fix(github-copilot): preserve reasoning IDs for Copilot Codex models (#71684)

* fix(github-copilot): preserve all reasoning IDs and add gpt-5.3-codex support

The existing guard (8fd15ed0e5) only skipped rewriting reasoning item IDs
when encrypted_content was a non-null string. When gpt-5.3-codex is used
via GitHub Copilot, the model falls through to the forward-compat catch-all
with reasoning:false, so encrypted_content is never requested and arrives
as null — bypassing the guard and causing a rewrite. Copilot validates
reasoning item IDs server-side regardless of whether the client includes
encrypted_content, so the rewritten id triggers the 400 error.

Two changes:

1. connection-bound-ids.ts: skip ALL reasoning items unconditionally.
   Reasoning items always reference server-side state bound to their
   original ID; rewriting any of them breaks Copilot's lookup.

2. models.ts + index.ts: extend the forward-compat cloning logic to
   cover gpt-5.3-codex (adds it to the template-target set and to
   CODEX_TEMPLATE_MODEL_IDS so it can also serve as a template source
   for gpt-5.4). Adds gpt-5.3-codex to COPILOT_XHIGH_MODEL_IDS for
   the thinking profile.

Thanks @InvalidPandaa.

* docs(github-copilot): clarify gpt-5.3-codex is a no-op template for itself

https://claude.ai/code/session_01EAFmq4WyKkiUkVAqRXp4Bm

* fix(github-copilot): remove dead reasoning prefix branch in deriveReplacementId

https://claude.ai/code/session_01EAFmq4WyKkiUkVAqRXp4Bm

* fix(github-copilot): align reasoning id replay tests

* test(plugin-sdk): use cjs sidecar for require fast path

---------

Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
This commit is contained in:
InvalidPanda ツ
2026-04-25 21:52:07 +02:00
committed by GitHub
parent 1d49b8cdaa
commit b64bfc5d9a
8 changed files with 101 additions and 36 deletions

View File

@@ -179,6 +179,7 @@ Docs: https://docs.openclaw.ai
- OpenAI/Codex image generation: canonicalize legacy `openai-codex.baseUrl` values such as `https://chatgpt.com/backend-api` to the Codex Responses backend before calling `gpt-image-2`, matching the chat transport. Fixes #71460. Thanks @GodsBoy.
- Control UI: make `/usage` use the fresh context snapshot for context percentage, and include cache-write tokens in the Usage overview cache-hit denominator. Fixes #47885. Thanks @imwyvern and @Ante042.
- GitHub Copilot: preserve encrypted Responses reasoning item IDs during replay so Copilot can validate encrypted reasoning payloads across requests. (#71448) Thanks @a410979729-sys.
- GitHub Copilot: never rewrite connection-bound reasoning item IDs regardless of whether `encrypted_content` is present, fixing a 400 "Encrypted content item_id did not match" error with `gpt-5.3-codex` and future Codex models that fall through to the forward-compat catch-all with `reasoning: false`. Also recognize Codex-named models as reasoning-capable so they inherit the correct capability flags. Refs #68735. Thanks @InvalidPandaa.
- Agents/replies: recover final-answer text when streamed assistant chunks contain only whitespace, preventing completed turns from surfacing as empty-payload errors. Fixes #71454. (#71467) Thanks @Sanjays2402.
- Feishu/TTS: transcode voice-intent MP3 and other audio replies to Ogg/Opus before sending native Feishu audio bubbles, while keeping ordinary MP3 attachments as files. Fixes #61249 and #37868. Thanks @sg1416-zg and @ycjlb2023-peteryi.
- WhatsApp/TTS: transcode MP3/WebM audio, including Microsoft Edge TTS output, to Ogg/Opus before sending PTT voice notes.

View File

@@ -5,14 +5,14 @@ import {
} from "./connection-bound-ids.js";
describe("github-copilot connection-bound response IDs", () => {
it("rewrites opaque response item IDs deterministically", () => {
const originalId = Buffer.from(`reasoning-${"x".repeat(24)}`).toString("base64");
const first = [{ id: originalId, type: "reasoning" }];
const second = [{ id: originalId, type: "reasoning" }];
it("rewrites opaque message response item IDs deterministically", () => {
const originalId = Buffer.from(`message-${"x".repeat(24)}`).toString("base64");
const first = [{ id: originalId, type: "message" }];
const second = [{ id: originalId, type: "message" }];
expect(rewriteCopilotConnectionBoundResponseIds(first)).toBe(true);
expect(rewriteCopilotConnectionBoundResponseIds(second)).toBe(true);
expect(first[0]?.id).toMatch(/^rs_[a-f0-9]{16}$/);
expect(first[0]?.id).toMatch(/^msg_[a-f0-9]{16}$/);
expect(first[0]?.id).toBe(second[0]?.id);
});
@@ -35,26 +35,20 @@ describe("github-copilot connection-bound response IDs", () => {
expect(input[4]?.id).toMatch(/^msg_[a-f0-9]{16}$/);
});
it("preserves reasoning IDs when encrypted_content is present", () => {
const originalId = Buffer.from(`reasoning-${"e".repeat(24)}`).toString("base64");
it("preserves reasoning IDs regardless of encrypted_content", () => {
const withEncrypted = Buffer.from(`reasoning-${"e".repeat(24)}`).toString("base64");
const withNull = Buffer.from(`reasoning-${"n".repeat(24)}`).toString("base64");
const withoutField = Buffer.from(`reasoning-${"a".repeat(24)}`).toString("base64");
const input = [
{
id: originalId,
type: "reasoning",
encrypted_content: "opaque-encrypted-payload",
},
{ id: withEncrypted, type: "reasoning", encrypted_content: "opaque-encrypted-payload" },
{ id: withNull, type: "reasoning", encrypted_content: null },
{ id: withoutField, type: "reasoning" },
];
expect(rewriteCopilotConnectionBoundResponseIds(input)).toBe(false);
expect(input[0]?.id).toBe(originalId);
});
it("still rewrites reasoning IDs when encrypted_content is absent", () => {
const originalId = Buffer.from(`reasoning-${"n".repeat(24)}`).toString("base64");
const input = [{ id: originalId, type: "reasoning" }];
expect(rewriteCopilotConnectionBoundResponseIds(input)).toBe(true);
expect(input[0]?.id).toMatch(/^rs_[a-f0-9]{16}$/);
expect(input[0]?.id).toBe(withEncrypted);
expect(input[1]?.id).toBe(withNull);
expect(input[2]?.id).toBe(withoutField);
});
it("patches response payload input arrays only", () => {

View File

@@ -18,7 +18,7 @@ function looksLikeConnectionBoundId(id: string): boolean {
}
function deriveReplacementId(type: string | undefined, originalId: string): string {
const prefix = type === "reasoning" ? "rs" : type === "function_call" ? "fc" : "msg";
const prefix = type === "function_call" ? "fc" : "msg";
const hex = createHash("sha256").update(originalId).digest("hex").slice(0, 16);
return `${prefix}_${hex}`;
}
@@ -35,7 +35,11 @@ export function rewriteCopilotConnectionBoundResponseIds(input: unknown): boolea
if (typeof id !== "string" || id.length === 0) {
continue;
}
if (item.type === "reasoning" && typeof item.encrypted_content === "string") {
// Reasoning items always reference server-side encrypted state bound to the
// original item ID. Rewriting the ID — even when encrypted_content is absent
// or null — breaks Copilot's server-side lookup and causes a 400 validation
// failure regardless of whether the client included encrypted_content.
if (item.type === "reasoning") {
continue;
}
if (looksLikeConnectionBoundId(id)) {

View File

@@ -12,7 +12,7 @@ import { buildGithubCopilotReplayPolicy } from "./replay-policy.js";
import { wrapCopilotProviderStream } from "./stream.js";
const COPILOT_ENV_VARS = ["COPILOT_GITHUB_TOKEN", "GH_TOKEN", "GITHUB_TOKEN"];
const COPILOT_XHIGH_MODEL_IDS = ["gpt-5.4", "gpt-5.2", "gpt-5.2-codex"] as const;
const COPILOT_XHIGH_MODEL_IDS = ["gpt-5.4", "gpt-5.3-codex", "gpt-5.2", "gpt-5.2-codex"] as const;
type GithubCopilotPluginConfig = {
discovery?: {

View File

@@ -135,6 +135,50 @@ describe("resolveCopilotForwardCompatModel", () => {
expect((result as unknown as Record<string, unknown>).reasoning).toBe(true);
});
it("clones gpt-5.3-codex template for gpt-5.3-codex when not in registry", () => {
const template = {
id: "gpt-5.2-codex",
name: "gpt-5.2-codex",
provider: "github-copilot",
api: "openai-responses",
reasoning: true,
contextWindow: 200_000,
};
const ctx = createMockCtx("gpt-5.3-codex", {
"github-copilot/gpt-5.2-codex": template,
});
const result = requireResolvedModel(ctx);
expect(result.id).toBe("gpt-5.3-codex");
expect(result.name).toBe("gpt-5.3-codex");
expect((result as unknown as Record<string, unknown>).reasoning).toBe(true);
});
it("prefers gpt-5.3-codex as template source over gpt-5.2-codex for gpt-5.4", () => {
const template53 = {
id: "gpt-5.3-codex",
name: "gpt-5.3-codex",
provider: "github-copilot",
api: "openai-responses",
reasoning: true,
contextWindow: 300_000,
};
const template52 = {
id: "gpt-5.2-codex",
name: "gpt-5.2-codex",
provider: "github-copilot",
api: "openai-responses",
reasoning: true,
contextWindow: 200_000,
};
const ctx = createMockCtx("gpt-5.4", {
"github-copilot/gpt-5.3-codex": template53,
"github-copilot/gpt-5.2-codex": template52,
});
const result = requireResolvedModel(ctx);
expect(result.id).toBe("gpt-5.4");
expect((result as unknown as Record<string, unknown>).contextWindow).toBe(300_000);
});
it("falls through to synthetic catch-all when codex template is missing", () => {
const ctx = createMockCtx("gpt-5.4");
const result = requireResolvedModel(ctx);
@@ -158,11 +202,20 @@ describe("resolveCopilotForwardCompatModel", () => {
}
});
it("infers reasoning=true for Codex model IDs", () => {
for (const id of ["gpt-5.4-codex", "gpt-5.5-codex", "gpt-5.4-codex-mini", "gpt-5.3-codex"]) {
const ctx = createMockCtx(id);
const result = requireResolvedModel(ctx);
expect((result as unknown as Record<string, unknown>).reasoning).toBe(true);
}
});
it("sets reasoning=false for non-reasoning model IDs including mid-string o1/o3", () => {
for (const id of [
"gpt-5.4-mini",
"claude-sonnet-4.6",
"gpt-4o",
"mycodexmodel",
"audio-o1-hd",
"turbo-o3-voice",
]) {

View File

@@ -6,12 +6,18 @@ import { normalizeModelCompat } from "openclaw/plugin-sdk/provider-model-shared"
import { normalizeOptionalLowercaseString } from "openclaw/plugin-sdk/text-runtime";
export const PROVIDER_ID = "github-copilot";
const CODEX_GPT_54_MODEL_ID = "gpt-5.4";
const CODEX_TEMPLATE_MODEL_IDS = ["gpt-5.2-codex"] as const;
const CODEX_FORWARD_COMPAT_TARGET_IDS = new Set(["gpt-5.4", "gpt-5.3-codex"]);
// gpt-5.3-codex is only a useful template when gpt-5.4 is the target; it is
// always a registry miss (and therefore skipped) when it is the target itself.
const CODEX_TEMPLATE_MODEL_IDS = ["gpt-5.3-codex", "gpt-5.2-codex"] as const;
const DEFAULT_CONTEXT_WINDOW = 128_000;
const DEFAULT_MAX_TOKENS = 8192;
function isCopilotCodexModelId(modelId: string): boolean {
return /(?:^|[-_.])codex(?:$|[-_.])/.test(modelId);
}
export function resolveCopilotTransportApi(
modelId: string,
): "anthropic-messages" | "openai-responses" {
@@ -35,9 +41,9 @@ export function resolveCopilotForwardCompatModel(
return undefined;
}
// For gpt-5.4 specifically, clone from the gpt-5.2-codex template
// to preserve any special settings the registry has for codex models.
if (lowerModelId === CODEX_GPT_54_MODEL_ID) {
// For gpt-5.4 and gpt-5.3-codex, clone from a registered codex template
// to inherit the correct reasoning and capability flags.
if (CODEX_FORWARD_COMPAT_TARGET_IDS.has(lowerModelId)) {
for (const templateId of CODEX_TEMPLATE_MODEL_IDS) {
const template = ctx.modelRegistry.find(
PROVIDER_ID,
@@ -60,7 +66,7 @@ export function resolveCopilotForwardCompatModel(
// model isn't available on the user's plan. This lets new models be used
// by simply adding them to agents.defaults.models in openclaw.json — no
// code change required.
const reasoning = /^o[13](\b|$)/.test(lowerModelId);
const reasoning = /^o[13](\b|$)/.test(lowerModelId) || isCopilotCodexModelId(lowerModelId);
return normalizeModelCompat({
id: trimmedModelId,
name: trimmedModelId,

View File

@@ -101,11 +101,17 @@ describe("wrapCopilotAnthropicStream", () => {
expect(baseStreamFn).toHaveBeenCalledWith(expect.anything(), expect.anything(), options);
});
it("adds Copilot headers and rewrites Responses connection-bound IDs before payload send", () => {
const connectionBoundId = Buffer.from(`reasoning-${"x".repeat(24)}`).toString("base64");
it("adds Copilot headers, preserves reasoning IDs, and rewrites message IDs before payload send", () => {
const reasoningId = Buffer.from(`reasoning-${"x".repeat(24)}`).toString("base64");
const messageId = Buffer.from(`message-${"y".repeat(24)}`).toString("base64");
const payloads: Array<{ input: Array<Record<string, unknown>> }> = [];
const baseStreamFn = vi.fn((_model, _context, options) => {
const payload = { input: [{ id: connectionBoundId, type: "reasoning" }] };
const payload = {
input: [
{ id: reasoningId, type: "reasoning" },
{ id: messageId, type: "message" },
],
};
options?.onPayload?.(payload, _model);
payloads.push(payload);
return {
@@ -144,7 +150,8 @@ describe("wrapCopilotAnthropicStream", () => {
"X-Test": "1",
},
});
expect(payloads[0]?.input[0]?.id).toMatch(/^rs_[a-f0-9]{16}$/);
expect(payloads[0]?.input[0]?.id).toBe(reasoningId);
expect(payloads[0]?.input[1]?.id).toMatch(/^msg_[a-f0-9]{16}$/);
});
it("rewrites Copilot Responses IDs returned by an existing payload hook", async () => {

View File

@@ -170,14 +170,14 @@ async function expectBuiltArtifactNodeRequireFastPath(
fs.mkdirSync(pluginRoot, { recursive: true });
const importerPath = path.join(pluginRoot, "index.js");
const sidecarPath = path.join(pluginRoot, "fast-path-sidecar.js");
const sidecarPath = path.join(pluginRoot, "fast-path-sidecar.cjs");
fs.writeFileSync(importerPath, "export default {};\n", "utf8");
// CommonJS so `nodeRequire` succeeds without falling back to jiti.
fs.writeFileSync(sidecarPath, "module.exports = { sentinel: 7 };\n", "utf8");
expect(
channelEntryContract.loadBundledEntryExportSync<number>(pathToFileURL(importerPath).href, {
specifier: "./fast-path-sidecar.js",
specifier: "./fast-path-sidecar.cjs",
exportName: "sentinel",
}),
).toBe(7);