pi-freeflow 1.27.0 → 1.28.1

package/CHANGELOG.md

@@ -1,5 +1,31 @@
 # Changelog
 
+## 1.28.1
+
+### Patch Changes
+
+- Retry relay reachability checks after deployment so edge routing propagation delays do not show a fresh relay as unreachable — the check retries with progress notices and only warns if the relay stays dark.
+
+  Fix upstream drift alerts to stop flagging commercial paid model families as free candidates, while known alpha and stealth free models keep their free status.
+
+## 1.28.0
+
+### Minor Changes
+
+- 526d43f: Refresh the free model catalog to the current live rotation (33 models).
+
+  - New: Space Bunny Alpha on the KiloCode Gateway free list (1M context window) — pick it with the short name `space-bunny-alpha`.
+  - Back: Step 3.7 Flash returns to the KiloCode Gateway free list after a live serving re-check.
+  - Retired from the picker: Nex N2.5 Pro, Nex N2.5 Mini, and GLM 5.2 free — all three are gone from the free list (only paid versions remain).
+
+- 526d43f: Add LongCat 2.5 Preview Free (`longcat-2.5-preview-free`) from OpenCode Zen with a 1M context window, reasoning, and vision.
+
+  Fix upstream drift detection to accurately recognize free and alpha models while excluding commercial paid model families from candidate suggestions.
+
+### Patch Changes
+
+- 526d43f: Retry relay reachability verification after deployment so edge DNS and routing propagation delays (such as newly created Cloudflare Workers routes) do not show a false unreachable warning.
+
 ## 1.27.0
 
 ### Minor Changes

package/README.md

@@ -16,7 +16,7 @@ Thin by design: a model list, a relay proxy, and a log. The host (`pi-ai`) handl
 
 | Feature | What it does | Cost |
 | :--- | :--- | :--- |
-| **33 free models** | 9 from OpenCode Zen, 19 from KiloCode Gateway, 5 from Cline, context windows up to 1M. Full list below. | **$0** |
+| **33 free models** | 10 from OpenCode Zen, 18 from KiloCode Gateway, 5 from Cline, context windows up to 1M. Full list below. | **$0** |
 | **Relay pool** | Route requests through your own Cloudflare Workers and Vercel Edge relays. Requests rotate across the pool. A relay that rate-limits, times out, or drops the connection cools down while healthy ones take its traffic. | **$0** beyond your platforms' free tiers |
 | **Automatic fallback** | When every relay is cooling down, requests go direct to upstream instead of failing. | **$0** |
 | **Short model names** | Every model has a slash-free, colon-free alias, plus an optional `:effort` suffix for thinking depth. You type `freeflow/<name>`. | **$0** |
@@ -190,7 +190,7 @@ The same command set works identically in OMP and Pi:
 /model → freeflow → pick
 ```
 
-#### OpenCode Zen (9 models), Responses and Chat API
+#### OpenCode Zen (10 models), Responses and Chat API
 
 Good defaults for long coding sessions and agentic work.
 
@@ -205,8 +205,9 @@ Good defaults for long coding sessions and agentic work.
 | `big-pickle` | Big Pickle | **200K** (200.000) | **32K** (32.000) | `high / max` | ❌ |
 | `ling-3.0-flash-fin-free` | Inclusion AI | **262K** (262.144) | **131K** (131.072) | `minimal … xhigh` | ❌ |
 | `space-bunny-free` | Stealth preview (lab undisclosed) | **1M** (1.048.576) | **512K** (524.288) | `low … max` | ✅ |
+| `longcat-2.5-preview-free` | Meituan LongCat | **1M** (1.000.000) | **131K** (131.072) | `minimal … max` | ✅ |
 
-#### KiloCode Gateway (19 models), OpenRouter compatible
+#### KiloCode Gateway (18 models), OpenRouter compatible
 
 Keyless access. Short aliases work for every row (the full ID is in parentheses).
 
@@ -226,11 +227,10 @@ Keyless access. Short aliases work for every row (the full ID is in parentheses)
 | `content-safety` (`nvidia/...:free`) | NVIDIA | **128K** (128.000) | **8K** (8.192) | ❌ *(non-thinking)* | ✅ |
 | `ling-3.0-flash-fin` (`inclusionai/ling-3.0-flash-fin:free`) | Inclusion AI | **262K** (262.144) | **32K** (32.768) | `minimal…xhigh`\* | ❌ |
 | `ling-3.0-flash-sante` (`inclusionai/ling-3.0-flash-sante:free`) | Inclusion AI | **262K** (262.144) | **32K** (32.768) | `minimal…xhigh`\* | ❌ |
-| `nex-n2.5-pro` (`nex-agi/nex-n2.5-pro:free`) | Nex AGI | **262K** (262.144) | **235K** (235.929) | `minimal…xhigh`\* | ✅ |
-| `nex-n2.5-mini` (`nex-agi/nex-n2.5-mini:free`) | Nex AGI | **262K** (262.144) | **235K** (235.929) | `minimal…xhigh`\* | ✅ |
+| `step-3.7-flash` (`stepfun/...:free`) | StepFun | **262K** (262.144) | **262K** (262.144) | `minimal…xhigh`\* | ✅ |
+| `space-bunny-alpha` (`stealth/...`) | Stealth preview (lab undisclosed) | **1M** (1.000.000) | **512K** (524.288) | `minimal…xhigh`\* | ✅ |
 | `inkling-small` (`thinkingmachines/inkling-small:free`) | Thinking Machines | **1M** (1.048.576) | **262K** (262.144) | `minimal…xhigh`\* | ✅ |
 | `qwen3.8-27b` (`qwen/...:free`) | Alibaba Qwen | **262K** (262.144) | **235K** (235.929) | `minimal…xhigh`\* | ✅ |
-| `glm-5.2` (`z-ai/...:free`) | Zhipu AI | **32K** (32.768) | **29K** (29.491) | `minimal…xhigh`\* | ❌ |
 
 \* Levels are forwarded as-is through the OpenRouter-style nested `reasoning` parameter; effort mapping is decided by each model. MiMo collapses `minimal→low` and `xhigh→high` upstream, so its selector shows 5 labels but only 3 distinct effort values.
 

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "pi-freeflow",
   "type": "module",
-  "version": "1.27.0",
+  "version": "1.28.1",
   "description": "Thin provider for OMP/Pi — model list + dumb relay proxy + log; host pi-ai owns thinking/normalization",
   "main": "extensions/index.ts",
   "types": "src/index.ts",

package/src/catalog.ts

@@ -30,9 +30,12 @@ import type {
 
 /**
  * Pruned model IDs that must never re-enter the catalog via disk cache or upstream merge.
- * - jev-1.13-free: non-chat decision model, chat-incompatible — never registered.
- * - deepseek-v4-flash-free: listed but currently unserved upstream — stays out until it answers.
- * - stepfun/step-3.7-flash:free: gone from the live Kilo free list — only the paid counterpart remains.
+ * - jev-1.13-free: non-chat decision model, chat-incompatible (live chat 500) — never registered.
+ * - deepseek-v4-flash-free: listed but unserved upstream (live chat 400 "Model is unavailable").
+ * - nex-agi/nex-n2.5-pro:free + nex-agi/nex-n2.5-mini:free: gone 2026-09-28 — zero nex-agi
+ *   IDs on the live Kilo list, keyless chat 404 "does not exist".
+ * - z-ai/glm-5.2:free: gone 2026-09-28 — :free ID absent (keyless 404); only the paid
+ *   z-ai/glm-5.2 remains (isFree:false, keyless 401 PAID_MODEL_AUTH_REQUIRED).
  * - inclusionai/ling-3.0-flash-vl:free: gone from the live Kilo free list — only the paid counterpart remains.
  */
 export const DEAD_MODEL_IDS = new Set<string>([
@@ -47,7 +50,9 @@ export const DEAD_MODEL_IDS = new Set<string>([
  "minimax/minimax-m3:free",
  "thinkingmachines/inkling:free",
  "union-alpha",
- "stepfun/step-3.7-flash:free",
+ "nex-agi/nex-n2.5-pro:free",
+ "nex-agi/nex-n2.5-mini:free",
+ "z-ai/glm-5.2:free",
  "inclusionai/ling-3.0-flash-vl:free",
 ]);
 /**
@@ -55,7 +60,7 @@ export const DEAD_MODEL_IDS = new Set<string>([
  * Upstream lists paid models alongside free ones (e.g. claude-fable-5-1,
  * claude-opus-4-*, gemini-3-*) so a bare upstream merge leaks paid entries that
  * fail with 401 Missing API key. Known static IDs without a free suffix
- * (e.g. big-pickle) stay allowed via MODEL_MAP.
+ * (e.g. big-pickle, stealth/space-bunny-alpha) stay allowed via MODEL_MAP.
  */
 export function isFreeCatalogId(id: string): boolean {
  if (typeof id !== "string" || id.length === 0) return false;

package/src/commands.ts

@@ -12,7 +12,7 @@ import path from "node:path";
 import { fileURLToPath } from "node:url";
 import { getClientPort, isJsRuntimeExecutable, resolveDaemonRuntime } from "./client.ts";
 import { refreshCatalog, setAliveCatalog } from "./catalog.ts";
-import { DAEMON_SPAWN_ENV, DEBUG_STATE_FILE, HOST, LOG_FILE, PORT, RELAY_STATE_FILE } from "./config.ts";
+import { DAEMON_SPAWN_ENV, DATA_DIR_ENV, DEBUG_STATE_FILE, HOST, LOG_FILE, PORT, RELAY_STATE_FILE } from "./config.ts";
 import {
  compareVersions,
  fetchLatestVersion,
@@ -84,6 +84,17 @@ export function updateStatusBar(ui?: ExtensionUIContext): void {
 export const STARTUP_TASK_NAME = "pi-freeflow-daemon";
 export const STARTUP_SERVICE_NAME = "pi-freeflow.service";
 
+let deployProbeRetries = 5;
+let deployProbeRetryDelayMs: number | null = null;
+
+/** Test-only: override post-deploy probe retry count and delay. */
+export function _setDeployProbeOptionsForTest(
+ opts: { retries?: number; retryDelayMs?: number } | null,
+): void {
+ deployProbeRetries = opts?.retries ?? 5;
+ deployProbeRetryDelayMs = opts?.retryDelayMs ?? null;
+}
+
 export interface StartupPlan {
  platform: string;
  taskName: string;
@@ -755,7 +766,18 @@ export function createCommandSpec(
      }
      let probeNote = "";
      try {
-      const probe = await probeRelay(finalUrl, auth);
+      const retryDelay =
+       deployProbeRetryDelayMs ?? (process.env[DATA_DIR_ENV] ? 10 : 2000);
+      const probe = await probeRelay(finalUrl, auth, {
+       retries: deployProbeRetries,
+       retryDelayMs: retryDelay,
+       onRetry: (attempt, total) => {
+        ctx.ui.notify(
+         `Waiting for edge routing to propagate (${attempt}/${total})…`,
+         "info",
+        );
+       },
+      });
       probeNote = probe.ok
        ? ` ✓ reachable (HTTP ${probe.status}, ${probe.latencyMs}ms)`
        : ` ⚠ deployed but unreachable (${probe.error || `HTTP ${probe.status}`}) — run /freeflow test ${finalUrl} to retry, or /freeflow remove ${finalUrl} to drop it`;

package/src/index.ts

@@ -1,7 +1,7 @@
 /**
  * pi-freeflow — Modular, high-resiliency LLM extension for Pi & Oh My Pi (OMP)
  *
- * Provides access to 33 free models (9 OpenCode Zen + 19 KiloCode Gateway + 5 Cline) with:
+ * Provides access to 33 free models (10 OpenCode Zen + 18 KiloCode Gateway + 5 Cline) with:
  * - Single-port daemon reuse on 28180 across concurrent subagents
  * - Multi-cloud rolling egress relays (Vercel Edge, Cloudflare, Deno)
  * - 0ms instant startup with verified static catalog and background live health checks

package/src/models.ts

@@ -1,9 +1,9 @@
 /**
  * Static model definitions and upstream routing catalogs for pi-freeflow
  *
- * Defines the 33 verified free models:
- * - 9 OpenCode Zen models (2 Responses API + 7 Chat Completions)
- * - 19 KiloCode Keyless Gateway models (18 OpenRouter format + 1 Standard format)
+ * Defines the 33 verified free models (live-verified 2026-09-28):
+ * - 10 OpenCode Zen models (2 Responses API + 8 Chat Completions)
+ * - 18 KiloCode Keyless Gateway models (OpenRouter format)
  * - 5 Cline direct-only models (per-user pool)
  */
 
@@ -169,6 +169,23 @@ export const OPENCODE_MODELS: ModelDef[] = [
    max: "max",
   },
  },
+ {
+  id: "longcat-2.5-preview-free",
+  name: "LongCat 2.5 Preview (1M) [OpenCode]",
+  reasoning: true,
+  contextWindow: 1_000_000,
+  maxTokens: 131_072,
+  input: ["text", "image"],
+  thinkingLevelMap: {
+   off: null,
+   minimal: "minimal",
+   low: "low",
+   medium: "medium",
+   high: "high",
+   xhigh: "xhigh",
+   max: "max",
+  },
+ },
 ];
 
 /**
@@ -321,21 +338,26 @@ export const KILO_MODELS: ModelDef[] = [
   thinkingLevelMap: KILO_REASONING_MAP,
  },
  {
-  id: "nex-agi/nex-n2.5-pro:free",
-  name: "Nex N2.5 Pro [Kilo]",
+  // Resurrected 2026-09-28: back on the live Kilo free list
+  // (isFree:true, 0/0 pricing, no expiry) + keyless chat 200.
+  id: "stepfun/step-3.7-flash:free",
+  name: "Step 3.7 Flash [Kilo]",
   reasoning: true,
   contextWindow: 262_144,
-  maxTokens: 235_929,
+  maxTokens: 262_144,
   input: ["text", "image"],
   thinkingFormat: "openrouter",
   thinkingLevelMap: KILO_REASONING_MAP,
  },
  {
-  id: "nex-agi/nex-n2.5-mini:free",
-  name: "Nex N2.5 Mini [Kilo]",
+  // Added 2026-09-28: live on the Kilo free list (isFree:true, 0/0
+  // pricing, 1M context, text+image in) + keyless chat 200. Passes
+  // isFreeCatalogId via MODEL_MAP (no :free suffix by design).
+  id: "stealth/space-bunny-alpha",
+  name: "Space Bunny Alpha (1M) [Kilo]",
   reasoning: true,
-  contextWindow: 262_144,
-  maxTokens: 235_929,
+  contextWindow: 1_000_000,
+  maxTokens: 524_288,
   input: ["text", "image"],
   thinkingFormat: "openrouter",
   thinkingLevelMap: KILO_REASONING_MAP,
@@ -370,16 +392,6 @@ export const KILO_MODELS: ModelDef[] = [
   thinkingFormat: "openrouter",
   thinkingLevelMap: KILO_REASONING_MAP,
  },
- {
-  id: "z-ai/glm-5.2:free",
-  name: "GLM 5.2 [Kilo]",
-  reasoning: true,
-  contextWindow: 32_768,
-  maxTokens: 29_491,
-  input: ["text"],
-  thinkingFormat: "openrouter",
-  thinkingLevelMap: KILO_REASONING_MAP,
- },
 ];
 
 /**
@@ -498,13 +510,14 @@ export const MODEL_ALIASES: Record<string, string> = {
  "content-safety": "nvidia/nemotron-3.5-content-safety:free",
  "ling-3.0-flash-fin": "inclusionai/ling-3.0-flash-fin:free",
  "ling-3.0-flash-sante": "inclusionai/ling-3.0-flash-sante:free",
- "nex-n2.5-pro": "nex-agi/nex-n2.5-pro:free",
- "nex-n2.5-mini": "nex-agi/nex-n2.5-mini:free",
+ "step-3.7-flash": "stepfun/step-3.7-flash:free",
+ "space-bunny-alpha": "stealth/space-bunny-alpha",
  "inkling-small": "thinkingmachines/inkling-small:free",
  "mimo-v2.6-flash": "mimo-v2.6-flash-free",
  "space-bunny": "space-bunny-free",
+ "longcat-2.5-preview": "longcat-2.5-preview-free",
+ "longcat": "longcat-2.5-preview-free",
  "qwen3.8-27b": "qwen/qwen3.8-27b:free",
- "glm-5.2": "z-ai/glm-5.2:free",
  // provider-prefixed short aliases (slash-normalized)
  "laguna-s-2.1:free": "poolside/laguna-s-2.1:free",
  "laguna-xs-2.1:free": "poolside/laguna-xs-2.1:free",

package/src/probe.ts

@@ -1,121 +1,153 @@
-import { opencodeHeaders } from "./config.ts";
-import { ensureChatFingerprintTools } from "./opencode-fingerprint.ts";
-
-/**
- * Relay reachability probe: verifies a deployed relay answers requests
- * by issuing a minimal round trip through it and timing the response.
- */
-
-export interface RelayProbeResult {
- ok: boolean;
- status: number;
- latencyMs: number;
- error?: string;
-}
-
-export interface RelayProbeOptions {
- /**
-  * Probe mode:
-  * - "models" (default): GET /v1/models (reachability probe)
-  * - "chat": POST /zen/v1/chat/completions (verifies free tier inference & spoofing)
-  */
- endpoint?: "models" | "chat";
- /** Model to use for chat probe (defaults to "nemotron-3.5-lightning-free") */
- model?: string;
- /** Timeout in milliseconds (defaults to 5000) */
- timeoutMs?: number;
-}
-
-/**
- * Probe a relay endpoint to confirm it is reachable.
- *
- * Sends a minimal request to `<url>/v1/models` carrying the x-relay-target /
- * x-relay-path headers the relay expects, timing the round trip. Never
- * throws: network failures, timeouts, and non-2xx statuses are all reported
- * on the returned result instead.
- *
- * The path mirrors the proxy's real relay contract (`relayFetch` in relay.ts):
- * `x-relay-target` is the upstream origin and `x-relay-path` the full path
- * including the `/zen` prefix — the workers only allow the exact origins
- * `https://opencode.ai` / `https://api.kilo.ai` and forward `target + path`.
- * @param auth Optional per-relay shared secret; sent as x-relay-auth so
- *   deployed relays (which enforce it) answer the probe instead of 401.
- */
-export async function probeRelay(
- url: string,
- auth?: string,
- opts?: RelayProbeOptions,
-): Promise<RelayProbeResult> {
- const cleanUrl = url.trim().replace(/\/+$/, "");
- const isDirectOpenCode =
-  cleanUrl === "https://opencode.ai" ||
-  cleanUrl === "opencode" ||
-  cleanUrl === "direct";
- const isChat = opts?.endpoint === "chat";
- const timeout = opts?.timeoutMs ?? (isChat ? 15_000 : 5_000);
- const start = Date.now();
-
- try {
-  const headers: Record<string, string> = {
-   ...opencodeHeaders(),
-   ...(auth ? { "x-relay-auth": auth } : {}),
-  };
-
-  let fetchUrl: string;
-  let method = "GET";
-  let body: string | undefined;
-
-  if (isDirectOpenCode) {
-   if (isChat) {
-    fetchUrl = "https://opencode.ai/zen/v1/chat/completions";
-    method = "POST";
-    headers["content-type"] = "application/json";
-    const probePayload: Record<string, unknown> = {
-     model: opts?.model || "nemotron-3.5-lightning-free",
-     max_tokens: 1,
-     stream: true,
-     messages: [{ role: "user", content: "ping" }],
-    };
-    ensureChatFingerprintTools(probePayload);
-    body = JSON.stringify(probePayload);
-   } else {
-    fetchUrl = "https://opencode.ai/zen/v1/models";
-   }
-  } else {
-   if (isChat) {
-    fetchUrl = `${cleanUrl}/zen/v1/chat/completions`;
-    method = "POST";
-    headers["content-type"] = "application/json";
-    headers["x-relay-target"] = "https://opencode.ai";
-    headers["x-relay-path"] = "/zen/v1/chat/completions";
-    const probePayload: Record<string, unknown> = {
-     model: opts?.model || "nemotron-3.5-lightning-free",
-     max_tokens: 1,
-     stream: true,
-     messages: [{ role: "user", content: "ping" }],
-    };
-    ensureChatFingerprintTools(probePayload);
-    body = JSON.stringify(probePayload);
-   } else {
-    fetchUrl = `${cleanUrl}/v1/models`;
-    headers["x-relay-target"] = "https://opencode.ai";
-    headers["x-relay-path"] = "/zen/v1/models";
-   }
-  }
-
-  const res = await fetch(fetchUrl, {
-   method,
-   headers,
-   body,
-   signal: AbortSignal.timeout(timeout),
-  });
-  return { ok: res.ok, status: res.status, latencyMs: Date.now() - start };
- } catch (e) {
-  return {
-   ok: false,
-   status: 0,
-   latencyMs: Date.now() - start,
-   error: (e as Error)?.message || String(e),
-  };
- }
-}
+import { opencodeHeaders } from "./config.ts";
+import { ensureChatFingerprintTools } from "./opencode-fingerprint.ts";
+
+/**
+ * Relay reachability probe: verifies a deployed relay answers requests
+ * by issuing a minimal round trip through it and timing the response.
+ */
+
+export interface RelayProbeResult {
+ ok: boolean;
+ status: number;
+ latencyMs: number;
+ error?: string;
+}
+
+export interface RelayProbeOptions {
+ /**
+  * Probe mode:
+  * - "models" (default): GET /v1/models (reachability probe)
+  * - "chat": POST /zen/v1/chat/completions (verifies free tier inference & spoofing)
+  */
+ endpoint?: "models" | "chat";
+ /** Model to use for chat probe (defaults to "nemotron-3.5-lightning-free") */
+ model?: string;
+ /** Timeout in milliseconds (defaults to 5000) */
+ timeoutMs?: number;
+ /**
+  * Number of retry attempts when the probe returns !ok or throws.
+  * Defaults to 0 (single attempt). Useful post-deploy for edge DNS propagation.
+  */
+ retries?: number;
+ /** Delay between retries in milliseconds (defaults to 2000). */
+ retryDelayMs?: number;
+ /** Optional progress callback before each retry */
+ onRetry?: (attempt: number, maxRetries: number, lastResult: RelayProbeResult) => void;
+}
+
+/**
+ * Probe a relay endpoint to confirm it is reachable.
+ *
+ * Sends a minimal request to `<url>/v1/models` carrying the x-relay-target /
+ * x-relay-path headers the relay expects, timing the round trip. Never
+ * throws: network failures, timeouts, and non-2xx statuses are all reported
+ * on the returned result instead.
+ *
+ * The path mirrors the proxy's real relay contract (`relayFetch` in relay.ts):
+ * `x-relay-target` is the upstream origin and `x-relay-path` the full path
+ * including the `/zen` prefix — the workers only allow the exact origins
+ * `https://opencode.ai` / `https://api.kilo.ai` and forward `target + path`.
+ * @param auth Optional per-relay shared secret; sent as x-relay-auth so
+ *   deployed relays (which enforce it) answer the probe instead of 401.
+ */
+async function probeRelayOnce(
+ url: string,
+ auth?: string,
+ opts?: RelayProbeOptions,
+): Promise<RelayProbeResult> {
+ const cleanUrl = url.trim().replace(/\/+$/, "");
+ const isDirectOpenCode =
+  cleanUrl === "https://opencode.ai" ||
+  cleanUrl === "opencode" ||
+  cleanUrl === "direct";
+ const isChat = opts?.endpoint === "chat";
+ const timeout = opts?.timeoutMs ?? (isChat ? 15_000 : 5_000);
+ const start = Date.now();
+
+ try {
+  const headers: Record<string, string> = {
+   ...opencodeHeaders(),
+   ...(auth ? { "x-relay-auth": auth } : {}),
+  };
+
+  let fetchUrl: string;
+  let method = "GET";
+  let body: string | undefined;
+
+  if (isDirectOpenCode) {
+   if (isChat) {
+    fetchUrl = "https://opencode.ai/zen/v1/chat/completions";
+    method = "POST";
+    headers["content-type"] = "application/json";
+    const probePayload: Record<string, unknown> = {
+     model: opts?.model || "nemotron-3.5-lightning-free",
+     max_tokens: 1,
+     stream: true,
+     messages: [{ role: "user", content: "ping" }],
+    };
+    ensureChatFingerprintTools(probePayload);
+    body = JSON.stringify(probePayload);
+   } else {
+    fetchUrl = "https://opencode.ai/zen/v1/models";
+   }
+  } else {
+   if (isChat) {
+    fetchUrl = `${cleanUrl}/zen/v1/chat/completions`;
+    method = "POST";
+    headers["content-type"] = "application/json";
+    headers["x-relay-target"] = "https://opencode.ai";
+    headers["x-relay-path"] = "/zen/v1/chat/completions";
+    const probePayload: Record<string, unknown> = {
+     model: opts?.model || "nemotron-3.5-lightning-free",
+     max_tokens: 1,
+     stream: true,
+     messages: [{ role: "user", content: "ping" }],
+    };
+    ensureChatFingerprintTools(probePayload);
+    body = JSON.stringify(probePayload);
+   } else {
+    fetchUrl = `${cleanUrl}/v1/models`;
+    headers["x-relay-target"] = "https://opencode.ai";
+    headers["x-relay-path"] = "/zen/v1/models";
+   }
+  }
+
+  const res = await fetch(fetchUrl, {
+   method,
+   headers,
+   body,
+   signal: AbortSignal.timeout(timeout),
+  });
+  return { ok: res.ok, status: res.status, latencyMs: Date.now() - start };
+ } catch (e) {
+  return {
+   ok: false,
+   status: 0,
+   latencyMs: Date.now() - start,
+   error: (e as Error)?.message || String(e),
+  };
+ }
+}
+
+export async function probeRelay(
+ url: string,
+ auth?: string,
+ opts?: RelayProbeOptions,
+): Promise<RelayProbeResult> {
+ const retries = Math.max(0, opts?.retries ?? 0);
+ const retryDelay = Math.max(10, opts?.retryDelayMs ?? 2000);
+
+ for (let attempt = 0; attempt <= retries; attempt++) {
+  if (attempt > 0) {
+   const { promise, resolve } = Promise.withResolvers<void>();
+   setTimeout(resolve, retryDelay);
+   await promise;
+  }
+  const result = await probeRelayOnce(url, auth, opts);
+  if (result.ok || attempt === retries) {
+   return result;
+  }
+  opts?.onRetry?.(attempt + 1, retries, result);
+ }
+ return { ok: false, status: 0, latencyMs: 0, error: "Probe failed" };
+}