: [{
-          // Row label intentionally empty: the segments carry their own labels.
-          label: "",
-          percent: 0,
-          segments,
-        }],
-        updatedAt: Date.now(),
-      });
-    }
+    const quota = opencodeGoQuotaFromSegments(opencodeGoSegments(api), Date.now());
+    if (quota) return report(name, "opencode-go:usage-api", quota);
   }
   // Fallback: dominant model's local request counts against published limits.
@@ -1023,36 +1029,30 @@ async function fetchOpencodeGoQuota(name: string, config: OcxProviderConfig): Pr
   if (!dominant) return null;
   const { fiveHour, weekly, monthly } = OPENCODE_GO_LIMITS[dominant]!;
   const now = Date.now();
-  return report(name, "opencode-go:docs-estimate", {
-    customWindows: [{
-      label: "",
-      percent: 0,
-      segments: [
-        {
-          label: "5h",
-          percent: normalizePercent(((estimate.fiveHourCounts.get(dominant) ?? 0) / fiveHour) * 100) ?? 0,
-          resetAt: now + OPENCODE_GO_FIVE_HOUR_MS,
-        },
-        {
-          label: "Weekly",
-          percent: normalizePercent(((estimate.weeklyCounts.get(dominant) ?? 0) / weekly) * 100) ?? 0,
-          resetAt: now + OPENCODE_GO_WEEK_MS,
-        },
-        {
-          label: "Monthly",
-          percent: normalizePercent(((estimate.monthlyCounts.get(dominant) ?? 0) / monthly) * 100) ?? 0,
-          resetAt: now + OPENCODE_GO_COST_WINDOW_MS,
-        },
-      ],
-    }],
-    updatedAt: now,
-  });
+  const quota = opencodeGoQuotaFromSegments([
+    {
+      label: "5h",
+      percent: normalizePercent(((estimate.fiveHourCounts.get(dominant) ?? 0) / fiveHour) * 100) ?? 0,
+      resetAt: now + OPENCODE_GO_FIVE_HOUR_MS,
+    },
+    {
+      label: "Weekly",
+      percent: normalizePercent(((estimate.weeklyCounts.get(dominant) ?? 0) / weekly) * 100) ?? 0,
+      resetAt: now + OPENCODE_GO_WEEK_MS,
+    },
+    {
+      label: "Monthly",
+      percent: normalizePercent(((estimate.monthlyCounts.get(dominant) ?? 0) / monthly) * 100) ?? 0,
+      resetAt: now + OPENCODE_GO_COST_WINDOW_MS,
+    },
+  ], now);
+  return quota ? report(name, "opencode-go:docs-estimate", quota) : null;
 }
 /**
- * Per-key monthly-allocation percent for every connected key: the usage endpoint
- * answers per key, so each pool key reports its own real monthly percent. Keys the
- * endpoint rejects fall back to the local 30-day request estimate.
+ * Per-key 5h/weekly/monthly allocation for every connected key. The usage
+ * endpoint is key-scoped, so each pool key reports its own live windows. Keys
+ * the endpoint rejects fall back to the local request-count estimate.
  */
 export async function opencodeGoKeyQuotaEstimates(config: OcxConfig, name: string): Promise<Record<string, ProviderQuota> | null> {
   const provider = config.providers[name];
@@ -1064,40 +1064,48 @@ export async function opencodeGoKeyQuotaEstimates(config: OcxConfig, name: strin
   const now = Date.now();
   const out: Record<string, ProviderQuota> = {};
   const pool = provider.apiKeyPool ?? [];
+  const activeKey = resolveEnvValue(provider.apiKey)?.trim() ?? provider.apiKey;
+  const activeKeyId = activeKey ? pool.find(entry => entry.key === activeKey)?.id : undefined;
   if (pool.length > 0) {
     const results = await Promise.all(pool.map(async entry => {
       const api = await fetchOpencodeGoUsageApi(entry.key).catch(() => null);
-      return [entry.id, api?.monthly] as const;
+      return [entry.id, api] as const;
     }));
-    for (const [keyId, monthly] of results) {
-      if (monthly?.percent === undefined) continue;
-      out[keyId] = {
-        customWindows: [{
-          label: "월간 할당",
-          percent: monthly.percent,
-          ...(monthly.resetAt !== undefined ? { resetAt: monthly.resetAt } : {}),
-        }],
-        updatedAt: now,
-      };
+    for (const [keyId, api] of results) {
+      if (!api) continue;
+      const quota = opencodeGoQuotaFromSegments(opencodeGoSegments(api), now);
+      if (quota) out[keyId] = quota;
     }
   }
   // Fallback for keys the endpoint did not answer.
   const estimate = estimateOpencodeGoUsage(name, provider);
   if (estimate) {
     const dominant = [...estimate.monthlyCounts.entries()]
-      .sort((a, b) => b[1] - a[1])[0]?.[0];
+      .sort((a, b) => b[1] - a[1] || (estimate.weeklyCounts.get(b[0]) ?? 0) - (estimate.weeklyCounts.get(a[0]) ?? 0))[0]?.[0];
     if (dominant) {
-      const monthlyLimit = OPENCODE_GO_LIMITS[dominant]!.monthly;
+      const limits = OPENCODE_GO_LIMITS[dominant]!;
       for (const [keyId, count] of estimate.perKeyMonthlyCounts) {
         if (out[keyId]) continue;
-        out[keyId] = {
-          customWindows: [{
-            label: "월간 할당",
-            percent: normalizePercent((count / monthlyLimit) * 100) ?? 0,
-            resetAt: now + OPENCODE_GO_COST_WINDOW_MS,
-          }],
-          updatedAt: now,
-        };
+        const segments: NonNullable<ProviderQuotaWindow["segments"]> = [];
+        if (keyId === activeKeyId) {
+          segments.push({
+            label: "5h",
+            percent: normalizePercent(((estimate.fiveHourCounts.get(dominant) ?? 0) / limits.fiveHour) * 100) ?? 0,
+            resetAt: now + OPENCODE_GO_FIVE_HOUR_MS,
+          });
+          segments.push({
+            label: "Weekly",
+            percent: normalizePercent(((estimate.weeklyCounts.get(dominant) ?? 0) / limits.weekly) * 100) ?? 0,
+            resetAt: now + OPENCODE_GO_WEEK_MS,
+          });
+        }
+        segments.push({
+          label: "Monthly",
+          percent: normalizePercent((count / limits.monthly) * 100) ?? 0,
+          resetAt: now + OPENCODE_GO_COST_WINDOW_MS,
+        });
+        const quota = opencodeGoQuotaFromSegments(segments, now);
+        if (quota) out[keyId] = quota;
       }
     }
   }

package/src/providers/registry.ts CHANGED Viewed

@@ -808,6 +808,9 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     // catalog authoritative so a spurious 2xx from runtime.../models cannot drop seeded ids
     // (e.g. newly listed GPT-5.6 tiers) via live-discovery reconciliation.
     liveModels: false,
+    // Kiro rejects request-level parallel tool calls; keep persisted presets and the Codex
+    // catalog aligned with the adapter's single-call capability guard.
+    parallelToolCalls: false,
     // Per-model context metadata is maintained next to the Kiro model list.
     modelContextWindows: KIRO_MODEL_CONTEXT_WINDOWS,
     modelReasoningEfforts: KIRO_MODEL_REASONING_EFFORTS,
@@ -862,6 +865,10 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     id: "opencode-go", label: "opencode go", adapter: "openai-chat", baseUrl: "https://opencode.ai/zen/go/v1",
     authKind: "key", featured: true, dashboardUrl: "https://opencode.ai/auth", defaultModel: "kimi-k2.7-code",
     jawcodeBundle: "opencode-go", note: "GLM, DeepSeek, Kimi, Qwen, MiMo…",
+    modelWireDefaults: {
+      "muse-spark-1.3-contributor": "openai-responses",
+      "muse-spark-1.2-contributor": "openai-responses",
+    },
     modelContextWindows: {
       "kimi-k3": KIMI_K3_STANDARD_CONTEXT_WINDOW,
       [OPENCODE_OX_ALPHA_FREE_MODEL]: OX_ALPHA_CONTEXT_WINDOW,

package/src/server/request-log-conversation.ts CHANGED Viewed

@@ -61,6 +61,14 @@ export function sessionIdHeaderFromRequest(headers: Headers): string | null {
   return headers.get("session_id") ?? headers.get("session-id");
 }
+export function sessionLaneIdFromRequest(headers: Headers): string | undefined {
+  const parent = headers.get("x-codex-parent-thread-id")?.trim();
+  const thread = headers.get("thread-id")?.trim();
+  const session = sessionIdHeaderFromRequest(headers)?.trim();
+  const lane = [parent, thread, session].filter(Boolean);
+  return lane.length > 0 ? lane.join("\0") : undefined;
+}
 export function conversationIdFromResponsesRequest(input: {
   clientThreadId?: string;
   sessionIdHeader?: string | null;

package/src/server/request-log.ts CHANGED Viewed

@@ -4,6 +4,9 @@ import {
   classifyError,
   httpStatusFromTerminalError as httpStatusFromClassifiedTerminalError,
   isClientClosedMessage,
+  isCyberPolicyCode,
+  isCyberPolicyMessage,
+  upstreamErrorMessageFromPayload,
 } from "../lib/errors";
 import { CODEX_CONFIG_PATH, readRootTomlString } from "../codex/paths";
 import { readCodexCatalogPath } from "../codex/catalog";
@@ -674,9 +677,7 @@ function captureUpstreamErrorParsed(
       logCtx.terminalIncompleteReason = reason.trim();
     }
     if (logCtx.upstreamError) return;
-    const message = json?.error?.message
-      ?? json?.last_error?.message
-      ?? json?.response?.error?.message;
+    const message = upstreamErrorMessageFromPayload(parsed);
     if (typeof message === "string" && message.trim()) {
       logCtx.upstreamError = redactSecretString(message).slice(0, 500);
       return;
@@ -939,7 +940,7 @@ function finalizedUsage(
   const usageFallback = !finalUsage && estimate !== undefined
     ? { inputTokens: estimate, outputTokens: 0, estimated: true }
     : undefined;
-  const loggedUsage = finalUsage && estimate !== undefined
+  const loggedUsage = finalUsage?.estimated && estimate !== undefined
     ? {
         ...finalUsage,
         inputTokens: Math.max(finalUsage.inputTokens, estimate),

package/src/server/responses/core.ts CHANGED Viewed

@@ -101,10 +101,15 @@ import type { InboundWire } from "../../providers/registry";
 import { hasKeyPoolFailover, rotateProviderTransportOn429 } from "../../providers/key-failover";
 import { shouldAttemptImageTierRetry } from "../image-retry";
 import { resolveProviderTransport } from "../../providers/xai-transport";
+import { resolveOpenCodeGoTransport } from "../../providers/opencode-go-transport";
 import type { WsData } from "../ws-bridge";
 import { trackActiveTurnLease, trackStreamLifetime } from "../lifecycle";
 import { redactSecretString } from "../../lib/redact";
 import { readBoundedResponseBody } from "../../lib/bounded-body";
+import {
+  isRateLimitOrQuotaFailureMessage,
+  upstreamErrorMessageFromPayload,
+} from "../../lib/errors";
 import type { AdmissionLease } from "../../lib/admission";
 import { supportedLadderFor } from "../effort-policy";
 import { isThreadSpawnRequest } from "../effort-policy";
@@ -130,6 +135,7 @@ import {
 import {
   conversationIdFromResponsesRequest,
   normalizeLogConversationId,
+  sessionLaneIdFromRequest,
   sessionIdHeaderFromRequest,
 } from "../request-log-conversation";
 import type { AttemptRecoveryKind } from "../../usage/log";
@@ -252,8 +258,40 @@ async function shouldRetryCodexPoolAccountModel400(
 }
 /** Pre-stream quota/billing rejections that warrant one alternate-account attempt (#584). */
-function shouldRetryCodexPoolAccountQuota(response: Response): boolean {
-  return response.status === 429 || response.status === 402;
+function codexQuotaFailureMessage(body: string): string | undefined {
+  try {
+    const payload = JSON.parse(body) as unknown;
+    const canonical = upstreamErrorMessageFromPayload(payload);
+    if (canonical !== undefined) return canonical;
+    if (typeof payload === "string") return payload;
+    if (!payload || typeof payload !== "object" || Array.isArray(payload)) return undefined;
+    const record = payload as Record<string, unknown>;
+    if (typeof record.message === "string") return record.message;
+    return typeof record.error === "string" ? record.error : undefined;
+  } catch {
+    // Plain-text gateways remain supported. Valid JSON is inspected only at recognized
+    // message fields so echoed request content elsewhere cannot trigger account cooldown.
+    return body;
+  }
+}
+export async function shouldRetryCodexPoolAccountQuota(
+  response: Response,
+  signal?: AbortSignal,
+): Promise<boolean> {
+  if (response.status === 402 || response.status === 429) return true;
+  if (response.status < 500 || response.status >= 600) return false;
+  try {
+    // Reject malformed UTF-8 instead of matching quota words around replacement characters.
+    const body = await readBoundedResponseBody(response.clone(), { signal, fatalUtf8: true });
+    const message = body.displaySafe && !body.truncated
+      ? codexQuotaFailureMessage(body.text)
+      : undefined;
+    return message !== undefined
+      && isRateLimitOrQuotaFailureMessage(message);
+  } catch {
+    return false;
+  }
 }
 interface CodexPoolAccountRetryArgs {
@@ -351,6 +389,19 @@ async function retryCodexPoolOnAlternateAccount(
     ) throw error;
   }
   if (retryAuthCtx?.kind !== "pool" && retryAuthCtx?.kind !== "main-pool") {
+    // A body-confirmed quota response may arrive under HTTP 5xx. Without an alternate,
+    // the ordinary terminal recorder sees only that wire status and would misclassify it
+    // as transient, leaving the exhausted account immediately selectable next turn.
+    if (outcomeStatus !== firstResponse.status && (outcomeStatus === 429 || outcomeStatus === 402)) {
+      recordCodexUpstreamOutcome(config, firstAuthCtx.accountId, outcomeStatus, {
+        ...codexQuotaOutcomeMeta(firstResponse),
+        threadId: req.headers.get("x-codex-parent-thread-id"),
+        modelId: route.modelId,
+        probeLeaseId: codexProbeLeaseId(firstAuthCtx),
+        probeQuotaScope: codexProbeQuotaScope(firstAuthCtx),
+        writerGeneration: firstAuthCtx.writerGeneration,
+      });
+    }
     return { kind: "no-alternate" };
   }
@@ -826,6 +877,7 @@ async function applyFinalRouteRequestNormalization(args: {
   }
   // Settle the wire once so logging, fast-mode, auth, and sidecars read the adapter
   // this request will actually use (#404).
+  route.provider = resolveOpenCodeGoTransport(route.provider, sessionLaneIdFromRequest(req.headers));
   route.provider = resolveWireProtocolOverride(route.providerName, route.modelId, route.provider, inboundWire);
   logCtx.model = route.modelId;
   logCtx.provider = route.providerName;
@@ -1671,9 +1723,14 @@ async function handleResponsesInner(
         options.abortSignal,
       )) {
         poolRetryOutcome = 400;
-      } else if (shouldRetryCodexPoolAccountQuota(upstreamResponse)) {
+      } else if (!authCtx.fixedAccount && await shouldRetryCodexPoolAccountQuota(
+        upstreamResponse,
+        options.abortSignal,
+      )) {
         // Pre-stream only: once SSE has begun, mid-stream quota stays terminal.
-        poolRetryOutcome = upstreamResponse.status;
+        // ChatGPT sometimes wraps quota exhaustion in a generic 5xx. Normalize only
+        // body-confirmed cases to quota evidence so cooldown and rotation both apply.
+        poolRetryOutcome = upstreamResponse.status >= 500 ? 429 : upstreamResponse.status;
       }
       if (poolRetryOutcome !== undefined) {

package/src/usage/log.ts CHANGED Viewed

@@ -147,8 +147,7 @@ export function usageTotalTokens(usage: OcxUsage | undefined): number | undefine
  * fallback for paths that only know the configured provider name (e.g. "cursor-mykey").
  */
 function isEstimatedUsageProvider(providerOrAdapter: string): boolean {
-  return providerOrAdapter === "kiro" || providerOrAdapter.startsWith("kiro-")
-    || providerOrAdapter === "cursor" || providerOrAdapter.startsWith("cursor-");
+  return providerOrAdapter === "cursor" || providerOrAdapter.startsWith("cursor-");
 }
 export function usageForFinalLog(provider: string, usage: OcxUsage | undefined): OcxUsage | undefined {

package/vendor/ocx-notch/win32-x64/ocx-notch.exe CHANGED Viewed

Binary file