Fix model test routing for image providers

2026-09-23 04:09:49 +00:00 · 2026-06-02 21:14:36 +08:00
parent cce8a50cac
commit e414975d0c
7 changed files with 309 additions and 163 deletions
@@ -0,0 +1,129 @@
+import { getApiKeys } from "@/lib/localDb";
+import { UPDATER_CONFIG } from "@/shared/constants/config";
+import { getConsistentMachineId } from "@/shared/utils/machineId";
+
+const CLI_TOKEN_SALT = "9r-cli-auth";
+
+async function getInternalHeaders() {
+  let apiKey = null;
+  try {
+    const keys = await getApiKeys();
+    apiKey = keys.find((k) => k.isActive !== false)?.key || null;
+  } catch {}
+
+  const headers = { "Content-Type": "application/json" };
+  if (apiKey) headers["Authorization"] = `Bearer ${apiKey}`;
+  headers["x-9r-cli-token"] = await getConsistentMachineId(CLI_TOKEN_SALT);
+  return headers;
+}
+
+export async function pingModelByKind(model, kind, baseUrl = `http://127.0.0.1:${process.env.PORT || UPDATER_CONFIG.appPort}`) {
+  const headers = await getInternalHeaders();
+  const start = Date.now();
+
+  if (kind === "embedding") {
+    const res = await fetch(`${baseUrl}/api/v1/embeddings`, {
+      method: "POST",
+      headers,
+      body: JSON.stringify({ model, input: "test" }),
+      signal: AbortSignal.timeout(15000),
+    });
+    const latencyMs = Date.now() - start;
+    const rawText = await res.text().catch(() => "");
+    let parsed = null;
+    try { parsed = rawText ? JSON.parse(rawText) : null; } catch {}
+
+    if (!res.ok) {
+      const detail = parsed?.error?.message || parsed?.error || rawText;
+      return { ok: false, latencyMs, error: `HTTP ${res.status}${detail ? `: ${String(detail).slice(0, 240)}` : ""}`, status: res.status };
+    }
+    const hasEmbedding = Array.isArray(parsed?.data) && parsed.data.length > 0 && Array.isArray(parsed.data[0]?.embedding);
+    if (!hasEmbedding) {
+      return { ok: false, latencyMs, status: res.status, error: "Provider returned no embedding data" };
+    }
+    return { ok: true, latencyMs, error: null, status: res.status };
+  }
+
+  if (kind === "image") {
+    const res = await fetch(`${baseUrl}/api/v1/images/generations`, {
+      method: "POST",
+      headers,
+      body: JSON.stringify({ model, prompt: "test" }),
+      signal: AbortSignal.timeout(15000),
+    });
+    const latencyMs = Date.now() - start;
+    const rawText = await res.text().catch(() => "");
+    let parsed = null;
+    try { parsed = rawText ? JSON.parse(rawText) : null; } catch {}
+
+    if (!res.ok) {
+      const detail = parsed?.error?.message || parsed?.msg || parsed?.message || parsed?.error || rawText;
+      return { ok: false, latencyMs, error: `HTTP ${res.status}${detail ? `: ${String(detail).slice(0, 240)}` : ""}`, status: res.status };
+    }
+
+    const hasImages = Array.isArray(parsed?.data) && parsed.data.length > 0;
+    if (!hasImages) {
+      return { ok: false, latencyMs, status: res.status, error: "Provider returned no image data for this model" };
+    }
+    return { ok: true, latencyMs, error: null, status: res.status };
+  }
+
+  const res = await fetch(`${baseUrl}/api/v1/chat/completions`, {
+    method: "POST",
+    headers,
+    body: JSON.stringify({
+      model,
+      max_tokens: 1,
+      stream: false,
+      messages: [{ role: "user", content: "hi" }],
+    }),
+    signal: AbortSignal.timeout(15000),
+  });
+  const latencyMs = Date.now() - start;
+
+  const rawText = await res.text().catch(() => "");
+  let parsed = null;
+  try { parsed = rawText ? JSON.parse(rawText) : null; } catch {}
+
+  if (!res.ok) {
+    const detail = parsed?.error?.message || parsed?.msg || parsed?.message || parsed?.error || rawText;
+    return { ok: false, latencyMs, error: `HTTP ${res.status}${detail ? `: ${String(detail).slice(0, 240)}` : ""}`, status: res.status };
+  }
+
+  const providerStatus = parsed?.status;
+  const providerMsg = parsed?.msg || parsed?.message;
+  const hasProviderErrorStatus = providerStatus !== undefined
+    && providerStatus !== null
+    && String(providerStatus) !== "200"
+    && String(providerStatus) !== "0";
+  if (hasProviderErrorStatus && providerMsg) {
+    return {
+      ok: false,
+      latencyMs,
+      status: res.status,
+      error: `Provider status ${providerStatus}: ${String(providerMsg).slice(0, 240)}`,
+    };
+  }
+
+  if (parsed?.error) {
+    const providerError = parsed?.error?.message || parsed?.error || "Provider returned an error";
+    return {
+      ok: false,
+      latencyMs,
+      status: res.status,
+      error: String(providerError).slice(0, 240),
+    };
+  }
+
+  const hasChoices = Array.isArray(parsed?.choices) && parsed.choices.length > 0;
+  if (!hasChoices) {
+    return {
+      ok: false,
+      latencyMs,
+      status: res.status,
+      error: "Provider returned no completion choices for this model",
+    };
+  }
+
+  return { ok: true, latencyMs, error: null, status: res.status };
+}
@@ -1,119 +1,13 @@
 import { NextResponse } from "next/server";
-import { getApiKeys } from "@/lib/localDb";
-import { UPDATER_CONFIG } from "@/shared/constants/config";
-import { getConsistentMachineId } from "@/shared/utils/machineId";
-
-const CLI_TOKEN_SALT = "9r-cli-auth";
+import { pingModelByKind } from "./ping";

 // POST /api/models/test - Ping a single model via internal completions or embeddings
 export async function POST(request) {
  try {
    const { model, kind } = await request.json();
    if (!model) return NextResponse.json({ error: "Model required" }, { status: 400 });
-
-    const baseUrl = `http://127.0.0.1:${process.env.PORT || UPDATER_CONFIG.appPort}`;
-
-    // Get an active internal API key for auth (if requireApiKey is enabled)
-    let apiKey = null;
-    try {
-      const keys = await getApiKeys();
-      apiKey = keys.find((k) => k.isActive !== false)?.key || null;
-    } catch {}
-
-    const headers = { "Content-Type": "application/json" };
-    if (apiKey) headers["Authorization"] = `Bearer ${apiKey}`;
-    // Bypass dashboardGuard for internal self-call via CLI token (machineId-based)
-    headers["x-9r-cli-token"] = await getConsistentMachineId(CLI_TOKEN_SALT);
-
-    const start = Date.now();
-
-    // Route to appropriate endpoint based on kind
-    if (kind === "embedding") {
-      const res = await fetch(`${baseUrl}/api/v1/embeddings`, {
-        method: "POST",
-        headers,
-        body: JSON.stringify({ model, input: "test" }),
-        signal: AbortSignal.timeout(15000),
-      });
-      const latencyMs = Date.now() - start;
-      const rawText = await res.text().catch(() => "");
-      let parsed = null;
-      try { parsed = rawText ? JSON.parse(rawText) : null; } catch {}
-
-      if (!res.ok) {
-        const detail = parsed?.error?.message || parsed?.error || rawText;
-        return NextResponse.json({ ok: false, latencyMs, error: `HTTP ${res.status}${detail ? `: ${String(detail).slice(0, 240)}` : ""}`, status: res.status });
-      }
-      const hasEmbedding = Array.isArray(parsed?.data) && parsed.data.length > 0 && Array.isArray(parsed.data[0]?.embedding);
-      if (!hasEmbedding) {
-        return NextResponse.json({ ok: false, latencyMs, status: res.status, error: "Provider returned no embedding data" });
-      }
-      return NextResponse.json({ ok: true, latencyMs, error: null, status: res.status });
-    }
-
-    // Default: chat completions
-    const res = await fetch(`${baseUrl}/api/v1/chat/completions`, {
-      method: "POST",
-      headers,
-      body: JSON.stringify({
-        model,
-        max_tokens: 1,
-        stream: false,
-        messages: [{ role: "user", content: "hi" }],
-      }),
-      signal: AbortSignal.timeout(15000),
-    });
-    const latencyMs = Date.now() - start;
-
-    const rawText = await res.text().catch(() => "");
-    let parsed = null;
-    try {
-      parsed = rawText ? JSON.parse(rawText) : null;
-    } catch {}
-
-    if (!res.ok) {
-      const detail = parsed?.error?.message || parsed?.msg || parsed?.message || parsed?.error || rawText;
-      const error = `HTTP ${res.status}${detail ? `: ${String(detail).slice(0, 240)}` : ""}`;
-      return NextResponse.json({ ok: false, latencyMs, error, status: res.status });
-    }
-
-    // Some providers may return HTTP 200 but not a real completion for invalid models.
-    const providerStatus = parsed?.status;
-    const providerMsg = parsed?.msg || parsed?.message;
-    const hasProviderErrorStatus = providerStatus !== undefined
-      && providerStatus !== null
-      && String(providerStatus) !== "200"
-      && String(providerStatus) !== "0";
-    if (hasProviderErrorStatus && providerMsg) {
-      return NextResponse.json({
-        ok: false,
-        latencyMs,
-        status: res.status,
-        error: `Provider status ${providerStatus}: ${String(providerMsg).slice(0, 240)}`,
-      });
-    }
-
-    if (parsed?.error) {
-      const providerError = parsed?.error?.message || parsed?.error || "Provider returned an error";
-      return NextResponse.json({
-        ok: false,
-        latencyMs,
-        status: res.status,
-        error: String(providerError).slice(0, 240),
-      });
-    }
-
-    const hasChoices = Array.isArray(parsed?.choices) && parsed.choices.length > 0;
-    if (!hasChoices) {
-      return NextResponse.json({
-        ok: false,
-        latencyMs,
-        status: res.status,
-        error: "Provider returned no completion choices for this model",
-      });
-    }
-
-    return NextResponse.json({ ok: true, latencyMs, error: null, status: res.status });
+    const result = await pingModelByKind(model, kind || "llm");
+    return NextResponse.json(result);
  } catch (err) {
    return NextResponse.json({ ok: false, error: err.message }, { status: 500 });
  }
@@ -1,59 +1,14 @@
 import { NextResponse } from "next/server";
-import { getProviderConnectionById, getApiKeys } from "@/lib/localDb";
+import { getProviderConnectionById } from "@/lib/localDb";
 import { getProviderModels, PROVIDER_ID_TO_ALIAS } from "open-sse/config/providerModels.js";
 import { isOpenAICompatibleProvider, isAnthropicCompatibleProvider } from "@/shared/constants/providers";
 import { UPDATER_CONFIG } from "@/shared/constants/config";
-import { getConsistentMachineId } from "@/shared/utils/machineId";
-
-const CLI_TOKEN_SALT = "9r-cli-auth";
-
-/**
- * Get an active API key to pass through auth when requireApiKey is enabled.
- */
-async function getInternalApiKey() {
-  const keys = await getApiKeys();
-  return keys.find((k) => k.isActive !== false)?.key || null;
-}
-
-/**
- * Ping a single model via internal completions endpoint (OpenAI format).
- * open-sse handles all provider translation automatically.
- */
-async function pingModel(modelId, baseUrl, apiKey, cliToken) {
-  const start = Date.now();
-  try {
-    const headers = { "Content-Type": "application/json" };
-    if (apiKey) headers["Authorization"] = `Bearer ${apiKey}`;
-    if (cliToken) headers["x-9r-cli-token"] = cliToken;
-    const res = await fetch(`${baseUrl}/api/v1/chat/completions`, {
-      method: "POST",
-      headers,
-      body: JSON.stringify({
-        model: modelId,
-        max_tokens: 1,
-        stream: false,
-        messages: [{ role: "user", content: "hi" }],
-      }),
-      signal: AbortSignal.timeout(15000),
-    });
-    const latencyMs = Date.now() - start;
-    // 200 = working; 400 = bad request but auth passed (model reachable)
-    const ok = res.status === 200 || res.status === 400;
-    let error = null;
-    if (!ok) {
-      const text = await res.text().catch(() => "");
-      error = `HTTP ${res.status}${text ? `: ${text.slice(0, 120)}` : ""}`;
-    }
-    return { ok, latencyMs, error };
-  } catch (err) {
-    return { ok: false, latencyMs: Date.now() - start, error: err.message };
-  }
-}
+import { pingModelByKind } from "@/app/api/models/test/ping";

 /**
 * POST /api/providers/[id]/test-models
 * id = connectionId — used only to resolve provider + model list.
- * Actual requests go through /api/v1/chat/completions (open-sse handles everything).
+ * Actual requests go through the internal endpoint that matches each model kind.
 */
 export async function POST(request, { params }) {
  try {
@@ -86,20 +41,17 @@ export async function POST(request, { params }) {
      return NextResponse.json({ error: "No models configured for this provider" }, { status: 400 });
    }

-    const apiKey = await getInternalApiKey();
-    // Bypass dashboardGuard for internal self-call via CLI token (machineId-based)
-    const cliToken = await getConsistentMachineId(CLI_TOKEN_SALT);
-
    // Warm up with first model to trigger token refresh (if needed) before parallel calls.
    // This prevents race condition where multiple requests concurrently refresh the same token.
    const [first, ...rest] = models;
-    const firstResult = await pingModel(`${alias}/${first.id}`, baseUrl, apiKey, cliToken);
+    const firstKind = first.type || "llm";
+    const firstResult = await pingModelByKind(`${alias}/${first.id}`, firstKind, baseUrl);
    const results = [{ modelId: first.id, name: first.name || first.id, ...firstResult }];

    if (rest.length > 0) {
      const restResults = await Promise.all(
        rest.map(async (model) => {
-          const result = await pingModel(`${alias}/${model.id}`, baseUrl, apiKey, cliToken);
+          const result = await pingModelByKind(`${alias}/${model.id}`, model.type || "llm", baseUrl);
          return { modelId: model.id, name: model.name || model.id, ...result };
        })
      );