n9router 0.4.44 → 0.4.46

package/app/open-sse/executors/vertex.js

@@ -1,11 +1,33 @@
 import { BaseExecutor } from "./base.js";
 import { PROVIDERS } from "../config/providers.js";
-import { parseVertexSaJson, refreshVertexToken } from "../services/tokenRefresh.js";
+import { parseVertexSaJson, refreshVertexToken, refreshGoogleToken } from "../services/tokenRefresh.js";
 import { proxyAwareFetch } from "../utils/proxyFetch.js";
 
 // Cache project IDs resolved from raw API keys { apiKey → projectId }
 const projectIdCache = new Map();
 
+/**
+ * Parse Google ADC user credential JSON from apiKey string.
+ * This is the format produced by `gcloud auth application-default login`.
+ */
+function parseVertexAdcJson(apiKey) {
+  if (typeof apiKey !== "string") return null;
+  try {
+    const parsed = JSON.parse(apiKey);
+    if (
+      parsed.type === "authorized_user" &&
+      parsed.client_id &&
+      parsed.client_secret &&
+      parsed.refresh_token
+    ) {
+      return parsed;
+    }
+    return null;
+  } catch {
+    return null;
+  }
+}
+
 /**
  * Resolve GCP project ID from a raw Vertex API key.
  * Sends a dummy 404 request and parses "projects/{id}" from the error message.
@@ -43,8 +65,13 @@ export class VertexExecutor extends BaseExecutor {
 
   buildUrl(model, stream, urlIndex = 0, credentials = null) {
     const saJson = parseVertexSaJson(credentials?.apiKey);
-    const rawKey = !saJson ? credentials?.apiKey : null;
-    const projectId = saJson?.project_id || credentials?.providerSpecificData?.projectId;
+    const adcJson = parseVertexAdcJson(credentials?.apiKey);
+    const usesOAuth = !!saJson || !!adcJson || !!credentials?.accessToken;
+    const rawKey = !usesOAuth ? credentials?.apiKey : null;
+    const projectId =
+      saJson?.project_id ||
+      adcJson?.quota_project_id ||
+      credentials?.providerSpecificData?.projectId;
 
     if (this.provider === "vertex-partner") {
       // Partner models require project_id in path regardless of auth method
@@ -56,8 +83,14 @@ export class VertexExecutor extends BaseExecutor {
     // Gemini on Vertex
     const action = stream ? "streamGenerateContent" : "generateContent";
 
-    if (saJson) {
-      // SA JSON + Bearer token: must use project-scoped path to avoid RESOURCE_PROJECT_INVALID
+    if (usesOAuth) {
+      // SA JSON / ADC / pre-set accessToken: must use project-scoped path to avoid RESOURCE_PROJECT_INVALID
+      if (!projectId) {
+        throw new Error(
+          "Vertex OAuth/ADC requires a project_id. " +
+          "Add quota_project_id to your ADC JSON or set providerSpecificData.projectId."
+        );
+      }
       const location = credentials?.providerSpecificData?.location || "us-central1";
       let url = `https://aiplatform.googleapis.com/v1/projects/${projectId}/locations/${location}/publishers/google/models/${model}:${action}`;
       if (stream) url += "?alt=sse";
@@ -97,16 +130,29 @@ export class VertexExecutor extends BaseExecutor {
 
   async execute({ model, body, stream, credentials, signal, log, proxyOptions = null }) {
     const saJson = parseVertexSaJson(credentials?.apiKey);
+    const adcJson = parseVertexAdcJson(credentials?.apiKey);
 
-    // SA JSON flow: mint Bearer token (cached)
+    // SA JSON flow: mint Bearer token via JWT assertion (cached)
     if (saJson) {
       const result = await refreshVertexToken(saJson, log);
       if (!result?.accessToken) throw new Error("Vertex: failed to mint access token from Service Account JSON");
       credentials.accessToken = result.accessToken;
     }
 
+    // ADC user credential flow: refresh Bearer token via Google OAuth2 token endpoint
+    if (adcJson) {
+      const result = await refreshGoogleToken(
+        adcJson.refresh_token,
+        adcJson.client_id,
+        adcJson.client_secret,
+        log
+      );
+      if (!result?.accessToken) throw new Error("Vertex: failed to refresh access token from ADC JSON (authorized_user)");
+      credentials.accessToken = result.accessToken;
+    }
+
     // vertex-partner with raw key: auto-resolve project_id if not provided
-    if (this.provider === "vertex-partner" && !saJson && !credentials?.providerSpecificData?.projectId) {
+    if (this.provider === "vertex-partner" && !saJson && !adcJson && !credentials?.providerSpecificData?.projectId) {
       const projectId = await resolveProjectId(credentials.apiKey);
       if (!projectId) throw new Error("Vertex: could not resolve project_id from API key. Please add it manually in provider settings.");
       log?.debug?.("VERTEX", `Resolved project_id: ${projectId}`);

package/app/open-sse/executors/xiaomi-tokenplan.js

@@ -1,18 +1,19 @@
 import { DefaultExecutor } from "./default.js";
 import { resolveXiaomiTokenplanBaseUrl } from "../config/providers.js";
-import { getModelTargetFormat } from "../config/providerModels.js";
-import { FORMATS } from "../translator/formats.js";
+// import { getModelTargetFormat } from "../config/providerModels.js";
+// import { FORMATS } from "../translator/formats.js";
 
 export class XiaomiTokenplanExecutor extends DefaultExecutor {
   constructor() {
     super("xiaomi-tokenplan");
   }
 
-  // Claude-native aliases route to the Anthropic-compatible messages endpoint
+  // Token Plan keys are region-specific. Route per sourceFormat-matched transport:
+  // claude → Anthropic /anthropic/v1/messages, openai → /chat/completions.
   buildUrl(model, stream, urlIndex = 0, credentials = null) {
     const baseUrl = resolveXiaomiTokenplanBaseUrl(credentials);
-    if (getModelTargetFormat(model, model) === FORMATS.CLAUDE) {
-      return `${baseUrl.replace(/\/v1\/?$/, "/anthropic/v1")}/messages`;
+    if (credentials?.runtimeTransport?.format === "claude") {
+      return `${baseUrl.replace(/\/v1\/?$/, "")}/anthropic/v1/messages`;
     }
     return `${baseUrl}/chat/completions`;
   }

package/app/open-sse/handlers/chatCore/nonStreamingHandler.js

@@ -1,5 +1,6 @@
 import { FORMATS } from "../../translator/formats.js";
 import { needsTranslation } from "../../translator/index.js";
+import { fromOpenAIFinish } from "../../translator/concerns/finishReason.js";
 import { ollamaBodyToOpenAI } from "../../translator/response/ollama-to-openai.js";
 import { addBufferToUsage, filterUsageForFormat } from "../../utils/usageTracking.js";
 import { createErrorResult } from "../../utils/error.js";
@@ -9,11 +10,75 @@ import { buildRequestDetail, extractRequestConfig, extractUsageFromResponse, sav
 import { appendRequestLog, saveRequestDetail } from "@/lib/usageDb.js";
 import { decloakToolNames } from "../../utils/claudeCloaking.js";
 
+function parseToolArguments(value) {
+  if (!value) return {};
+  if (typeof value === "object") return value;
+  try {
+    return JSON.parse(value);
+  } catch {
+    return {};
+  }
+}
+
+function unwrapOpenAIDataEnvelope(responseBody) {
+  const inner = responseBody?.data;
+  if (!inner || typeof inner !== "object" || !Array.isArray(inner.choices)) return responseBody;
+  return {
+    ...inner,
+    object: inner.object || responseBody.object,
+    created: inner.created || responseBody.created,
+  };
+}
+
+function openAICompletionToClaudeMessage(responseBody) {
+  if (!responseBody?.choices?.[0]) return responseBody;
+  const choice = responseBody.choices[0];
+  const message = choice.message || {};
+  const content = [];
+
+  const reasoning = message.reasoning_content || message.provider_specific_fields?.reasoning_content || "";
+  if (reasoning) {
+    content.push({ type: "thinking", thinking: reasoning });
+  }
+  if (typeof message.content === "string" && message.content.length > 0) {
+    content.push({ type: "text", text: message.content });
+  }
+  for (const toolCall of message.tool_calls || []) {
+    const fn = toolCall.function || {};
+    content.push({
+      type: "tool_use",
+      id: toolCall.id || `toolu_${Date.now()}_${content.length}`,
+      name: fn.name || toolCall.name || "",
+      input: parseToolArguments(fn.arguments || toolCall.arguments),
+    });
+  }
+  if (content.length === 0) content.push({ type: "text", text: "" });
+
+  const usage = responseBody.usage || {};
+  return {
+    id: String(responseBody.id || `msg_${Date.now()}`).replace(/^chatcmpl-/, ""),
+    type: "message",
+    role: "assistant",
+    model: responseBody.model || "unknown",
+    content,
+    stop_reason: fromOpenAIFinish(choice.finish_reason, FORMATS.CLAUDE),
+    stop_sequence: null,
+    usage: {
+      input_tokens: usage.prompt_tokens || usage.input_tokens || 0,
+      output_tokens: usage.completion_tokens || usage.output_tokens || 0,
+    },
+  };
+}
+
 /**
  * Translate non-streaming response body from provider format → OpenAI format.
  */
 export function translateNonStreamingResponse(responseBody, targetFormat, sourceFormat) {
-  if (targetFormat === sourceFormat || targetFormat === FORMATS.OPENAI) return responseBody;
+  if (targetFormat === sourceFormat) return responseBody;
+  if (targetFormat === FORMATS.OPENAI && sourceFormat === FORMATS.CLAUDE) {
+    return openAICompletionToClaudeMessage(responseBody);
+  }
+  if (targetFormat === FORMATS.OPENAI) return responseBody;
 
   // Gemini / Antigravity
   if (targetFormat === FORMATS.GEMINI || targetFormat === FORMATS.ANTIGRAVITY || targetFormat === FORMATS.GEMINI_CLI || targetFormat === FORMATS.VERTEX) {
@@ -37,6 +102,12 @@ export function translateNonStreamingResponse(responseBody, targetFormat, source
             function: { name: part.functionCall.name, arguments: JSON.stringify(part.functionCall.args || {}) }
           });
         }
+        // Handle inline image data (from image generation models)
+        const inlineData = part.inlineData || part.inline_data;
+        if (inlineData?.data) {
+          const mimeType = inlineData.mimeType || inlineData.mime_type || "image/png";
+          textContent += `\n![image](data:${mimeType};base64,${inlineData.data})\n`;
+        }
       }
     }
 
@@ -76,7 +147,12 @@ export function translateNonStreamingResponse(responseBody, targetFormat, source
     // missing/null (e.g. M3 with max_tokens:1 spends the budget on thinking
     // and returns `content: null`). Returning the raw body would leave the
     // OpenAI client without a `choices` array and surface as a UI test error.
-    if (responseBody.content && !Array.isArray(responseBody.content)) return responseBody;
+    // Early return if the response is already in OpenAI format (has choices array)
+    // or if it has content as a non-array value (likely a different non-Claude format).
+    // Some providers (e.g. xiaomi-tokenplan) return OpenAI-format responses even when
+    // the request was translated to Claude format — the targetFormat is Claude but the
+    // actual response is OpenAI-native and needs no further translation.
+    if (responseBody.choices || (responseBody.content && !Array.isArray(responseBody.content))) return responseBody;
 
     let textContent = "", thinkingContent = "";
     const toolCalls = [];
@@ -156,10 +232,17 @@ export async function handleNonStreamingResponse({ providerResponse, provider, m
   }
 
   reqLogger.logProviderResponse(providerResponse.status, providerResponse.statusText, providerResponse.headers, responseBody);
-  if (onRequestSuccess) await onRequestSuccess();
+  if (onRequestSuccess) {
+    Promise.resolve()
+      .then(onRequestSuccess)
+      .catch(err => {
+        console.error("[ChatCore] onRequestSuccess failed:", err?.message || err);
+      });
+  }
 
   // Decloak tool_use names once on raw Claude body, before any translation (INPUT side)
   responseBody = decloakToolNames(responseBody, toolNameMap);
+  responseBody = unwrapOpenAIDataEnvelope(responseBody);
 
   const usage = extractUsageFromResponse(responseBody);
   appendLog({ tokens: usage, status: "200 OK" });
@@ -168,6 +251,7 @@ export async function handleNonStreamingResponse({ providerResponse, provider, m
   const translatedResponse = needsTranslation(targetFormat, sourceFormat)
     ? translateNonStreamingResponse(responseBody, targetFormat, sourceFormat)
     : responseBody;
+  const isClaudeMessageResponse = sourceFormat === FORMATS.CLAUDE && translatedResponse?.type === "message";
 
   // Fix finish_reason for tool_calls: some providers return non-standard values (e.g. "other")
   if (translatedResponse?.choices?.[0]) {
@@ -180,24 +264,31 @@ export async function handleNonStreamingResponse({ providerResponse, provider, m
   }
 
   // Ensure OpenAI-required fields
-  if (!translatedResponse.object) translatedResponse.object = "chat.completion";
-  if (!translatedResponse.created) translatedResponse.created = Math.floor(Date.now() / 1000);
+  if (!isClaudeMessageResponse) {
+    if (!translatedResponse.object) translatedResponse.object = "chat.completion";
+    if (!translatedResponse.created) translatedResponse.created = Math.floor(Date.now() / 1000);
+  }
 
   // Strip Azure-specific fields
-  delete translatedResponse.prompt_filter_results;
-  if (translatedResponse?.choices) {
-    for (const choice of translatedResponse.choices) delete choice.content_filter_results;
+  if (!isClaudeMessageResponse) {
+    delete translatedResponse.prompt_filter_results;
+    if (translatedResponse?.choices) {
+      for (const choice of translatedResponse.choices) delete choice.content_filter_results;
+    }
   }
 
   if (translatedResponse?.usage) {
     translatedResponse.usage = filterUsageForFormat(addBufferToUsage(translatedResponse.usage), sourceFormat);
   }
 
-  // Strip reasoning_content — some clients (e.g. Firecrawl AI SDK) have JSON parsers that
-  // break on this non-standard field, even though OpenAI allows it in extensions.
-  if (translatedResponse?.choices) {
+  // Strip reasoning_content only when content is non-empty.
+  // When content is empty (e.g. thinking models that used all tokens for reasoning),
+  // reasoning_content is the only useful output and must be preserved.
+  if (!isClaudeMessageResponse && translatedResponse?.choices) {
     for (const choice of translatedResponse.choices) {
-      if (choice?.message) delete choice.message.reasoning_content;
+      if (choice?.message?.reasoning_content && choice.message.content) {
+        delete choice.message.reasoning_content;
+      }
     }
   }
 

package/app/open-sse/handlers/chatCore/requestDetail.js

@@ -1,5 +1,6 @@
 import { saveRequestUsage, appendRequestLog, saveRequestDetail } from "@/lib/usageDb.js";
 import { COLORS } from "../../utils/stream.js";
+import { canonicalizeUsage } from "../../utils/usageTracking.js";
 
 const OPTIONAL_PARAMS = [
   "temperature", "top_p", "top_k",
@@ -48,8 +49,8 @@ export function extractUsageFromResponse(responseBody) {
     return {
       prompt_tokens: responseBody.usageMetadata.promptTokenCount || 0,
       completion_tokens: responseBody.usageMetadata.candidatesTokenCount || 0,
-      cached_tokens: responseBody.usageMetadata.cachedContentTokenCount,
-      reasoning_tokens: responseBody.usageMetadata.thoughtsTokenCount
+      cached_tokens: responseBody.usageMetadata.cachedContentTokenCount || 0,
+      reasoning_tokens: responseBody.usageMetadata.thoughtsTokenCount || 0
     };
   }
 
@@ -86,16 +87,11 @@ export function saveUsageStats({ provider, model, tokens, connectionId, apiKey,
   const accountSuffix = connectionId ? ` | account=${connectionId.slice(0, 8)}...` : "";
   console.log(`${COLORS.green}[${time}] 📊 [${label}] ${provider.toUpperCase()} | in=${inTokens} | out=${outTokens}${accountSuffix}${COLORS.reset}`);
 
-  // Normalize to OpenAI token shape for storage
-  const cacheRead = tokens.cache_read_input_tokens || tokens.cached_tokens || tokens.prompt_tokens_details?.cached_tokens || 0;
-  const cacheCreation = tokens.cache_creation_input_tokens || 0;
-  const reasoning = tokens.reasoning_tokens || tokens.completion_tokens_details?.reasoning_tokens || 0;
-  const normalized = {
+  // Canonicalize to one storage convention (prompt_tokens cache-inclusive) so
+  // cached/cache-creation tokens survive to cost calc + stats. See canonicalizeUsage.
+  const normalized = canonicalizeUsage(tokens) || {
     prompt_tokens: tokens.prompt_tokens ?? tokens.input_tokens ?? 0,
     completion_tokens: tokens.completion_tokens ?? tokens.output_tokens ?? 0,
-    ...(cacheRead > 0 && { cache_read_input_tokens: cacheRead }),
-    ...(cacheCreation > 0 && { cache_creation_input_tokens: cacheCreation }),
-    ...(reasoning > 0 && { reasoning_tokens: reasoning }),
   };
 
   saveRequestUsage({

package/app/open-sse/handlers/chatCore/sseToJsonHandler.js

@@ -2,7 +2,11 @@ import { convertResponsesStreamToJson } from "../../transformer/streamToJsonConv
 import { createErrorResult } from "../../utils/error.js";
 import { HTTP_STATUS } from "../../config/runtimeConfig.js";
 import { FORMATS } from "../../translator/formats.js";
+import { PROVIDERS } from "../../config/providers.js";
 import { buildRequestDetail, extractRequestConfig, saveUsageStats } from "./requestDetail.js";
+
+// Responses-API providers (e.g. codex) may emit SSE without content-type + use Responses output shape
+const isResponsesProvider = (p) => PROVIDERS[p]?.format === FORMATS.OPENAI_RESPONSES;
 import { saveRequestDetail, appendRequestLog } from "@/lib/usageDb.js";
 
 function textFromResponsesMessageItem(item) {
@@ -100,7 +104,7 @@ export function parseSSEToOpenAIResponse(rawSSE, fallbackModel) {
  */
 export async function handleForcedSSEToJson({ providerResponse, sourceFormat, provider, model, body, stream, translatedBody, finalBody, requestStartTime, connectionId, apiKey, clientRawRequest, onRequestSuccess, trackDone, appendLog }) {
   const contentType = providerResponse.headers.get("content-type") || "";
-  const isSSE = contentType.includes("text/event-stream") || (contentType === "" && provider === "codex");
+  const isSSE = contentType.includes("text/event-stream") || (contentType === "" && isResponsesProvider(provider));
   if (!isSSE) return null; // not handled here
 
   trackDone();
@@ -112,7 +116,7 @@ export async function handleForcedSSEToJson({ providerResponse, sourceFormat, pr
   };
 
   // Codex/Responses API SSE path
-  const isCodexResponsesApi = provider === "codex" || sourceFormat === FORMATS.OPENAI_RESPONSES;
+  const isCodexResponsesApi = isResponsesProvider(provider) || sourceFormat === FORMATS.OPENAI_RESPONSES;
   if (isCodexResponsesApi) {
     try {
       const jsonResponse = await convertResponsesStreamToJson(providerResponse.body);

package/app/open-sse/handlers/chatCore/streamingHandler.js

@@ -2,15 +2,21 @@ import { FORMATS } from "../../translator/formats.js";
 import { needsTranslation } from "../../translator/index.js";
 import { createSSETransformStreamWithLogger, createPassthroughStreamWithLogger } from "../../utils/stream.js";
 import { pipeWithDisconnect } from "../../utils/streamHandler.js";
+import { PROVIDERS } from "../../config/providers.js";
+import { STREAM_STALL_TIMEOUT_MS } from "../../config/runtimeConfig.js";
 import { buildAbortedResponsesTerminalBytes } from "../../utils/responsesStreamHelpers.js";
 import { buildRequestDetail, extractRequestConfig, saveUsageStats } from "./requestDetail.js";
 import { saveRequestDetail } from "@/lib/usageDb.js";
+import { SSE_HEADERS_CORS as SSE_HEADERS } from "../../utils/sseConstants.js";
 
-const SSE_HEADERS = {
-  "Content-Type": "text/event-stream",
-  "Cache-Control": "no-cache",
-  "Connection": "keep-alive",
-  "Access-Control-Allow-Origin": "*"
+// Codex returns Responses API SSE → which client format to translate INTO, by request sourceFormat.
+// Gemini-family all map to ANTIGRAVITY decoder; unknown sources fall back to OPENAI.
+const CODEX_SOURCE_TO_TARGET = {
+  [FORMATS.OPENAI_RESPONSES]: FORMATS.OPENAI_RESPONSES,
+  [FORMATS.CLAUDE]: FORMATS.CLAUDE,
+  [FORMATS.ANTIGRAVITY]: FORMATS.ANTIGRAVITY,
+  [FORMATS.GEMINI]: FORMATS.ANTIGRAVITY,
+  [FORMATS.GEMINI_CLI]: FORMATS.ANTIGRAVITY,
 };
 
 /**
@@ -18,15 +24,12 @@ const SSE_HEADERS = {
  */
 function buildTransformStream({ provider, sourceFormat, targetFormat, userAgent, reqLogger, toolNameMap, model, connectionId, body, onStreamComplete, apiKey }) {
   const isDroidCLI = userAgent?.toLowerCase().includes("droid") || userAgent?.toLowerCase().includes("codex-cli");
-  const needsCodexTranslation = provider === "codex" && targetFormat === FORMATS.OPENAI_RESPONSES && !isDroidCLI;
+  // Responses-API providers (e.g. codex) emit Responses SSE → translate into client format
+  const isResponsesProvider = PROVIDERS[provider]?.format === FORMATS.OPENAI_RESPONSES;
+  const needsCodexTranslation = isResponsesProvider && targetFormat === FORMATS.OPENAI_RESPONSES && !isDroidCLI;
 
   if (needsCodexTranslation) {
-    // Codex returns Responses API SSE → translate to client format
-    let codexTarget;
-    if (sourceFormat === FORMATS.OPENAI_RESPONSES) codexTarget = FORMATS.OPENAI_RESPONSES;
-    else if (sourceFormat === FORMATS.CLAUDE) codexTarget = FORMATS.CLAUDE;
-    else if (sourceFormat === FORMATS.ANTIGRAVITY || sourceFormat === FORMATS.GEMINI || sourceFormat === FORMATS.GEMINI_CLI) codexTarget = FORMATS.ANTIGRAVITY;
-    else codexTarget = FORMATS.OPENAI;
+    const codexTarget = CODEX_SOURCE_TO_TARGET[sourceFormat] || FORMATS.OPENAI;
     return createSSETransformStreamWithLogger(FORMATS.OPENAI_RESPONSES, codexTarget, provider, reqLogger, toolNameMap, model, connectionId, body, onStreamComplete, apiKey);
   }
 
@@ -40,15 +43,48 @@ function buildTransformStream({ provider, sourceFormat, targetFormat, userAgent,
 /**
  * Handle streaming response — pipe provider SSE through transform stream to client.
  */
-export function handleStreamingResponse({ providerResponse, provider, model, sourceFormat, targetFormat, userAgent, body, stream, translatedBody, finalBody, requestStartTime, connectionId, apiKey, clientRawRequest, onRequestSuccess, reqLogger, toolNameMap, streamController, onStreamComplete, streamDetailId, streamWatchdogEnabled = true }) {
-  if (onRequestSuccess) onRequestSuccess();
+export async function handleStreamingResponse({ providerResponse, provider, model, sourceFormat, targetFormat, userAgent, body, stream, translatedBody, finalBody, requestStartTime, connectionId, apiKey, clientRawRequest, onRequestSuccess, reqLogger, toolNameMap, streamController, onStreamComplete, streamDetailId, streamWatchdogEnabled = true }) {
+  if (onRequestSuccess) {
+    Promise.resolve()
+      .then(onRequestSuccess)
+      .catch(err => {
+        console.error("[ChatCore] onRequestSuccess failed:", err?.message || err);
+      });
+  }
+
+  // When upstream returns HTML/text instead of SSE (e.g. Cloudflare 5xx error
+  // page), piping it through the SSE transform stream causes Next.js
+  // "failed to pipe response" and crashes the chat router. Read the body,
+  // pull a short human-readable message from the <title>, sanitize it, and
+  // return a clean JSON error instead. The message is stripped of HTML tags
+  // and clamped so untrusted upstream text never reaches the client verbatim
+  // (the UI may render error.message as HTML).
+  const upstreamContentType = (providerResponse.headers.get('content-type') || '').toLowerCase();
+  if (upstreamContentType && !upstreamContentType.includes('text/event-stream') && !upstreamContentType.includes('application/json')) {
+    const bodyText = await providerResponse.text().catch(() => '');
+    const titleMatch = bodyText.match(/<title>([^<]+)<\/title>/i);
+    const sanitizedTitle = (titleMatch?.[1] || '').replace(/<[^>]*>/g, '').replace(/[\r\n]+/g, ' ').trim().slice(0, 160);
+    const shortMsg = sanitizedTitle
+      || (bodyText.length < 200 ? bodyText.replace(/<[^>]*>/g, '').trim().slice(0, 160) : `Upstream returned non-SSE response (${upstreamContentType})`);
+    const status = providerResponse.status || 502;
+    console.warn(`[STREAM] ${provider} | ${model} | blocked pipe: ${shortMsg} [${status}]`);
+    streamController?.handleError?.(new Error(`upstream non-SSE: ${status}`));
+    return {
+      success: false,
+      response: new Response(JSON.stringify({ error: { message: `[${status}]: ${shortMsg}` } }), {
+        status,
+        headers: { 'Content-Type': 'application/json', 'Access-Control-Allow-Origin': '*' },
+      }),
+    };
+  }
 
   const transformStream = buildTransformStream({ provider, sourceFormat, targetFormat, userAgent, reqLogger, toolNameMap, model, connectionId, body, onStreamComplete, apiKey });
 
   // Responses passthrough: synthesize response.failed + [DONE] if the stream aborts/stalls before a terminal event
   const isResponsesPassthrough = sourceFormat === FORMATS.OPENAI_RESPONSES && targetFormat === FORMATS.OPENAI_RESPONSES;
   const onAbortTerminal = isResponsesPassthrough ? buildAbortedResponsesTerminalBytes : null;
-  const transformedBody = pipeWithDisconnect(providerResponse, transformStream, streamController, streamWatchdogEnabled, onAbortTerminal);
+  const stallTimeoutMs = PROVIDERS[provider]?.stallTimeoutMs || STREAM_STALL_TIMEOUT_MS;
+  const transformedBody = pipeWithDisconnect(providerResponse, transformStream, streamController, streamWatchdogEnabled, onAbortTerminal, stallTimeoutMs);
 
   saveRequestDetail(buildRequestDetail({
     provider, model, connectionId, apiKey,

package/app/open-sse/handlers/chatCore.js

@@ -1,11 +1,13 @@
-import { detectFormat, getTargetFormat } from "../services/provider.js";
+import { detectFormat, getTargetFormat, resolveTransport } from "../services/provider.js";
 import { translateRequest } from "../translator/index.js";
 import { FORMATS } from "../translator/formats.js";
+import { normalizeClaudePassthrough } from "../translator/formats/claude.js";
 import { COLORS } from "../utils/stream.js";
 import { createStreamController } from "../utils/streamHandler.js";
 import { refreshWithRetry } from "../services/tokenRefresh.js";
 import { createRequestLogger } from "../utils/requestLogger.js";
 import { getModelTargetFormat, getModelStrip, getModelUpstreamId, getModelType, PROVIDER_ID_TO_ALIAS } from "../config/providerModels.js";
+import { PROVIDERS } from "../config/providers.js";
 import { createErrorResult, parseUpstreamError, formatProviderError } from "../utils/error.js";
 import { HTTP_STATUS } from "../config/runtimeConfig.js";
 import { handleBypassRequest } from "../utils/bypassHandler.js";
@@ -18,7 +20,12 @@ import { handleStreamingResponse, buildOnStreamComplete } from "./chatCore/strea
 import { detectClientTool, isNativePassthrough } from "../utils/clientDetector.js";
 import { dedupeTools } from "../utils/toolDeduper.js";
 import { injectCaveman } from "../rtk/caveman.js";
+import { injectPonytail } from "../rtk/ponytail.js";
 import { compressMessages, formatRtkLog } from "../rtk/index.js";
+import { compressWithHeadroom, formatHeadroomLog, formatHeadroomSizeLog, isHeadroomPhantomSavings } from "../rtk/headroom.js";
+import { getCapabilitiesForModel } from "../providers/capabilities.js";
+import { stripUnsupportedModalities } from "../translator/concerns/modality.js";
+import { prefetchRemoteImages } from "../translator/concerns/prefetch.js";
 
 /**
  * Core chat handler - shared between SSE and Worker
@@ -27,7 +34,7 @@ import { compressMessages, formatRtkLog } from "../rtk/index.js";
  * @param {object} options.credentials - Provider credentials
  * @param {string} options.sourceFormatOverride - Override detected source format (e.g. "openai-responses")
  */
-export async function handleChatCore({ body, modelInfo, credentials, log, onCredentialsRefreshed, onRequestSuccess, onDisconnect, clientRawRequest, connectionId, userAgent, apiKey, ccFilterNaming, rtkEnabled, cavemanEnabled, cavemanLevel, sourceFormatOverride, providerThinking, streamWatchdogEnabled = true }) {
+export async function handleChatCore({ body, modelInfo, credentials, log, onCredentialsRefreshed, onRequestSuccess, onDisconnect, clientRawRequest, connectionId, userAgent, apiKey, ccFilterNaming, rtkEnabled, headroomEnabled, headroomUrl, headroomCompressUserMessages, cavemanEnabled, cavemanLevel, ponytailEnabled, ponytailLevel, sourceFormatOverride, providerThinking, streamWatchdogEnabled = true }) {
   const { provider, model } = modelInfo;
   const requestStartTime = Date.now();
 
@@ -39,7 +46,10 @@ export async function handleChatCore({ body, modelInfo, credentials, log, onCred
 
   const alias = PROVIDER_ID_TO_ALIAS[provider] || provider;
   const modelTargetFormat = getModelTargetFormat(alias, model);
-  const targetFormat = modelTargetFormat || getTargetFormat(provider);
+  // Multi-endpoint providers: pick transport matching sourceFormat → zero translation
+  const runtimeTransport = resolveTransport(provider, sourceFormat);
+  const targetFormat = modelTargetFormat || runtimeTransport?.format || getTargetFormat(provider);
+  if (runtimeTransport && credentials) credentials.runtimeTransport = runtimeTransport;
   const stripList = getModelStrip(alias, model);
   const upstreamModel = getModelUpstreamId(alias, model);
 
@@ -58,9 +68,16 @@ export async function handleChatCore({ body, modelInfo, credentials, log, onCred
   }
 
   const clientRequestedStreaming = body.stream === true || sourceFormat === FORMATS.ANTIGRAVITY || sourceFormat === FORMATS.GEMINI || sourceFormat === FORMATS.GEMINI_CLI;
-  const providerRequiresStreaming = provider === "openai" || provider === "codex" || provider === "commandcode";
+  const providerRequiresStreaming = PROVIDERS[provider]?.forceStream === true;
   let stream = providerRequiresStreaming ? true : (body.stream !== false);
 
+  // Image generation models require non-streaming (Google v1internal:generateContent)
+  const modelType = getModelType(alias, model);
+  const isImageGenModel = modelType === "imageGen" || /image|imagen|image-generation/i.test(model);
+  if (isImageGenModel && (provider === "antigravity" || provider === "gemini-cli")) {
+    stream = false;
+  }
+
   // DeepSeek-TUI: interactive TUI panel sends stream:true and needs SSE.
   // Non-interactive mode (-p flag) sends without stream and can't parse SSE.
   // Only force non-streaming when client didn't explicitly request it.
@@ -72,7 +89,7 @@ export async function handleChatCore({ body, modelInfo, credentials, log, onCred
   const acceptHeader = clientRawRequest?.headers?.accept || "";
   const clientPrefersJson = acceptHeader.includes("application/json");
   const clientPrefersSSE = acceptHeader.includes("text/event-stream");
-  if (clientPrefersJson && !clientPrefersSSE && body.stream !== true) {
+  if (clientPrefersJson && !clientPrefersSSE && body.stream !== true && !providerRequiresStreaming) {
     stream = false;
   }
 
@@ -86,11 +103,29 @@ export async function handleChatCore({ body, modelInfo, credentials, log, onCred
   const clientTool = detectClientTool(clientRawRequest?.headers || {}, body);
   const passthrough = isNativePassthrough(clientTool, provider);
 
+  // Expose raw client headers to translators/executors for session-id resolution
+  if (credentials) credentials.rawHeaders = clientRawRequest?.headers || {};
+
+  // Auto-strip media blocks the model can't read (vision/audio/pdf) before translation.
+  if (!passthrough) {
+    const caps = getCapabilitiesForModel(provider, model);
+    if (stripUnsupportedModalities(body, sourceFormat, caps)) {
+      log?.debug?.("MODALITY", `stripped unsupported media for ${provider}/${model}`);
+    }
+    // Convert remote image URLs to base64 for targets that can't fetch URLs.
+    try {
+      const n = await prefetchRemoteImages(body, sourceFormat, targetFormat, { signal: undefined });
+      if (n > 0) log?.debug?.("MODALITY", `prefetched ${n} remote image(s) for ${targetFormat}`);
+    } catch (e) { log?.warn?.("MODALITY", `image prefetch failed: ${e.message}`); }
+  }
+
   let translatedBody;
   let toolNameMap;
   if (passthrough) {
     log?.debug?.("PASSTHROUGH", `${clientTool} → ${provider} | native lossless`);
     translatedBody = { ...body, model: upstreamModel };
+    // Normalize newer Cowork/CC beta shapes (adaptive thinking, mid-conversation system) the API rejects
+    if (clientTool === "claude") normalizeClaudePassthrough(translatedBody, upstreamModel);
   } else {
     translatedBody = translateRequest(sourceFormat, targetFormat, upstreamModel, body, stream, credentials, provider, reqLogger, stripList, connectionId, clientTool);
     if (!translatedBody) {
@@ -126,12 +161,30 @@ export async function handleChatCore({ body, modelInfo, credentials, log, onCred
   const rtkLine = formatRtkLog(rtkStats);
   if (rtkLine) console.log(rtkLine);
 
+  // Headroom: optional external proxy compression; fail open if proxy is absent.
+  const headroomDiagnostics = {};
+  const headroomStats = await compressWithHeadroom(translatedBody, { enabled: headroomEnabled, url: headroomUrl, model: upstreamModel, format: finalFormat, compressUserMessages: headroomCompressUserMessages, diagnostics: headroomDiagnostics });
+  const headroomLine = formatHeadroomLog(headroomStats);
+  const headroomSizeLine = formatHeadroomSizeLog(headroomDiagnostics);
+  if (headroomLine) {
+    log?.info?.("HEADROOM", `${headroomLine}${headroomSizeLine ? ` | ${headroomSizeLine}` : ""}`);
+    if (isHeadroomPhantomSavings(headroomStats, headroomDiagnostics)) {
+      log?.warn?.("HEADROOM", `reported token delta, but outbound JSON shrank <5%; provider may bill near-original payload | ${headroomSizeLine}`);
+    }
+  } else if (headroomEnabled) log?.warn?.("HEADROOM", `skipped: ${headroomDiagnostics.reason || "compression unavailable"}${headroomDiagnostics.endpoint ? ` (${headroomDiagnostics.endpoint})` : ""}`);
+
   // Caveman: inject terse-style system prompt
   if (cavemanEnabled && cavemanLevel) {
     injectCaveman(translatedBody, finalFormat, cavemanLevel);
     log?.debug?.("CAVEMAN", `${cavemanLevel} | ${finalFormat}`);
   }
 
+  // Ponytail: inject lazy-senior-dev system prompt
+  if (ponytailEnabled && ponytailLevel) {
+    injectPonytail(translatedBody, finalFormat, ponytailLevel);
+    log?.debug?.("PONYTAIL", `${ponytailLevel} | ${finalFormat}`);
+  }
+
   const executor = getExecutor(provider);
   trackPendingRequest(model, provider, connectionId, true);
   appendRequestLog({ model, provider, connectionId, status: "PENDING" }).catch(() => { });

package/app/open-sse/handlers/embeddingProviders/index.js

@@ -6,6 +6,7 @@ import openaiCompatNode from "./openaiCompatNode.js";
 const OPENAI_COMPAT_PROVIDERS = [
   "openai", "openrouter", "mistral", "voyage-ai", "fireworks",
   "together", "nebius", "github", "nvidia", "jina-ai",
+  "vercel-ai-gateway",
 ];
 
 const ADAPTERS = {