Fix Antigravity streams and Claude Code route fallback

CHANGELOG.md

@@ -9,6 +9,14 @@ Release tags use the form `vX.Y.Z` and match `package.json`. GitHub Releases car
 
 ## [Unreleased]
 
+## [0.5.10] - 2026-07-29
+
+### Fixed
+
+- Antigravity Flash and Pro streaming responses now accept Google's CRLF-delimited SSE frames instead of completing with empty output.
+- Claude Code's canonical Fable and Claude Opus 4.8 model IDs now resolve through matching `fable` and `opus-4.8` named routes, preserving configured cross-provider fallback.
+- Anthropic overload responses now create short model-scoped transient cooldowns instead of misleading quota-wide account locks.
+
 ## [0.5.9] - 2026-07-23
 
 ### Fixed
@@ -100,7 +108,8 @@ Release tags use the form `vX.Y.Z` and match `package.json`. GitHub Releases car
 
 See [GitHub Releases](https://github.com/gitcommit90/rerouted/releases) for artifact digests and notes prior to the Keep a Changelog narrative. Notable themes in late 0.4.x included signed/notarized distribution, in-app updates, named routes, OAuth account pools, OpenAI chat completions and Responses routing, and launch hardening.
 
-[Unreleased]: https://github.com/gitcommit90/rerouted/compare/v0.5.9...HEAD
+[Unreleased]: https://github.com/gitcommit90/rerouted/compare/v0.5.10...HEAD
+[0.5.10]: https://github.com/gitcommit90/rerouted/releases/tag/v0.5.10
 [0.5.9]: https://github.com/gitcommit90/rerouted/releases/tag/v0.5.9
 [0.5.8]: https://github.com/gitcommit90/rerouted/releases/tag/v0.5.8
 [0.5.7]: https://github.com/gitcommit90/rerouted/releases/tag/v0.5.7

package.json

@@ -1,7 +1,7 @@
 {
   "name": "@gitcommit90/rerouted",
   "productName": "ReRouted",
-  "version": "0.5.9",
+  "version": "0.5.10",
   "description": "A local AI router for connected accounts, models, named routes, and automatic fallback.",
   "author": "gitcommit90",
   "license": "MIT",

src/lib/router.js

@@ -25,6 +25,10 @@ const COOLDOWN_MS = {
   transient: 30_000,
 };
 const PREOUTPUT_INSPECTION_BYTES = 64 * 1024;
+const CLAUDE_CODE_CANONICAL_ROUTES = new Map([
+  ["claude-fable-5", "fable"],
+  ["claude-opus-4-8", "opus-4.8"],
+]);
 
 function createRequestLog(max = 50) {
   const items = [];
@@ -94,8 +98,26 @@ function makeMember(cfg, provider, upstreamModel, opts) {
   };
 }
 
+function canonicalRouteModelId(cfg, requestedModelId) {
+  const requested = String(requestedModelId || "").trim();
+  const upstreamModel = requested.startsWith("claude/")
+    ? requested.slice("claude/".length)
+    : requested;
+
+  // Account-qualified IDs must continue to address that exact account.
+  if (requested !== upstreamModel && upstreamModel.includes("/")) return requested;
+
+  const routeName = CLAUDE_CODE_CANONICAL_ROUTES.get(upstreamModel.toLowerCase());
+  if (!routeName) return requested;
+  const combo = (cfg.combos || []).find(
+    (entry) => publicComboId(entry).trim().toLowerCase() === routeName
+  );
+  return combo ? publicComboId(combo) : requested;
+}
+
 function resolveTargets(cfg, modelId) {
-  const combo = (cfg.combos || []).find((c) => comboMatchesId(c, modelId));
+  const routeModelId = canonicalRouteModelId(cfg, modelId);
+  const combo = (cfg.combos || []).find((c) => comboMatchesId(c, routeModelId));
   if (combo) {
     const members = (combo.members || [])
       .map((m) => {
@@ -134,7 +156,7 @@ function resolveTargets(cfg, modelId) {
       members,
     };
   }
-  const single = resolveSingle(cfg, modelId);
+  const single = resolveSingle(cfg, routeModelId);
   if (!single) return null;
   return { kind: "single", members: [single], strategy: "fallback" };
 }
@@ -243,7 +265,7 @@ function classifyFailure(status, errorText) {
   const text = String(errorText || "").toLowerCase();
   const quota =
     status === 429 ||
-    /rate[ _-]?limit|too many requests|quota|usage[ _-]?limit|resource[ _-]?exhaust|capacity|overload/.test(text);
+    /rate[ _-]?limit|too many requests|quota|usage[ _-]?limit|resource[ _-]?exhaust|capacity/.test(text);
   if (quota) return { eligible: true, kind: "quota", defaultCooldownMs: COOLDOWN_MS.quota };
   const request =
     /context[ _-]?window|context[ _-]?length[ _-]?exceed|maximum[ _-]?context[ _-]?length|input.{0,80}(?:too[ _-]?long|too[ _-]?large|exceed.{0,40}(?:context|token))|request[ _-]?too[ _-]?large/.test(
@@ -1438,6 +1460,7 @@ module.exports = {
   resolveTargets,
   resolveSingle,
   accountCandidatesFor,
+  canonicalRouteModelId,
   compareAccounts,
   orderMembers,
   isRetryableStatus,

src/lib/sse.js

@@ -98,13 +98,13 @@ function createSseParser() {
     push(chunk) {
       buf += chunkToString(chunk);
       const events = [];
-      let idx;
-      while ((idx = buf.indexOf("\n\n")) !== -1) {
-        const block = buf.slice(0, idx);
-        buf = buf.slice(idx + 2);
+      let boundary;
+      while ((boundary = /\r?\n\r?\n/.exec(buf))) {
+        const block = buf.slice(0, boundary.index);
+        buf = buf.slice(boundary.index + boundary[0].length);
         let event = "message";
         let data = "";
-        for (const line of block.split("\n")) {
+        for (const line of block.split(/\r?\n/)) {
           if (line.startsWith("event:")) event = line.slice(6).trim();
           else if (line.startsWith("data:")) data += (data ? "\n" : "") + line.slice(5).trimStart();
         }

tests/antigravity-tools.test.js

@@ -299,6 +299,50 @@ describe("Antigravity tool calling", () => {
     assert.equal(body.request.contents[1].parts[1].thoughtSignature, undefined);
   });
 
+  it("parses CRLF-delimited Gemini SSE text, finish state, and usage", async () => {
+    const frames = [
+      {
+        response: {
+          candidates: [
+            {
+              content: { role: "model", parts: [{ text: "OK" }] },
+              finishReason: "STOP",
+            },
+          ],
+          usageMetadata: {
+            promptTokenCount: 3,
+            candidatesTokenCount: 1,
+            totalTokenCount: 4,
+            cachedContentTokenCount: 2,
+          },
+        },
+      },
+    ].map((payload) => `data: ${JSON.stringify(payload)}\r\n\r\n`);
+    const writes = [];
+
+    const usage = await antigravity.pipeGeminiSse(
+      Readable.from(frames),
+      { write(chunk) { writes.push(chunk); } },
+      "gemini-3-flash-agent"
+    );
+
+    const chunks = parseSseWrites(writes);
+    assert.equal(
+      chunks
+        .map((chunk) => chunk.choices[0].delta.content)
+        .filter(Boolean)
+        .join(""),
+      "OK"
+    );
+    assert.equal(chunks.at(-1).choices[0].finish_reason, "stop");
+    assert.deepEqual(usage, {
+      prompt_tokens: 3,
+      completion_tokens: 1,
+      total_tokens: 4,
+      prompt_tokens_details: { cached_tokens: 2 },
+    });
+  });
+
   it("deduplicates cumulative Gemini SSE text and buffers the final signed tool call", async () => {
     const events = [
       `data: ${JSON.stringify({

tests/gateway.test.js

@@ -548,6 +548,78 @@ describe("gateway Responses API", () => {
       await new Promise((resolve) => server.close(resolve));
     }
   });
+
+  it("streams visible Antigravity text through Responses from CRLF Gemini SSE", async () => {
+    const store = createStore(tmpConfig());
+    store.seed({
+      providers: [
+        {
+          id: "prov_antigravity",
+          type: "antigravity",
+          name: "Antigravity",
+          accessToken: "antigravity-token",
+          projectId: "test-project",
+          enabled: true,
+          createdAt: 100,
+          models: [
+            { id: "gemini-3-flash-agent", name: "Gemini 3 Flash", enabled: true },
+          ],
+        },
+      ],
+    });
+    const apiKey = store.load().apiKey;
+    let upstreamCalls = 0;
+    const router = createRouter({
+      store,
+      fetchImpl: async () => {
+        upstreamCalls += 1;
+        return new Response(
+          `data: ${JSON.stringify({
+            response: {
+              candidates: [
+                {
+                  content: { role: "model", parts: [{ text: "OK" }] },
+                  finishReason: "STOP",
+                },
+              ],
+              usageMetadata: {
+                promptTokenCount: 3,
+                candidatesTokenCount: 1,
+                totalTokenCount: 4,
+              },
+            },
+          })}\r\n\r\n`,
+          { status: 200, headers: { "Content-Type": "text/event-stream" } }
+        );
+      },
+    });
+    const gateway = createGateway({ store, router });
+    const server = http.createServer((req, res) => gateway.handle(req, res));
+    await new Promise((resolve) => server.listen(0, "127.0.0.1", resolve));
+
+    try {
+      const response = await fetch(`http://127.0.0.1:${server.address().port}/v1/responses`, {
+        method: "POST",
+        headers: { Authorization: `Bearer ${apiKey}`, "Content-Type": "application/json" },
+        body: JSON.stringify({
+          model: "antigravity/gemini-3-flash-agent",
+          input: "Reply with OK",
+          stream: true,
+        }),
+      });
+      const text = await response.text();
+
+      assert.equal(response.status, 200);
+      assert.equal(upstreamCalls, 1);
+      assert.match(text, /event: response\.output_text\.delta/);
+      assert.match(text, /"delta":"OK"/);
+      assert.match(text, /event: response\.completed/);
+      assert.match(text, /"input_tokens":3/);
+      assert.match(text, /"output_tokens":1/);
+    } finally {
+      await new Promise((resolve) => server.close(resolve));
+    }
+  });
 });
 
 describe("gateway request limits", () => {