Never stop coding. Free MIT AI gateway: one endpoint, 375 providers (150+ free), 1,400+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 900+ contributors
TypeScript a2aai-agentsai-gatewayanthropicclaudeclaude-codeclinecodexcopilotcursordeepseekfree-aigeminikimillm-gatewaymcpopenaiopenai-proxyqwentoken-saver
fix(routing): honor client reasoning.effort for gpt-5.5 + route suffixed variants to codex (#2877)
diegosouzapwf5d74cd76dfd
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex e52c402baf..cf4a57adb4 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -59,6 +59,14 @@### Fixed+- **routing/codex:** fix two gpt-5.5 Codex defects (#2877). (A) For a Codex-only+ account, a bare `gpt-5.5` Responses request was rerouted to codex with the+ model hardcoded to `gpt-5.5-medium` (`chatHelpers.ts`); the executor read that+ `-medium` suffix as an explicit `modelEffort` that (per #2331) overrode a+ client `reasoning.effort=xhigh`, silently demoting it — now it keeps the bare+ `gpt-5.5` id so the client effort wins. (B) `gpt-5.5-xhigh`/`-high`/`-low`+ misrouted to `openai` (→ "No credentials" for codex-only users); the suffixed+ variants are now in `CODEX_PREFERRED_UNPREFIXED_MODELS` so they infer codex.- **sse/chatCore:** remove a duplicate `const settings` declaration in`handleChatCore` (introduced alongside the per-key stream-default-modefeature). The same-scope redeclaration made esbuild/tsx fail withdiff --git a/open-sse/services/model.ts b/open-sse/services/model.tsindex 5c15c0ac84..2f3ca61d1a 100644--- a/open-sse/services/model.ts+++ b/open-sse/services/model.ts@@ -112,7 +112,16 @@ for (const [aliasOrId, models] of Object.entries(PROVIDER_MODELS)) {}}const KNOWN_MODEL_IDS = new Set(MODEL_TO_PROVIDERS.keys());-const CODEX_PREFERRED_UNPREFIXED_MODELS = new Set(["gpt-5.5"]);+// #2877(B): include the effort-suffixed variants so a bare `gpt-5.5-xhigh`+// (and -high/-medium/-low) infers the codex provider instead of falling through+// the `/^gpt-/` → openai fallback (which 500s for codex-only credentials).+const CODEX_PREFERRED_UNPREFIXED_MODELS = new Set([+ "gpt-5.5",+ "gpt-5.5-xhigh",+ "gpt-5.5-high",+ "gpt-5.5-medium",+ "gpt-5.5-low",+]);const CODEX_PREFERRED_UNPREFIXED_MODEL_ALIASES = new Map([["gpt-5.5", "gpt-5.5-medium"]]);export const CODEX_NATIVE_UNPREFIXED_MODELS = new Set(["codex-auto-review"]);diff --git a/src/sse/handlers/chatHelpers.ts b/src/sse/handlers/chatHelpers.tsindex 15e4bb2d7f..ebaa542927 100644--- a/src/sse/handlers/chatHelpers.ts+++ b/src/sse/handlers/chatHelpers.ts@@ -130,9 +130,14 @@ export async function resolveModelOrError(!isCodexNativeResponsesRequest(body, endpointPath, requestHeaders) &&(await hasOnlyActiveCodexAccount())) {- log.info("ROUTING", `${modelStr} → codex/gpt-5.5-medium (Codex-only active account)`);+ // #2877: keep the bare model id (do NOT bake a `-medium` suffix). The Codex+ // executor reads a model-name suffix as an explicit `modelEffort` that (per+ // #2331) overrides the client's `reasoning.effort`, so injecting `-medium`+ // here silently demoted a genuine `reasoning.effort=xhigh`. The default+ // effort still comes from the connection fallback when the client sends none.+ log.info("ROUTING", `${modelStr} → codex/gpt-5.5 (Codex-only active account)`);modelInfo.provider = "codex";- modelInfo.model = "gpt-5.5-medium";+ modelInfo.model = "gpt-5.5";}// Forced-rewrite: codex provider doesn't serve DeepSeek/Qwen/Kimi/etc. Reroutediff --git a/tests/unit/codex-gpt55-effort-routing.test.ts b/tests/unit/codex-gpt55-effort-routing.test.tsnew file mode 100644index 0000000000..17b0abdb40--- /dev/null+++ b/tests/unit/codex-gpt55-effort-routing.test.ts@@ -0,0 +1,71 @@+/**+ * Issue #2877 — two routing/effort defects for the gpt-5.5 Codex family:+ *+ * (B) `gpt-5.5-xhigh` (and -high/-low) misrouted to the `openai` provider+ * because only bare `gpt-5.5` was in CODEX_PREFERRED_UNPREFIXED_MODELS — the+ * suffixed variants fell through to the `/^gpt-/` → openai fallback, so a+ * Codex-OAuth-only user got "No credentials for provider: openai". Fixed by+ * adding the variants to the set (`open-sse/services/model.ts`).+ *+ * (A) For a Codex-only account, a bare `gpt-5.5` Responses request was rerouted+ * to codex but with the model hardcoded to `gpt-5.5-medium`+ * (`src/sse/handlers/chatHelpers.ts`). The Codex executor reads that `-medium`+ * suffix as an explicit `modelEffort`, which (per #2331) overrides the+ * client's `reasoning.effort=xhigh` — silently demoting it. Fixed by keeping+ * the bare `gpt-5.5` id; the executor's modelEffort-first precedence (#2331)+ * is left untouched.+ */+import test from "node:test";+import assert from "node:assert/strict";+import fs from "node:fs";+import os from "node:os";+import path from "node:path";++const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-gpt55-routing-"));+process.env.DATA_DIR = TEST_DATA_DIR;++const core = await import("../../src/lib/db/core.ts");+const providersDb = await import("../../src/lib/db/providers.ts");+const { getModelInfoCore } = await import("../../open-sse/services/model.ts");+const { resolveModelOrError } = await import("../../src/sse/handlers/chatHelpers.ts");++test.before(async () => {+ // Codex-only active account (no openai connection).+ await providersDb.createProviderConnection({+ provider: "codex",+ authType: "oauth",+ email: "[email protected]",+ providerSpecificData: { workspaceId: "ws-1" },+ });+});++test.after(() => {+ core.resetDbInstance();+ fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });+});++// ── Defect B: suffixed bare names infer codex, not openai ─────────────────────+for (const variant of ["gpt-5.5-xhigh", "gpt-5.5-high", "gpt-5.5-low"]) {+ test(`#2877(B) ${variant} infers codex (not openai)`, async () => {+ const info = await getModelInfoCore(variant, null);+ assert.equal(info.provider, "codex", `${variant} must infer the codex provider`);+ assert.equal(info.model, variant, "the explicit effort suffix must be preserved");+ });+}++// ── Defect A: Codex-only bare gpt-5.5 reroute must NOT bake a -medium suffix ───+test("#2877(A) Codex-only bare gpt-5.5 Responses request keeps the bare model id", async () => {+ const result = (await resolveModelOrError(+ "gpt-5.5",+ { input: [{ role: "user", content: [{ type: "input_text", text: "hi" }] }] },+ "/v1/responses",+ null+ )) as { provider?: string; model?: string };++ assert.equal(result.provider, "codex", "Codex-only account must reroute gpt-5.5 to codex");+ assert.equal(+ result.model,+ "gpt-5.5",+ "must NOT inject a -medium suffix (that would override a client reasoning.effort=xhigh)"+ );+});