feat(proxy): pass account proxy through all proxy route handlers

Thread account-level proxyUrl from extraConfig into
resolveChannelProxyUrl for endpointFlow and the third parameter
of withSiteRecordProxyRequestInit for direct fetch calls across
chat, responses, completions, embeddings, images, search, and
videos proxy routes.

src/server/routes/proxy/chat.ts

@@ -8,7 +8,8 @@ import { isTokenExpiredError } from '../../services/alertRules.js';
 import { shouldRetryProxyRequest } from '../../services/proxyRetryPolicy.js';
 import { resolveProxyUsageWithSelfLogFallback } from '../../services/proxyUsageFallbackService.js';
 import { mergeProxyUsage, parseProxyUsage } from '../../services/proxyUsageParser.js';
-import { resolveProxyUrlForSite, withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
+import { resolveChannelProxyUrl, withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
+import { getProxyUrlFromExtraConfig } from '../../services/accountExtraConfig.js';
 import { type DownstreamFormat } from '../../transformers/shared/normalized.js';
 import {
   buildUpstreamEndpointRequest,
@@ -168,7 +169,7 @@ async function handleChatProxyRequest(
           method: 'POST',
           headers: compatibilityRequest.headers,
           body: JSON.stringify(compatibilityRequest.body),
-        }),
+        }, getProxyUrlFromExtraConfig(selected.account.extraConfig)),
       ),
     });
     let startTime = Date.now();
@@ -176,7 +177,7 @@ async function handleChatProxyRequest(
     try {
       const endpointResult = await executeEndpointFlow({
         siteUrl: selected.site.url,
-        proxyUrl: resolveProxyUrlForSite(selected.site),
+        proxyUrl: resolveChannelProxyUrl(selected.site, selected.account.extraConfig),
         endpointCandidates,
         buildRequest: (endpoint) => buildEndpointRequest(endpoint),
         tryRecover: endpointStrategy.tryRecover,

src/server/routes/proxy/completions.ts

@@ -1,45 +1,46 @@
 import { FastifyInstance, FastifyReply, FastifyRequest } from 'fastify';
 import { fetch } from 'undici';
 import { db, schema } from '../../db/index.js';
-import { tokenRouter } from '../../services/tokenRouter.js';
+import { tokenRouter } from '../../services/tokenRouter.js';
-import { refreshModelsAndRebuildRoutes } from '../../services/modelService.js';
+import { refreshModelsAndRebuildRoutes } from '../../services/modelService.js';
-import { reportProxyAllFailed, reportTokenExpired } from '../../services/alertService.js';
+import { reportProxyAllFailed, reportTokenExpired } from '../../services/alertService.js';
-import { isTokenExpiredError } from '../../services/alertRules.js';
+import { isTokenExpiredError } from '../../services/alertRules.js';
-import { shouldRetryProxyRequest } from '../../services/proxyRetryPolicy.js';
+import { shouldRetryProxyRequest } from '../../services/proxyRetryPolicy.js';
-import { resolveProxyUsageWithSelfLogFallback } from '../../services/proxyUsageFallbackService.js';
+import { resolveProxyUsageWithSelfLogFallback } from '../../services/proxyUsageFallbackService.js';
-import { mergeProxyUsage, parseProxyUsage, pullSseDataEvents } from '../../services/proxyUsageParser.js';
+import { mergeProxyUsage, parseProxyUsage, pullSseDataEvents } from '../../services/proxyUsageParser.js';
-import { ensureModelAllowedForDownstreamKey, getDownstreamRoutingPolicy, recordDownstreamCostUsage } from './downstreamPolicy.js';
+import { ensureModelAllowedForDownstreamKey, getDownstreamRoutingPolicy, recordDownstreamCostUsage } from './downstreamPolicy.js';
-import { withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
+import { withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
-import { composeProxyLogMessage } from './logPathMeta.js';
+import { getProxyUrlFromExtraConfig } from '../../services/accountExtraConfig.js';
-import { formatUtcSqlDateTime } from '../../services/localTimeService.js';
+import { composeProxyLogMessage } from './logPathMeta.js';
-import { resolveProxyLogBilling } from './proxyBilling.js';
+import { formatUtcSqlDateTime } from '../../services/localTimeService.js';
-import { buildUpstreamUrl } from './upstreamUrl.js';
+import { resolveProxyLogBilling } from './proxyBilling.js';
+import { buildUpstreamUrl } from './upstreamUrl.js';
 
 const MAX_RETRIES = 2;
 
 export async function completionsProxyRoute(app: FastifyInstance) {
   app.post('/v1/completions', async (request: FastifyRequest, reply: FastifyReply) => {
     const body = request.body as any;
-    const requestedModel = body?.model;
+    const requestedModel = body?.model;
-    if (!requestedModel) {
+    if (!requestedModel) {
-      return reply.code(400).send({ error: { message: 'model is required', type: 'invalid_request_error' } });
+      return reply.code(400).send({ error: { message: 'model is required', type: 'invalid_request_error' } });
-    }
+    }
-    if (!await ensureModelAllowedForDownstreamKey(request, reply, requestedModel)) return;
+    if (!await ensureModelAllowedForDownstreamKey(request, reply, requestedModel)) return;
-    const downstreamPolicy = getDownstreamRoutingPolicy(request);
+    const downstreamPolicy = getDownstreamRoutingPolicy(request);
-
+
-    const isStream = body.stream === true;
+    const isStream = body.stream === true;
     const excludeChannelIds: number[] = [];
     let retryCount = 0;
 
-    while (retryCount <= MAX_RETRIES) {
+    while (retryCount <= MAX_RETRIES) {
-      let selected = retryCount === 0
+      let selected = retryCount === 0
-        ? await tokenRouter.selectChannel(requestedModel, downstreamPolicy)
+        ? await tokenRouter.selectChannel(requestedModel, downstreamPolicy)
-        : await tokenRouter.selectNextChannel(requestedModel, excludeChannelIds, downstreamPolicy);
+        : await tokenRouter.selectNextChannel(requestedModel, excludeChannelIds, downstreamPolicy);
-
+
-      if (!selected && retryCount === 0) {
+      if (!selected && retryCount === 0) {
-        await refreshModelsAndRebuildRoutes();
+        await refreshModelsAndRebuildRoutes();
-        selected = await tokenRouter.selectChannel(requestedModel, downstreamPolicy);
+        selected = await tokenRouter.selectChannel(requestedModel, downstreamPolicy);
-      }
+      }
 
       if (!selected) {
         await reportProxyAllFailed({
@@ -51,21 +52,21 @@ export async function completionsProxyRoute(app: FastifyInstance) {
         });
       }
 
-      excludeChannelIds.push(selected.channel.id);
+      excludeChannelIds.push(selected.channel.id);
-
+
-      const targetUrl = buildUpstreamUrl(selected.site.url, '/v1/completions');
+      const targetUrl = buildUpstreamUrl(selected.site.url, '/v1/completions');
-      const forwardBody = { ...body, model: selected.actualModel };
+      const forwardBody = { ...body, model: selected.actualModel };
-      const startTime = Date.now();
+      const startTime = Date.now();
 
       try {
-        const upstream = await fetch(targetUrl, withSiteRecordProxyRequestInit(selected.site, {
+        const upstream = await fetch(targetUrl, withSiteRecordProxyRequestInit(selected.site, {
-          method: 'POST',
+          method: 'POST',
-          headers: {
+          headers: {
-            'Content-Type': 'application/json',
+            'Content-Type': 'application/json',
-            'Authorization': `Bearer ${selected.tokenValue}`,
+            'Authorization': `Bearer ${selected.tokenValue}`,
-          },
+          },
-          body: JSON.stringify(forwardBody),
+          body: JSON.stringify(forwardBody),
-        }));
+        }, getProxyUrlFromExtraConfig(selected.account.extraConfig)));
 
         if (!upstream.ok) {
           const errText = await upstream.text().catch(() => 'unknown error');
@@ -81,10 +82,10 @@ export async function completionsProxyRoute(app: FastifyInstance) {
             });
           }
 
-          if (shouldRetryProxyRequest(upstream.status, errText) && retryCount < MAX_RETRIES) {
+          if (shouldRetryProxyRequest(upstream.status, errText) && retryCount < MAX_RETRIES) {
-            retryCount++;
+            retryCount++;
-            continue;
+            continue;
-          }
+          }
 
           await reportProxyAllFailed({
             model: requestedModel,
@@ -106,110 +107,110 @@ export async function completionsProxyRoute(app: FastifyInstance) {
             return;
           }
 
-          const decoder = new TextDecoder();
+          const decoder = new TextDecoder();
-          let parsedUsage: ReturnType<typeof parseProxyUsage> = {
+          let parsedUsage: ReturnType<typeof parseProxyUsage> = {
-            promptTokens: 0,
+            promptTokens: 0,
-            completionTokens: 0,
+            completionTokens: 0,
-            totalTokens: 0,
+            totalTokens: 0,
-            cacheReadTokens: 0,
+            cacheReadTokens: 0,
-            cacheCreationTokens: 0,
+            cacheCreationTokens: 0,
-            promptTokensIncludeCache: null,
+            promptTokensIncludeCache: null,
-          };
+          };
-          let sseBuffer = '';
+          let sseBuffer = '';
-          try {
+          try {
-            while (true) {
+            while (true) {
-              const { done, value } = await reader.read();
+              const { done, value } = await reader.read();
-              if (done) break;
+              if (done) break;
-              const chunk = decoder.decode(value, { stream: true });
+              const chunk = decoder.decode(value, { stream: true });
-              reply.raw.write(chunk);
+              reply.raw.write(chunk);
-
-              sseBuffer += chunk;
-              const pulled = pullSseDataEvents(sseBuffer);
-              sseBuffer = pulled.rest;
-              for (const eventPayload of pulled.events) {
-                try {
-                  parsedUsage = mergeProxyUsage(parsedUsage, parseProxyUsage(JSON.parse(eventPayload)));
-                } catch {}
-              }
-            }
-            if (sseBuffer.trim().length > 0) {
-              const pulled = pullSseDataEvents(`${sseBuffer}\n\n`);
-              for (const eventPayload of pulled.events) {
-                try {
-                  parsedUsage = mergeProxyUsage(parsedUsage, parseProxyUsage(JSON.parse(eventPayload)));
-                } catch {}
-              }
-            }
-          } finally {
-            reader.releaseLock();
-            reply.raw.end();
-          }
 
-          const latency = Date.now() - startTime;
+              sseBuffer += chunk;
-          const resolvedUsage = await resolveProxyUsageWithSelfLogFallback({
+              const pulled = pullSseDataEvents(sseBuffer);
-            site: selected.site,
+              sseBuffer = pulled.rest;
-            account: selected.account,
+              for (const eventPayload of pulled.events) {
-            tokenValue: selected.tokenValue,
+                try {
-            tokenName: selected.tokenName,
+                  parsedUsage = mergeProxyUsage(parsedUsage, parseProxyUsage(JSON.parse(eventPayload)));
-            modelName: selected.actualModel || requestedModel,
+                } catch {}
-            requestStartedAtMs: startTime,
+              }
-            requestEndedAtMs: startTime + latency,
+            }
-            localLatencyMs: latency,
+            if (sseBuffer.trim().length > 0) {
-            usage: {
+              const pulled = pullSseDataEvents(`${sseBuffer}\n\n`);
-              promptTokens: parsedUsage.promptTokens,
+              for (const eventPayload of pulled.events) {
-              completionTokens: parsedUsage.completionTokens,
+                try {
-              totalTokens: parsedUsage.totalTokens,
+                  parsedUsage = mergeProxyUsage(parsedUsage, parseProxyUsage(JSON.parse(eventPayload)));
-            },
+                } catch {}
-          });
+              }
-          const { estimatedCost, billingDetails } = await resolveProxyLogBilling({
+            }
-            site: selected.site,
+          } finally {
-            account: selected.account,
+            reader.releaseLock();
-            modelName: selected.actualModel || requestedModel,
+            reply.raw.end();
-            parsedUsage,
+          }
-            resolvedUsage,
+
-          });
+          const latency = Date.now() - startTime;
-          tokenRouter.recordSuccess(selected.channel.id, latency, estimatedCost);
+          const resolvedUsage = await resolveProxyUsageWithSelfLogFallback({
-          recordDownstreamCostUsage(request, estimatedCost);
+            site: selected.site,
-          logProxy(
+            account: selected.account,
-            selected, requestedModel, 'success', 200, latency, null, retryCount,
+            tokenValue: selected.tokenValue,
-            resolvedUsage.promptTokens, resolvedUsage.completionTokens, resolvedUsage.totalTokens, estimatedCost, billingDetails,
+            tokenName: selected.tokenName,
-          );
+            modelName: selected.actualModel || requestedModel,
-          return;
+            requestStartedAtMs: startTime,
-        }
+            requestEndedAtMs: startTime + latency,
-
+            localLatencyMs: latency,
-        const data = await upstream.json() as any;
+            usage: {
-        const latency = Date.now() - startTime;
+              promptTokens: parsedUsage.promptTokens,
-        const parsedUsage = parseProxyUsage(data);
+              completionTokens: parsedUsage.completionTokens,
-        const resolvedUsage = await resolveProxyUsageWithSelfLogFallback({
+              totalTokens: parsedUsage.totalTokens,
-          site: selected.site,
+            },
-          account: selected.account,
+          });
-          tokenValue: selected.tokenValue,
+          const { estimatedCost, billingDetails } = await resolveProxyLogBilling({
-          tokenName: selected.tokenName,
+            site: selected.site,
-          modelName: selected.actualModel || requestedModel,
+            account: selected.account,
-          requestStartedAtMs: startTime,
+            modelName: selected.actualModel || requestedModel,
-          requestEndedAtMs: startTime + latency,
+            parsedUsage,
-          localLatencyMs: latency,
+            resolvedUsage,
-          usage: {
+          });
-            promptTokens: parsedUsage.promptTokens,
+          tokenRouter.recordSuccess(selected.channel.id, latency, estimatedCost);
-            completionTokens: parsedUsage.completionTokens,
+          recordDownstreamCostUsage(request, estimatedCost);
-            totalTokens: parsedUsage.totalTokens,
+          logProxy(
-          },
+            selected, requestedModel, 'success', 200, latency, null, retryCount,
-        });
+            resolvedUsage.promptTokens, resolvedUsage.completionTokens, resolvedUsage.totalTokens, estimatedCost, billingDetails,
-        const { estimatedCost, billingDetails } = await resolveProxyLogBilling({
+          );
-          site: selected.site,
+          return;
-          account: selected.account,
+        }
-          modelName: selected.actualModel || requestedModel,
+
-          parsedUsage,
+        const data = await upstream.json() as any;
-          resolvedUsage,
+        const latency = Date.now() - startTime;
-        });
+        const parsedUsage = parseProxyUsage(data);
-
+        const resolvedUsage = await resolveProxyUsageWithSelfLogFallback({
-        tokenRouter.recordSuccess(selected.channel.id, latency, estimatedCost);
+          site: selected.site,
-        recordDownstreamCostUsage(request, estimatedCost);
+          account: selected.account,
-        logProxy(
+          tokenValue: selected.tokenValue,
-          selected, requestedModel, 'success', 200, latency, null, retryCount,
+          tokenName: selected.tokenName,
-          resolvedUsage.promptTokens, resolvedUsage.completionTokens, resolvedUsage.totalTokens, estimatedCost, billingDetails,
+          modelName: selected.actualModel || requestedModel,
-        );
+          requestStartedAtMs: startTime,
-        return reply.send(data);
+          requestEndedAtMs: startTime + latency,
+          localLatencyMs: latency,
+          usage: {
+            promptTokens: parsedUsage.promptTokens,
+            completionTokens: parsedUsage.completionTokens,
+            totalTokens: parsedUsage.totalTokens,
+          },
+        });
+        const { estimatedCost, billingDetails } = await resolveProxyLogBilling({
+          site: selected.site,
+          account: selected.account,
+          modelName: selected.actualModel || requestedModel,
+          parsedUsage,
+          resolvedUsage,
+        });
+
+        tokenRouter.recordSuccess(selected.channel.id, latency, estimatedCost);
+        recordDownstreamCostUsage(request, estimatedCost);
+        logProxy(
+          selected, requestedModel, 'success', 200, latency, null, retryCount,
+          resolvedUsage.promptTokens, resolvedUsage.completionTokens, resolvedUsage.totalTokens, estimatedCost, billingDetails,
+        );
+        return reply.send(data);
       } catch (err: any) {
         tokenRouter.recordFailure(selected.channel.id);
         logProxy(selected, requestedModel, 'failed', 0, Date.now() - startTime, err.message, retryCount);
@@ -237,38 +238,38 @@ async function logProxy(
   latencyMs: number,
   errorMessage: string | null,
   retryCount: number,
-  promptTokens = 0,
+  promptTokens = 0,
-  completionTokens = 0,
+  completionTokens = 0,
-  totalTokens = 0,
+  totalTokens = 0,
-  estimatedCost = 0,
+  estimatedCost = 0,
-  billingDetails: unknown = null,
+  billingDetails: unknown = null,
-) {
+) {
-  try {
+  try {
-    const createdAt = formatUtcSqlDateTime(new Date());
+    const createdAt = formatUtcSqlDateTime(new Date());
-    const normalizedErrorMessage = composeProxyLogMessage({
+    const normalizedErrorMessage = composeProxyLogMessage({
-      downstreamPath: '/v1/completions',
+      downstreamPath: '/v1/completions',
-      errorMessage,
+      errorMessage,
-    });
+    });
-    await db.insert(schema.proxyLogs).values({
+    await db.insert(schema.proxyLogs).values({
-      routeId: selected.channel.routeId,
+      routeId: selected.channel.routeId,
-      channelId: selected.channel.id,
+      channelId: selected.channel.id,
-      accountId: selected.account.id,
+      accountId: selected.account.id,
-      modelRequested,
+      modelRequested,
-      modelActual: selected.actualModel,
+      modelActual: selected.actualModel,
-      status,
+      status,
-      httpStatus,
+      httpStatus,
-      latencyMs,
+      latencyMs,
-      promptTokens,
+      promptTokens,
-      completionTokens,
+      completionTokens,
-      totalTokens,
+      totalTokens,
-      estimatedCost,
+      estimatedCost,
-      billingDetails: billingDetails ? JSON.stringify(billingDetails) : null,
+      billingDetails: billingDetails ? JSON.stringify(billingDetails) : null,
-      errorMessage: normalizedErrorMessage,
+      errorMessage: normalizedErrorMessage,
-      retryCount,
+      retryCount,
-      createdAt,
+      createdAt,
-    }).run();
+    }).run();
-  } catch (error) {
+  } catch (error) {
-    console.warn('[proxy/completions] failed to write proxy log', error);
+    console.warn('[proxy/completions] failed to write proxy log', error);
-  }
+  }
-}
+}
 

src/server/routes/proxy/embeddings.ts

@@ -1,44 +1,45 @@
 import { FastifyInstance, FastifyReply, FastifyRequest } from 'fastify';
 import { fetch } from 'undici';
 import { db, schema } from '../../db/index.js';
-import { tokenRouter } from '../../services/tokenRouter.js';
+import { tokenRouter } from '../../services/tokenRouter.js';
-import { refreshModelsAndRebuildRoutes } from '../../services/modelService.js';
+import { refreshModelsAndRebuildRoutes } from '../../services/modelService.js';
-import { reportProxyAllFailed, reportTokenExpired } from '../../services/alertService.js';
+import { reportProxyAllFailed, reportTokenExpired } from '../../services/alertService.js';
-import { isTokenExpiredError } from '../../services/alertRules.js';
+import { isTokenExpiredError } from '../../services/alertRules.js';
-import { shouldRetryProxyRequest } from '../../services/proxyRetryPolicy.js';
+import { shouldRetryProxyRequest } from '../../services/proxyRetryPolicy.js';
-import { resolveProxyUsageWithSelfLogFallback } from '../../services/proxyUsageFallbackService.js';
+import { resolveProxyUsageWithSelfLogFallback } from '../../services/proxyUsageFallbackService.js';
-import { parseProxyUsage } from '../../services/proxyUsageParser.js';
+import { parseProxyUsage } from '../../services/proxyUsageParser.js';
-import { ensureModelAllowedForDownstreamKey, getDownstreamRoutingPolicy, recordDownstreamCostUsage } from './downstreamPolicy.js';
+import { ensureModelAllowedForDownstreamKey, getDownstreamRoutingPolicy, recordDownstreamCostUsage } from './downstreamPolicy.js';
-import { withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
+import { withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
-import { composeProxyLogMessage } from './logPathMeta.js';
+import { getProxyUrlFromExtraConfig } from '../../services/accountExtraConfig.js';
-import { formatUtcSqlDateTime } from '../../services/localTimeService.js';
+import { composeProxyLogMessage } from './logPathMeta.js';
-import { resolveProxyLogBilling } from './proxyBilling.js';
+import { formatUtcSqlDateTime } from '../../services/localTimeService.js';
-import { buildUpstreamUrl } from './upstreamUrl.js';
+import { resolveProxyLogBilling } from './proxyBilling.js';
+import { buildUpstreamUrl } from './upstreamUrl.js';
 
 const MAX_RETRIES = 2;
 
 export async function embeddingsProxyRoute(app: FastifyInstance) {
   app.post('/v1/embeddings', async (request: FastifyRequest, reply: FastifyReply) => {
     const body = request.body as any;
-    const requestedModel = body?.model;
+    const requestedModel = body?.model;
-    if (!requestedModel) {
+    if (!requestedModel) {
-      return reply.code(400).send({ error: { message: 'model is required', type: 'invalid_request_error' } });
+      return reply.code(400).send({ error: { message: 'model is required', type: 'invalid_request_error' } });
-    }
+    }
-    if (!await ensureModelAllowedForDownstreamKey(request, reply, requestedModel)) return;
+    if (!await ensureModelAllowedForDownstreamKey(request, reply, requestedModel)) return;
-    const downstreamPolicy = getDownstreamRoutingPolicy(request);
+    const downstreamPolicy = getDownstreamRoutingPolicy(request);
-
+
-    const excludeChannelIds: number[] = [];
+    const excludeChannelIds: number[] = [];
     let retryCount = 0;
 
-    while (retryCount <= MAX_RETRIES) {
+    while (retryCount <= MAX_RETRIES) {
-      let selected = retryCount === 0
+      let selected = retryCount === 0
-        ? await tokenRouter.selectChannel(requestedModel, downstreamPolicy)
+        ? await tokenRouter.selectChannel(requestedModel, downstreamPolicy)
-        : await tokenRouter.selectNextChannel(requestedModel, excludeChannelIds, downstreamPolicy);
+        : await tokenRouter.selectNextChannel(requestedModel, excludeChannelIds, downstreamPolicy);
-
+
-      if (!selected && retryCount === 0) {
+      if (!selected && retryCount === 0) {
-        await refreshModelsAndRebuildRoutes();
+        await refreshModelsAndRebuildRoutes();
-        selected = await tokenRouter.selectChannel(requestedModel, downstreamPolicy);
+        selected = await tokenRouter.selectChannel(requestedModel, downstreamPolicy);
-      }
+      }
 
       if (!selected) {
         await reportProxyAllFailed({
@@ -48,21 +49,21 @@ export async function embeddingsProxyRoute(app: FastifyInstance) {
         return reply.code(503).send({ error: { message: 'No available channels', type: 'server_error' } });
       }
 
-      excludeChannelIds.push(selected.channel.id);
+      excludeChannelIds.push(selected.channel.id);
-
+
-      const targetUrl = buildUpstreamUrl(selected.site.url, '/v1/embeddings');
+      const targetUrl = buildUpstreamUrl(selected.site.url, '/v1/embeddings');
-      const forwardBody = { ...body, model: selected.actualModel };
+      const forwardBody = { ...body, model: selected.actualModel };
-      const startTime = Date.now();
+      const startTime = Date.now();
 
       try {
-        const upstream = await fetch(targetUrl, withSiteRecordProxyRequestInit(selected.site, {
+        const upstream = await fetch(targetUrl, withSiteRecordProxyRequestInit(selected.site, {
-          method: 'POST',
+          method: 'POST',
-          headers: {
+          headers: {
-            'Content-Type': 'application/json',
+            'Content-Type': 'application/json',
-            'Authorization': `Bearer ${selected.tokenValue}`,
+            'Authorization': `Bearer ${selected.tokenValue}`,
-          },
+          },
-          body: JSON.stringify(forwardBody),
+          body: JSON.stringify(forwardBody),
-        }));
+        }, getProxyUrlFromExtraConfig(selected.account.extraConfig)));
 
         const text = await upstream.text();
         if (!upstream.ok) {
@@ -78,10 +79,10 @@ export async function embeddingsProxyRoute(app: FastifyInstance) {
             });
           }
 
-          if (shouldRetryProxyRequest(upstream.status, text) && retryCount < MAX_RETRIES) {
+          if (shouldRetryProxyRequest(upstream.status, text) && retryCount < MAX_RETRIES) {
-            retryCount++;
+            retryCount++;
-            continue;
+            continue;
-          }
+          }
 
           await reportProxyAllFailed({
             model: requestedModel,
@@ -90,40 +91,40 @@ export async function embeddingsProxyRoute(app: FastifyInstance) {
           return reply.code(upstream.status).send({ error: { message: text, type: 'upstream_error' } });
         }
 
-        let data: any = {};
+        let data: any = {};
-        try { data = JSON.parse(text); } catch { data = {}; }
+        try { data = JSON.parse(text); } catch { data = {}; }
-        const latency = Date.now() - startTime;
+        const latency = Date.now() - startTime;
-        const parsedUsage = parseProxyUsage(data);
+        const parsedUsage = parseProxyUsage(data);
-        const resolvedUsage = await resolveProxyUsageWithSelfLogFallback({
+        const resolvedUsage = await resolveProxyUsageWithSelfLogFallback({
-          site: selected.site,
+          site: selected.site,
-          account: selected.account,
+          account: selected.account,
-          tokenValue: selected.tokenValue,
+          tokenValue: selected.tokenValue,
-          tokenName: selected.tokenName,
+          tokenName: selected.tokenName,
-          modelName: selected.actualModel || requestedModel,
+          modelName: selected.actualModel || requestedModel,
-          requestStartedAtMs: startTime,
+          requestStartedAtMs: startTime,
-          requestEndedAtMs: startTime + latency,
+          requestEndedAtMs: startTime + latency,
-          localLatencyMs: latency,
+          localLatencyMs: latency,
-          usage: {
+          usage: {
-            promptTokens: parsedUsage.promptTokens,
+            promptTokens: parsedUsage.promptTokens,
-            completionTokens: parsedUsage.completionTokens,
+            completionTokens: parsedUsage.completionTokens,
-            totalTokens: parsedUsage.totalTokens,
+            totalTokens: parsedUsage.totalTokens,
-          },
+          },
-        });
+        });
-        const { estimatedCost, billingDetails } = await resolveProxyLogBilling({
+        const { estimatedCost, billingDetails } = await resolveProxyLogBilling({
-          site: selected.site,
+          site: selected.site,
-          account: selected.account,
+          account: selected.account,
-          modelName: selected.actualModel || requestedModel,
+          modelName: selected.actualModel || requestedModel,
-          parsedUsage,
+          parsedUsage,
-          resolvedUsage,
+          resolvedUsage,
-        });
+        });
-
+
-        tokenRouter.recordSuccess(selected.channel.id, latency, estimatedCost);
+        tokenRouter.recordSuccess(selected.channel.id, latency, estimatedCost);
-        recordDownstreamCostUsage(request, estimatedCost);
+        recordDownstreamCostUsage(request, estimatedCost);
-        logProxy(
+        logProxy(
-          selected, requestedModel, 'success', upstream.status, latency, null, retryCount,
+          selected, requestedModel, 'success', upstream.status, latency, null, retryCount,
-          resolvedUsage.promptTokens, resolvedUsage.completionTokens, resolvedUsage.totalTokens, estimatedCost, billingDetails,
+          resolvedUsage.promptTokens, resolvedUsage.completionTokens, resolvedUsage.totalTokens, estimatedCost, billingDetails,
-        );
+        );
-        return reply.code(upstream.status).send(data);
+        return reply.code(upstream.status).send(data);
       } catch (err: any) {
         tokenRouter.recordFailure(selected.channel.id);
         logProxy(selected, requestedModel, 'failed', 0, Date.now() - startTime, err.message, retryCount);
@@ -149,38 +150,38 @@ async function logProxy(
   latencyMs: number,
   errorMessage: string | null,
   retryCount: number,
-  promptTokens = 0,
+  promptTokens = 0,
-  completionTokens = 0,
+  completionTokens = 0,
-  totalTokens = 0,
+  totalTokens = 0,
-  estimatedCost = 0,
+  estimatedCost = 0,
-  billingDetails: unknown = null,
+  billingDetails: unknown = null,
-) {
+) {
-  try {
+  try {
-    const createdAt = formatUtcSqlDateTime(new Date());
+    const createdAt = formatUtcSqlDateTime(new Date());
-    const normalizedErrorMessage = composeProxyLogMessage({
+    const normalizedErrorMessage = composeProxyLogMessage({
-      downstreamPath: '/v1/embeddings',
+      downstreamPath: '/v1/embeddings',
-      errorMessage,
+      errorMessage,
-    });
+    });
-    await db.insert(schema.proxyLogs).values({
+    await db.insert(schema.proxyLogs).values({
-      routeId: selected.channel.routeId,
+      routeId: selected.channel.routeId,
-      channelId: selected.channel.id,
+      channelId: selected.channel.id,
-      accountId: selected.account.id,
+      accountId: selected.account.id,
-      modelRequested,
+      modelRequested,
-      modelActual: selected.actualModel,
+      modelActual: selected.actualModel,
-      status,
+      status,
-      httpStatus,
+      httpStatus,
-      latencyMs,
+      latencyMs,
-      promptTokens,
+      promptTokens,
-      completionTokens,
+      completionTokens,
-      totalTokens,
+      totalTokens,
-      estimatedCost,
+      estimatedCost,
-      billingDetails: billingDetails ? JSON.stringify(billingDetails) : null,
+      billingDetails: billingDetails ? JSON.stringify(billingDetails) : null,
-      errorMessage: normalizedErrorMessage,
+      errorMessage: normalizedErrorMessage,
-      retryCount,
+      retryCount,
-      createdAt,
+      createdAt,
-    }).run();
+    }).run();
-  } catch (error) {
+  } catch (error) {
-    console.warn('[proxy/embeddings] failed to write proxy log', error);
+    console.warn('[proxy/embeddings] failed to write proxy log', error);
-  }
+  }
-}
+}
 

src/server/routes/proxy/images.ts

@@ -2,40 +2,41 @@
 import { fetch } from 'undici';
 import { db, schema } from '../../db/index.js';
 import { tokenRouter } from '../../services/tokenRouter.js';
-import { refreshModelsAndRebuildRoutes } from '../../services/modelService.js';
+import { refreshModelsAndRebuildRoutes } from '../../services/modelService.js';
-import { reportProxyAllFailed, reportTokenExpired } from '../../services/alertService.js';
+import { reportProxyAllFailed, reportTokenExpired } from '../../services/alertService.js';
-import { isTokenExpiredError } from '../../services/alertRules.js';
+import { isTokenExpiredError } from '../../services/alertRules.js';
-import { estimateProxyCost } from '../../services/modelPricingService.js';
+import { estimateProxyCost } from '../../services/modelPricingService.js';
-import { shouldRetryProxyRequest } from '../../services/proxyRetryPolicy.js';
+import { shouldRetryProxyRequest } from '../../services/proxyRetryPolicy.js';
-import { ensureModelAllowedForDownstreamKey, getDownstreamRoutingPolicy, recordDownstreamCostUsage } from './downstreamPolicy.js';
+import { ensureModelAllowedForDownstreamKey, getDownstreamRoutingPolicy, recordDownstreamCostUsage } from './downstreamPolicy.js';
-import { withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
+import { withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
-import { composeProxyLogMessage } from './logPathMeta.js';
+import { getProxyUrlFromExtraConfig } from '../../services/accountExtraConfig.js';
-import { formatUtcSqlDateTime } from '../../services/localTimeService.js';
+import { composeProxyLogMessage } from './logPathMeta.js';
-import { cloneFormDataWithOverrides, ensureMultipartBufferParser, parseMultipartFormData } from './multipart.js';
+import { formatUtcSqlDateTime } from '../../services/localTimeService.js';
-import { buildUpstreamUrl } from './upstreamUrl.js';
+import { cloneFormDataWithOverrides, ensureMultipartBufferParser, parseMultipartFormData } from './multipart.js';
+import { buildUpstreamUrl } from './upstreamUrl.js';
 
 const MAX_RETRIES = 2;
 
-export async function imagesProxyRoute(app: FastifyInstance) {
+export async function imagesProxyRoute(app: FastifyInstance) {
-  ensureMultipartBufferParser(app);
+  ensureMultipartBufferParser(app);
-
+
-  app.post('/v1/images/generations', async (request: FastifyRequest, reply: FastifyReply) => {
+  app.post('/v1/images/generations', async (request: FastifyRequest, reply: FastifyReply) => {
-    const body = request.body as any;
+    const body = request.body as any;
-    const requestedModel = body?.model || 'gpt-image-1';
+    const requestedModel = body?.model || 'gpt-image-1';
-    if (!await ensureModelAllowedForDownstreamKey(request, reply, requestedModel)) return;
+    if (!await ensureModelAllowedForDownstreamKey(request, reply, requestedModel)) return;
-    const downstreamPolicy = getDownstreamRoutingPolicy(request);
+    const downstreamPolicy = getDownstreamRoutingPolicy(request);
-    const excludeChannelIds: number[] = [];
+    const excludeChannelIds: number[] = [];
-    let retryCount = 0;
+    let retryCount = 0;
-
+
-    while (retryCount <= MAX_RETRIES) {
+    while (retryCount <= MAX_RETRIES) {
-      let selected = retryCount === 0
+      let selected = retryCount === 0
-        ? await tokenRouter.selectChannel(requestedModel, downstreamPolicy)
+        ? await tokenRouter.selectChannel(requestedModel, downstreamPolicy)
-        : await tokenRouter.selectNextChannel(requestedModel, excludeChannelIds, downstreamPolicy);
+        : await tokenRouter.selectNextChannel(requestedModel, excludeChannelIds, downstreamPolicy);
-
+
-      if (!selected && retryCount === 0) {
+      if (!selected && retryCount === 0) {
-        await refreshModelsAndRebuildRoutes();
+        await refreshModelsAndRebuildRoutes();
-        selected = await tokenRouter.selectChannel(requestedModel, downstreamPolicy);
+        selected = await tokenRouter.selectChannel(requestedModel, downstreamPolicy);
-      }
+      }
 
       if (!selected) {
         await reportProxyAllFailed({
@@ -49,19 +50,19 @@ export async function imagesProxyRoute(app: FastifyInstance) {
 
       excludeChannelIds.push(selected.channel.id);
 
-      const targetUrl = buildUpstreamUrl(selected.site.url, '/v1/images/generations');
+      const targetUrl = buildUpstreamUrl(selected.site.url, '/v1/images/generations');
       const forwardBody = { ...body, model: selected.actualModel };
       const startTime = Date.now();
 
       try {
-        const upstream = await fetch(targetUrl, withSiteRecordProxyRequestInit(selected.site, {
+        const upstream = await fetch(targetUrl, withSiteRecordProxyRequestInit(selected.site, {
-          method: 'POST',
+          method: 'POST',
-          headers: {
+          headers: {
-            'Content-Type': 'application/json',
+            'Content-Type': 'application/json',
-            'Authorization': `Bearer ${selected.tokenValue}`,
+            'Authorization': `Bearer ${selected.tokenValue}`,
-          },
+          },
-          body: JSON.stringify(forwardBody),
+          body: JSON.stringify(forwardBody),
-        }));
+        }, getProxyUrlFromExtraConfig(selected.account.extraConfig)));
 
         const text = await upstream.text();
         if (!upstream.ok) {
@@ -75,10 +76,10 @@ export async function imagesProxyRoute(app: FastifyInstance) {
               detail: `HTTP ${upstream.status}`,
             });
           }
-          if (shouldRetryProxyRequest(upstream.status, text) && retryCount < MAX_RETRIES) {
+          if (shouldRetryProxyRequest(upstream.status, text) && retryCount < MAX_RETRIES) {
-            retryCount++;
+            retryCount++;
-            continue;
+            continue;
-          }
+          }
           await reportProxyAllFailed({
             model: requestedModel,
             reason: `upstream returned HTTP ${upstream.status}`,
@@ -90,18 +91,18 @@ export async function imagesProxyRoute(app: FastifyInstance) {
         try { data = JSON.parse(text); } catch { data = { data: [] }; }
 
         const latency = Date.now() - startTime;
-        const estimatedCost = await estimateProxyCost({
+        const estimatedCost = await estimateProxyCost({
-          site: selected.site,
+          site: selected.site,
-          account: selected.account,
+          account: selected.account,
-          modelName: selected.actualModel || requestedModel,
+          modelName: selected.actualModel || requestedModel,
           promptTokens: 0,
           completionTokens: 0,
           totalTokens: 0,
-        });
+        });
-        tokenRouter.recordSuccess(selected.channel.id, latency, estimatedCost);
+        tokenRouter.recordSuccess(selected.channel.id, latency, estimatedCost);
-        recordDownstreamCostUsage(request, estimatedCost);
+        recordDownstreamCostUsage(request, estimatedCost);
-        logProxy(selected, requestedModel, 'success', upstream.status, latency, null, retryCount, estimatedCost);
+        logProxy(selected, requestedModel, 'success', upstream.status, latency, null, retryCount, estimatedCost);
-        return reply.code(upstream.status).send(data);
+        return reply.code(upstream.status).send(data);
       } catch (err: any) {
         tokenRouter.recordFailure(selected.channel.id);
         logProxy(selected, requestedModel, 'failed', 0, Date.now() - startTime, err.message, retryCount);
@@ -116,175 +117,176 @@ export async function imagesProxyRoute(app: FastifyInstance) {
         return reply.code(502).send({
           error: { message: `Upstream error: ${err.message}`, type: 'upstream_error' },
         });
-      }
+      }
-    }
+    }
-  });
+  });
-
+
-  app.post('/v1/images/edits', async (request: FastifyRequest, reply: FastifyReply) => {
+  app.post('/v1/images/edits', async (request: FastifyRequest, reply: FastifyReply) => {
-    const multipartForm = await parseMultipartFormData(request);
+    const multipartForm = await parseMultipartFormData(request);
-    const jsonBody = (!multipartForm && request.body && typeof request.body === 'object')
+    const jsonBody = (!multipartForm && request.body && typeof request.body === 'object')
-      ? request.body as Record<string, unknown>
+      ? request.body as Record<string, unknown>
-      : null;
+      : null;
-    const requestedModel = typeof multipartForm?.get('model') === 'string'
+    const requestedModel = typeof multipartForm?.get('model') === 'string'
-      ? String(multipartForm.get('model')).trim()
+      ? String(multipartForm.get('model')).trim()
-      : (typeof jsonBody?.model === 'string' ? jsonBody.model.trim() : '') || 'gpt-image-1';
+      : (typeof jsonBody?.model === 'string' ? jsonBody.model.trim() : '') || 'gpt-image-1';
-
+
-    if (!await ensureModelAllowedForDownstreamKey(request, reply, requestedModel)) return;
+    if (!await ensureModelAllowedForDownstreamKey(request, reply, requestedModel)) return;
-    const downstreamPolicy = getDownstreamRoutingPolicy(request);
+    const downstreamPolicy = getDownstreamRoutingPolicy(request);
-    const excludeChannelIds: number[] = [];
+    const excludeChannelIds: number[] = [];
-    let retryCount = 0;
+    let retryCount = 0;
-
+
-    while (retryCount <= MAX_RETRIES) {
+    while (retryCount <= MAX_RETRIES) {
-      let selected = retryCount === 0
+      let selected = retryCount === 0
-        ? await tokenRouter.selectChannel(requestedModel, downstreamPolicy)
+        ? await tokenRouter.selectChannel(requestedModel, downstreamPolicy)
-        : await tokenRouter.selectNextChannel(requestedModel, excludeChannelIds, downstreamPolicy);
+        : await tokenRouter.selectNextChannel(requestedModel, excludeChannelIds, downstreamPolicy);
-
+
-      if (!selected && retryCount === 0) {
+      if (!selected && retryCount === 0) {
-        await refreshModelsAndRebuildRoutes();
+        await refreshModelsAndRebuildRoutes();
-        selected = await tokenRouter.selectChannel(requestedModel, downstreamPolicy);
+        selected = await tokenRouter.selectChannel(requestedModel, downstreamPolicy);
-      }
+      }
-
+
-      if (!selected) {
+      if (!selected) {
-        await reportProxyAllFailed({
+        await reportProxyAllFailed({
-          model: requestedModel,
+          model: requestedModel,
-          reason: 'No available channels after retries',
+          reason: 'No available channels after retries',
-        });
+        });
-        return reply.code(503).send({
+        return reply.code(503).send({
-          error: { message: 'No available channels for this model', type: 'server_error' },
+          error: { message: 'No available channels for this model', type: 'server_error' },
-        });
+        });
-      }
+      }
-
+
-      excludeChannelIds.push(selected.channel.id);
+      excludeChannelIds.push(selected.channel.id);
-      const targetUrl = buildUpstreamUrl(selected.site.url, '/v1/images/edits');
+      const targetUrl = buildUpstreamUrl(selected.site.url, '/v1/images/edits');
-      const startTime = Date.now();
+      const startTime = Date.now();
-
+
-      try {
+      try {
-        const requestInit = multipartForm
+        const accountProxy = getProxyUrlFromExtraConfig(selected.account.extraConfig);
-          ? withSiteRecordProxyRequestInit(selected.site, {
+        const requestInit = multipartForm
-            method: 'POST',
+          ? withSiteRecordProxyRequestInit(selected.site, {
-            headers: {
+            method: 'POST',
-              Authorization: `Bearer ${selected.tokenValue}`,
+            headers: {
-            },
+              Authorization: `Bearer ${selected.tokenValue}`,
-            body: cloneFormDataWithOverrides(multipartForm, {
+            },
-              model: selected.actualModel || requestedModel,
+            body: cloneFormDataWithOverrides(multipartForm, {
-            }) as any,
+              model: selected.actualModel || requestedModel,
-          })
+            }) as any,
-          : withSiteRecordProxyRequestInit(selected.site, {
+          }, accountProxy)
-            method: 'POST',
+          : withSiteRecordProxyRequestInit(selected.site, {
-            headers: {
+            method: 'POST',
-              'Content-Type': 'application/json',
+            headers: {
-              Authorization: `Bearer ${selected.tokenValue}`,
+              'Content-Type': 'application/json',
-            },
+              Authorization: `Bearer ${selected.tokenValue}`,
-            body: JSON.stringify({
+            },
-              ...(jsonBody || {}),
+            body: JSON.stringify({
-              model: selected.actualModel || requestedModel,
+              ...(jsonBody || {}),
-            }),
+              model: selected.actualModel || requestedModel,
-          });
+            }),
-
+          }, accountProxy);
-        const upstream = await fetch(targetUrl, requestInit);
+
-        const text = await upstream.text();
+        const upstream = await fetch(targetUrl, requestInit);
-        if (!upstream.ok) {
+        const text = await upstream.text();
-          tokenRouter.recordFailure(selected.channel.id);
+        if (!upstream.ok) {
-          logProxy(selected, requestedModel, 'failed', upstream.status, Date.now() - startTime, text, retryCount, 0, '/v1/images/edits');
+          tokenRouter.recordFailure(selected.channel.id);
-          if (isTokenExpiredError({ status: upstream.status, message: text })) {
+          logProxy(selected, requestedModel, 'failed', upstream.status, Date.now() - startTime, text, retryCount, 0, '/v1/images/edits');
-            await reportTokenExpired({
+          if (isTokenExpiredError({ status: upstream.status, message: text })) {
-              accountId: selected.account.id,
+            await reportTokenExpired({
-              username: selected.account.username,
+              accountId: selected.account.id,
-              siteName: selected.site.name,
+              username: selected.account.username,
-              detail: `HTTP ${upstream.status}`,
+              siteName: selected.site.name,
-            });
+              detail: `HTTP ${upstream.status}`,
-          }
+            });
-          if (shouldRetryProxyRequest(upstream.status, text) && retryCount < MAX_RETRIES) {
+          }
-            retryCount++;
+          if (shouldRetryProxyRequest(upstream.status, text) && retryCount < MAX_RETRIES) {
-            continue;
+            retryCount++;
-          }
+            continue;
-          await reportProxyAllFailed({
+          }
-            model: requestedModel,
+          await reportProxyAllFailed({
-            reason: `upstream returned HTTP ${upstream.status}`,
+            model: requestedModel,
-          });
+            reason: `upstream returned HTTP ${upstream.status}`,
-          return reply.code(upstream.status).send({ error: { message: text, type: 'upstream_error' } });
+          });
-        }
+          return reply.code(upstream.status).send({ error: { message: text, type: 'upstream_error' } });
-
+        }
-        let data: any = {};
+
-        try { data = JSON.parse(text); } catch { data = { data: [] }; }
+        let data: any = {};
-
+        try { data = JSON.parse(text); } catch { data = { data: [] }; }
-        const latency = Date.now() - startTime;
+
-        const estimatedCost = await estimateProxyCost({
+        const latency = Date.now() - startTime;
-          site: selected.site,
+        const estimatedCost = await estimateProxyCost({
-          account: selected.account,
+          site: selected.site,
-          modelName: selected.actualModel || requestedModel,
+          account: selected.account,
-          promptTokens: 0,
+          modelName: selected.actualModel || requestedModel,
-          completionTokens: 0,
+          promptTokens: 0,
-          totalTokens: 0,
+          completionTokens: 0,
-        });
+          totalTokens: 0,
-        tokenRouter.recordSuccess(selected.channel.id, latency, estimatedCost);
+        });
-        recordDownstreamCostUsage(request, estimatedCost);
+        tokenRouter.recordSuccess(selected.channel.id, latency, estimatedCost);
-        logProxy(selected, requestedModel, 'success', upstream.status, latency, null, retryCount, estimatedCost, '/v1/images/edits');
+        recordDownstreamCostUsage(request, estimatedCost);
-        return reply.code(upstream.status).send(data);
+        logProxy(selected, requestedModel, 'success', upstream.status, latency, null, retryCount, estimatedCost, '/v1/images/edits');
-      } catch (err: any) {
+        return reply.code(upstream.status).send(data);
-        tokenRouter.recordFailure(selected.channel.id);
+      } catch (err: any) {
-        logProxy(selected, requestedModel, 'failed', 0, Date.now() - startTime, err.message, retryCount, 0, '/v1/images/edits');
+        tokenRouter.recordFailure(selected.channel.id);
-        if (retryCount < MAX_RETRIES) {
+        logProxy(selected, requestedModel, 'failed', 0, Date.now() - startTime, err.message, retryCount, 0, '/v1/images/edits');
-          retryCount++;
+        if (retryCount < MAX_RETRIES) {
-          continue;
+          retryCount++;
-        }
+          continue;
-        await reportProxyAllFailed({
+        }
-          model: requestedModel,
+        await reportProxyAllFailed({
-          reason: err.message || 'network failure',
+          model: requestedModel,
-        });
+          reason: err.message || 'network failure',
-        return reply.code(502).send({
+        });
-          error: { message: `Upstream error: ${err.message}`, type: 'upstream_error' },
+        return reply.code(502).send({
-        });
+          error: { message: `Upstream error: ${err.message}`, type: 'upstream_error' },
-      }
+        });
-    }
+      }
-  });
+    }
-
+  });
-  app.post('/v1/images/variations', async (_request: FastifyRequest, reply: FastifyReply) => {
+
-    return reply.code(400).send({
+  app.post('/v1/images/variations', async (_request: FastifyRequest, reply: FastifyReply) => {
-      error: {
+    return reply.code(400).send({
-        message: 'Image variations are not supported',
+      error: {
-        type: 'invalid_request_error',
+        message: 'Image variations are not supported',
-      },
+        type: 'invalid_request_error',
-    });
+      },
-  });
+    });
-}
+  });
+}
 
 async function logProxy(
   selected: any,
   modelRequested: string,
   status: string,
   httpStatus: number,
-  latencyMs: number,
+  latencyMs: number,
-  errorMessage: string | null,
+  errorMessage: string | null,
-  retryCount: number,
+  retryCount: number,
-  estimatedCost = 0,
+  estimatedCost = 0,
-  downstreamPath = '/v1/images/generations',
+  downstreamPath = '/v1/images/generations',
-) {
+) {
-  try {
+  try {
-    const createdAt = formatUtcSqlDateTime(new Date());
+    const createdAt = formatUtcSqlDateTime(new Date());
-    const normalizedErrorMessage = composeProxyLogMessage({
+    const normalizedErrorMessage = composeProxyLogMessage({
-      downstreamPath,
+      downstreamPath,
-      errorMessage,
+      errorMessage,
-    });
+    });
-    await db.insert(schema.proxyLogs).values({
+    await db.insert(schema.proxyLogs).values({
-      routeId: selected.channel.routeId,
+      routeId: selected.channel.routeId,
-      channelId: selected.channel.id,
+      channelId: selected.channel.id,
-      accountId: selected.account.id,
+      accountId: selected.account.id,
-      modelRequested,
+      modelRequested,
-      modelActual: selected.actualModel,
+      modelActual: selected.actualModel,
-      status,
+      status,
-      httpStatus,
+      httpStatus,
-      latencyMs,
+      latencyMs,
-      promptTokens: 0,
+      promptTokens: 0,
-      completionTokens: 0,
+      completionTokens: 0,
-      totalTokens: 0,
+      totalTokens: 0,
-      estimatedCost,
+      estimatedCost,
-      errorMessage: normalizedErrorMessage,
+      errorMessage: normalizedErrorMessage,
-      retryCount,
+      retryCount,
-      createdAt,
+      createdAt,
-    }).run();
+    }).run();
-  } catch (error) {
+  } catch (error) {
-    console.warn('[proxy/images] failed to write proxy log', error);
+    console.warn('[proxy/images] failed to write proxy log', error);
-  }
+  }
-}
+}
 

src/server/routes/proxy/responses.ts

@@ -8,7 +8,8 @@ import { isTokenExpiredError } from '../../services/alertRules.js';
 import { shouldRetryProxyRequest } from '../../services/proxyRetryPolicy.js';
 import { resolveProxyUsageWithSelfLogFallback } from '../../services/proxyUsageFallbackService.js';
 import { mergeProxyUsage, parseProxyUsage } from '../../services/proxyUsageParser.js';
-import { resolveProxyUrlForSite, withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
+import { resolveChannelProxyUrl, withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
+import { getProxyUrlFromExtraConfig } from '../../services/accountExtraConfig.js';
 import { openAiResponsesTransformer } from '../../transformers/openai/responses/index.js';
 import {
   buildUpstreamEndpointRequest,
@@ -259,7 +260,7 @@ export async function responsesProxyRoute(app: FastifyInstance) {
             method: 'POST',
             headers: compatibilityRequest.headers,
             body: JSON.stringify(compatibilityRequest.body),
-          }),
+          }, getProxyUrlFromExtraConfig(selected.account.extraConfig)),
         ),
       });
 
@@ -268,7 +269,7 @@ export async function responsesProxyRoute(app: FastifyInstance) {
       try {
         const endpointResult = await executeEndpointFlow({
           siteUrl: selected.site.url,
-          proxyUrl: resolveProxyUrlForSite(selected.site),
+          proxyUrl: resolveChannelProxyUrl(selected.site, selected.account.extraConfig),
           endpointCandidates,
           buildRequest: (endpoint) => buildEndpointRequest(endpoint),
           tryRecover: endpointStrategy.tryRecover,

src/server/routes/proxy/search.ts

@@ -8,6 +8,7 @@ import { isTokenExpiredError } from '../../services/alertRules.js';
 import { shouldRetryProxyRequest } from '../../services/proxyRetryPolicy.js';
 import { ensureModelAllowedForDownstreamKey, getDownstreamRoutingPolicy, recordDownstreamCostUsage } from './downstreamPolicy.js';
 import { withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
+import { getProxyUrlFromExtraConfig } from '../../services/accountExtraConfig.js';
 import { composeProxyLogMessage } from './logPathMeta.js';
 import { formatUtcSqlDateTime } from '../../services/localTimeService.js';
 import { buildUpstreamUrl } from './upstreamUrl.js';
@@ -92,7 +93,7 @@ export async function searchProxyRoute(app: FastifyInstance) {
             Authorization: `Bearer ${selected.tokenValue}`,
           },
           body: JSON.stringify(forwardBody),
-        }));
+        }, getProxyUrlFromExtraConfig(selected.account.extraConfig)));
 
         const text = await upstream.text();
         if (!upstream.ok) {

src/server/routes/proxy/videos.ts

@@ -8,6 +8,7 @@ import { estimateProxyCost } from '../../services/modelPricingService.js';
 import { shouldRetryProxyRequest } from '../../services/proxyRetryPolicy.js';
 import { ensureModelAllowedForDownstreamKey, getDownstreamRoutingPolicy, recordDownstreamCostUsage } from './downstreamPolicy.js';
 import { withSiteProxyRequestInit, withSiteRecordProxyRequestInit } from '../../services/siteProxy.js';
+import { getProxyUrlFromExtraConfig } from '../../services/accountExtraConfig.js';
 import { cloneFormDataWithOverrides, ensureMultipartBufferParser, parseMultipartFormData } from './multipart.js';
 import { buildUpstreamUrl } from './upstreamUrl.js';
 import {
@@ -75,6 +76,7 @@ export async function videosProxyRoute(app: FastifyInstance) {
       const startTime = Date.now();
 
       try {
+        const accountProxy = getProxyUrlFromExtraConfig(selected.account.extraConfig);
         const requestInit = multipartForm
           ? withSiteRecordProxyRequestInit(selected.site, {
             method: 'POST',
@@ -84,7 +86,7 @@ export async function videosProxyRoute(app: FastifyInstance) {
             body: cloneFormDataWithOverrides(multipartForm, {
               model: selected.actualModel || requestedModel,
             }) as any,
-          })
+          }, accountProxy)
           : withSiteRecordProxyRequestInit(selected.site, {
             method: 'POST',
             headers: {
@@ -95,7 +97,7 @@ export async function videosProxyRoute(app: FastifyInstance) {
               ...(jsonBody || {}),
               model: selected.actualModel || requestedModel,
             }),
-          });
+          }, accountProxy);
 
         const upstream = await fetch(targetUrl, requestInit);
         const text = await upstream.text();