ima2-gen 3.20.0 → 3.21.0

package/lib/providers/adapters/openaiOperations.js

@@ -7,7 +7,9 @@ import { retryPromptOnlyJsonImage } from "../../responsesFallback.js";
 import { AUTO_PROMPT_FIDELITY_SUFFIX, DIRECT_PROMPT_FIDELITY_SUFFIX, EDIT_DEVELOPER_PROMPT, EDIT_NO_SEARCH_DEVELOPER_PROMPT, GENERATE_DEVELOPER_PROMPT, GENERATE_NO_SEARCH_DEVELOPER_PROMPT, MULTIMODE_DEVELOPER_PROMPT, MULTIMODE_NO_SEARCH_DEVELOPER_PROMPT, buildEditTextPrompt, buildMultimodeSequencePrompt, buildUserTextPrompt, } from "../../oauthProxy.js";
 import { normalizeOAuthParams, normalizeImageToolModel } from "../../oauthNormalize.js";
 import { postResponses } from "../../responsesTransport.js";
-function normalizeRef(ref) {
+import { runOAuthImageJob } from "../../oauthImages.js";
+import { API_FALLBACK_IMAGE_MODEL, FALLBACK_IMAGE_MODEL, migrateOAuthImageModel } from "../../imageModels.js";
+function refImage(ref) {
     const b64 = typeof ref === "string" ? ref : ref?.b64;
     const detectedMime = typeof ref === "object" && ref?.detectedMime
         ? ref.detectedMime
@@ -18,11 +20,26 @@ function normalizeRef(ref) {
         : ["image/png", "image/jpeg", "image/webp"].includes(declaredMime)
             ? declaredMime
             : "image/png";
+    return { b64: b64, mime: mime };
+}
+function normalizeRef(ref) {
+    const { b64, mime } = refImage(ref);
     return { type: "input_image", image_url: `data:${mime};base64,${b64}` };
 }
+/** GPT OAuth calls the Codex backend in process: a GPT-6 planner plus gpt-image-2 (lib/oauthImages.ts). */
+function isOAuthLane(provider) {
+    return provider !== "api";
+}
+/** The request model: a legacy OAuth id moves to its GPT-6 tier; the API lane keeps its own default. */
+function requestModel(provider, ctx, requested) {
+    if (!isOAuthLane(provider)) {
+        return requested || ctx.config?.apiProvider?.defaultImageModel || API_FALLBACK_IMAGE_MODEL;
+    }
+    return migrateOAuthImageModel(requested || ctx.config?.imageModels?.default || FALLBACK_IMAGE_MODEL);
+}
 export async function generateViaResponses(provider, prompt, quality, size, moderation = "low", references = [], requestId = null, mode = "auto", ctxRaw = {}, options = {}) {
     const ctx = requireRuntimeContext(ctxRaw);
-    const model = options.model || ctx.config?.imageModels?.default || "gpt-5.6-luna";
+    const model = requestModel(provider, ctx, options.model);
     const webSearchEnabled = options.webSearchEnabled !== false && options.searchMode !== "off";
     const toolModel = normalizeImageToolModel(provider ?? "oauth", options.imageToolModel);
     if (toolModel.error)
@@ -41,6 +58,40 @@ export async function generateViaResponses(provider, prompt, quality, size, mode
     const toolChoice = imageToolChoice(options.forceImageToolChoice ?? ctx.config?.oauth?.forceImageToolChoice !== false);
     const toolChoiceKind = imageToolChoiceKind(toolChoice);
     const referenceInputs = references.map(normalizeRef);
+    if (isOAuthLane(provider)) {
+        const result = await runOAuthImageJob({
+            ctx,
+            scope: "oauth",
+            requestId,
+            signal: options.signal,
+            model,
+            mode,
+            reasoningEffort: options.reasoningEffort || "low",
+            webSearchEnabled,
+            developerPrompt: webSearchEnabled ? GENERATE_DEVELOPER_PROMPT : GENERATE_NO_SEARCH_DEVELOPER_PROMPT,
+            userText: buildUserTextPrompt(prompt, mode, { webSearchEnabled, size }),
+            directPrompt: prompt ?? "",
+            images: references.map(refImage),
+            maxImages: 1,
+            quality,
+            size,
+            moderation,
+            background: options.background,
+            onFinalImage: options.onFinalImage,
+        });
+        const image = result.images[0];
+        if (!image?.b64) {
+            throw emptyResponseError("No image data received from the OAuth image lane", result, {
+                provider, model, quality, size, moderation, webSearchEnabled,
+                refsCount: referenceInputs.length,
+                inputImageCount: referenceInputs.length,
+                promptChars: typeof prompt === "string" ? prompt.length : 0,
+                toolTypes: ["function"],
+                toolChoiceKind: "auto",
+            });
+        }
+        return { b64: image.b64, usage: result.usage, webSearchCalls: result.webSearchCalls, revisedPrompt: image.revisedPrompt, text: result.text };
+    }
     const userContent = referenceInputs.length
         ? [...referenceInputs, { type: "input_text", text: buildUserTextPrompt(prompt, mode, { webSearchEnabled, size }) }]
         : buildUserTextPrompt(prompt, mode, { webSearchEnabled, size });
@@ -110,7 +161,7 @@ export async function generateMultimodeViaResponses(provider, prompt, quality, s
     const ctx = requireRuntimeContext(ctxRaw);
     const maxGeneratedImages = Math.max(1, Math.trunc(Number(ctx.config.limits.maxGeneratedImages) || 24));
     const maxImages = Math.min(maxGeneratedImages, Math.max(1, Math.trunc(Number(options.maxImages) || 1)));
-    const model = options.model || ctx.config?.imageModels?.default || "gpt-5.6-luna";
+    const model = requestModel(provider, ctx, options.model);
     const webSearchEnabled = options.webSearchEnabled !== false && options.searchMode !== "off";
     const toolModel = normalizeImageToolModel(provider ?? "oauth", options.imageToolModel);
     if (toolModel.error)
@@ -126,6 +177,27 @@ export async function generateMultimodeViaResponses(provider, prompt, quality, s
     const userContent = referenceInputs.length
         ? [...referenceInputs, { type: "input_text", text: userText }]
         : userText;
+    if (isOAuthLane(provider)) {
+        return await runOAuthImageJob({
+            ctx,
+            scope: "oauth-multimode",
+            requestId,
+            signal: options.signal,
+            model,
+            mode,
+            reasoningEffort: options.reasoningEffort || "low",
+            webSearchEnabled,
+            developerPrompt: webSearchEnabled ? MULTIMODE_DEVELOPER_PROMPT : MULTIMODE_NO_SEARCH_DEVELOPER_PROMPT,
+            userText,
+            directPrompt: prompt ?? "",
+            images: references.map(refImage),
+            maxImages,
+            quality,
+            size,
+            moderation,
+            onFinalImage: options.onFinalImage,
+        });
+    }
     return await postResponses({
         ctx,
         provider,
@@ -150,7 +222,7 @@ export async function generateMultimodeViaResponses(provider, prompt, quality, s
 }
 export async function editViaResponses(provider, prompt, imageB64, quality, size, moderation = "low", mode = "auto", ctxRaw = {}, requestId = null, options = {}) {
     const ctx = requireRuntimeContext(ctxRaw);
-    const model = options.model || ctx.config?.imageModels?.default || "gpt-5.6-luna";
+    const model = requestModel(provider, ctx, options.model);
     const webSearchEnabled = options.webSearchEnabled !== false && options.searchMode !== "off";
     const toolModel = normalizeImageToolModel(provider ?? "oauth", options.imageToolModel);
     if (toolModel.error)
@@ -181,6 +253,46 @@ export async function editViaResponses(provider, prompt, imageB64, quality, size
         ...maskContent,
         { type: "input_text", text: buildEditTextPrompt(prompt, mode, { webSearchEnabled, size }) },
     ];
+    if (isOAuthLane(provider)) {
+        const hasMask = maskContent.length > 0;
+        const result = await runOAuthImageJob({
+            ctx,
+            scope: "oauth-edit",
+            requestId,
+            signal: options.signal,
+            model,
+            mode,
+            reasoningEffort: options.reasoningEffort || "low",
+            webSearchEnabled,
+            developerPrompt: webSearchEnabled ? EDIT_DEVELOPER_PROMPT : EDIT_NO_SEARCH_DEVELOPER_PROMPT,
+            userText: buildEditTextPrompt(prompt, mode, { webSearchEnabled, size })
+                + (hasMask ? "\n\nThe last attached image is an edit mask guide marking where the edit should apply; it is not a visible element of the result." : ""),
+            directPrompt: prompt ?? "",
+            // Edit source first, then references, then the mask guide; the edits endpoint takes five.
+            images: [
+                { b64: imageForRequest.b64, mime: "image/jpeg" },
+                ...referenceImages.slice(0, hasMask ? 3 : 4).map(({ b64 }) => ({ b64, mime: "image/jpeg" })),
+                ...(hasMask ? [{ b64: options.mask, mime: "image/png" }] : []),
+            ],
+            maxImages: 1,
+            quality,
+            size,
+            moderation,
+            background: options.background,
+        });
+        const image = result.images[0];
+        if (!image?.b64) {
+            throw emptyResponseError("No image data received from the OAuth image lane edit", result, {
+                provider, model, quality, size, moderation, webSearchEnabled,
+                refsCount: referenceImages.length,
+                inputImageCount: 1 + referenceImages.length + (hasMask ? 1 : 0),
+                promptChars: typeof prompt === "string" ? prompt.length : 0,
+                toolTypes: ["function"],
+                toolChoiceKind: "auto",
+            });
+        }
+        return { b64: image.b64, usage: result.usage, revisedPrompt: image.revisedPrompt, webSearchCalls: result.webSearchCalls };
+    }
     const result = await postResponses({
         ctx,
         provider,

package/lib/providers/registry.js

@@ -6,7 +6,9 @@ const EDIT = { generate: true, edit: true, mask: false, streaming: false };
 // lib/oauthProxy/multimodeGenerators.ts rejects masks unconditionally, but no
 // route reaches it, so it does not describe current behavior.
 const RESPONSES = { generate: true, edit: true, mask: true, streaming: true };
-const UNSUPPORTED = { generate: false, edit: false, mask: false, streaming: false };
+// GPT OAuth on the Codex backend (lib/oauthImages.ts): a GPT-6 model plans, the Images API
+// (gpt-image-2) renders. Edits and mask guidance work; the Images API has no partial frames.
+const OAUTH_IMAGES = { generate: true, edit: true, mask: true, streaming: false };
 const GENERATE_ONLY = { generate: true, edit: false, mask: false, streaming: false };
 export const REGISTRY = [
     {
@@ -20,14 +22,9 @@ export const REGISTRY = [
                 configKey: "oauth",
             }],
         models: [
-            { id: "gpt-5.5", kind: "image", supports: RESPONSES },
-            { id: "gpt-5.4", kind: "image", supports: RESPONSES },
-            { id: "gpt-5.4-mini", kind: "image", supports: RESPONSES },
-            { id: "gpt-5.6-sol", aliases: ["sol"], kind: "image", supports: RESPONSES },
-            { id: "gpt-5.6-terra", aliases: ["terra"], kind: "image", supports: RESPONSES },
-            { id: "gpt-5.6-luna", aliases: ["luna"], kind: "image", supports: RESPONSES },
-            { id: "gpt-6-astra", aliases: ["astra"], kind: "image", supports: RESPONSES },
-            { id: "gpt-5.3-codex-spark", aliases: ["spark"], kind: "image", supports: UNSUPPORTED },
+            { id: "gpt-6-luna", aliases: ["luna"], kind: "image", supports: OAUTH_IMAGES },
+            { id: "gpt-6-sol", aliases: ["sol"], kind: "image", supports: OAUTH_IMAGES },
+            { id: "gpt-6-astra", aliases: ["astra"], kind: "image", supports: OAUTH_IMAGES },
         ],
         referenceLimits: {},
         elementTaxonomy: "gpt",
@@ -50,10 +47,10 @@ export const REGISTRY = [
             { id: "gpt-5.5", kind: "image", supports: RESPONSES },
             { id: "gpt-5.4", kind: "image", supports: RESPONSES },
             { id: "gpt-5.4-mini", kind: "image", supports: RESPONSES },
-            { id: "gpt-5.6-sol", aliases: ["sol"], kind: "image", supports: RESPONSES },
-            { id: "gpt-5.6-terra", aliases: ["terra"], kind: "image", supports: RESPONSES },
-            { id: "gpt-5.6-luna", aliases: ["luna"], kind: "image", supports: RESPONSES },
-            { id: "gpt-6-astra", aliases: ["astra"], kind: "image", supports: RESPONSES },
+            { id: "gpt-5.6-sol", kind: "image", supports: RESPONSES },
+            { id: "gpt-5.6-terra", kind: "image", supports: RESPONSES },
+            { id: "gpt-5.6-luna", kind: "image", supports: RESPONSES },
+            { id: "gpt-6-astra", kind: "image", supports: RESPONSES },
         ],
         referenceLimits: {},
         elementTaxonomy: "gpt",

package/lib/responsesDoctor.js

@@ -1,7 +1,9 @@
 import { config as defaultConfig } from "../config.js";
 import { errInfo } from "./errInfo.js";
 import { parseJson, parseStream, safeDiagnosticLabel } from "./responsesParse.js";
-import { GENERATE_DEVELOPER_PROMPT, GENERATE_NO_SEARCH_DEVELOPER_PROMPT, buildUserTextPrompt, waitForOAuthReady, } from "./oauthProxy.js";
+import { runOAuthImageJob } from "./oauthImages.js";
+import { migrateOAuthImageModel } from "./oauthLegacyModels.js";
+import { GENERATE_DEVELOPER_PROMPT, GENERATE_NO_SEARCH_DEVELOPER_PROMPT, buildUserTextPrompt, } from "./oauthProxy.js";
 const BUILTIN_PROMPT_ID = "builtin_cat";
 const BUILTIN_CAT_PROMPT = "고양이";
 const TEXT_OK_PROMPT = "Reply with exactly OK.";
@@ -138,17 +140,9 @@ async function endpointFor(provider, options) {
             headers,
         };
     }
-    await waitForOAuthReady(ctx);
-    const port = ctx.config?.oauth?.proxyPort || defaultConfig.oauth.proxyPort;
-    const baseUrl = safeOAuthBaseUrl(options.oauthUrl || `http://127.0.0.1:${port}`);
-    const headers = {
-        "Content-Type": "application/json",
-        Accept: "text/event-stream, application/json",
-    };
-    return {
-        url: `${baseUrl}/v1/responses`,
-        headers,
-    };
+    // GPT OAuth probes go through runOAuthLaneProbe; only the API key lane reaches here.
+    void ctx;
+    throw Object.assign(new Error(`hosted-tool probes are API-only (provider ${provider})`), { code: "PROBE_PROVIDER_UNSUPPORTED" });
 }
 function apiAuthorizationHeader(apiKey) {
     const key = typeof apiKey === "string" ? apiKey.trim() : "";
@@ -164,17 +158,6 @@ function apiAuthorizationHeader(apiKey) {
     }
     return `Bearer ${key}`;
 }
-function safeOAuthBaseUrl(value) {
-    try {
-        const parsed = new URL(value);
-        parsed.username = "";
-        parsed.password = "";
-        return parsed.toString().replace(/\/$/, "");
-    }
-    catch {
-        return value.replace(/\/$/, "");
-    }
-}
 function sanitizeProbeErrorMessage(value) {
     return value
         .replace(/([a-z][a-z0-9+.-]*:\/\/)([^/\s@]+)@/gi, "$1[redacted]@")
@@ -348,9 +331,56 @@ async function runSingleProbe(provider, options, spec) {
         return { ...partial, diagnosticReason: reasonFrom(partial) };
     }
 }
+/**
+ * GPT OAuth has no hosted image tool on GPT-6: the lane plans with the selected model and
+ * renders through the Images API (lib/oauthImages.ts). Probe that lane end to end instead of
+ * the hosted-tool variants, which would all fail by construction.
+ */
+async function runOAuthLaneProbe(options, id, mode, settings) {
+    const start = Date.now();
+    const ctx = {
+        ...(options.ctx ?? { config: defaultConfig }),
+        ...(options.oauthUrl ? { oauthUrl: options.oauthUrl, oauthTransport: "proxy" } : { oauthTransport: "native" }),
+    };
+    const request = { stream: true, toolTypes: ["function"], toolChoiceKind: "none", promptId: options.prompt ? "custom" : BUILTIN_PROMPT_ID, promptChars: settings.prompt.length };
+    try {
+        const result = await runOAuthImageJob({
+            ctx, scope: "doctor-oauth", model: settings.model, mode, reasoningEffort: "low", webSearchEnabled: false,
+            developerPrompt: GENERATE_NO_SEARCH_DEVELOPER_PROMPT,
+            userText: buildUserTextPrompt(settings.prompt, mode, { webSearchEnabled: false, size: settings.size }),
+            directPrompt: settings.prompt, images: [], maxImages: 1, quality: settings.quality, size: settings.size,
+        });
+        const image = result.images[0];
+        const partial = {
+            id, ok: Boolean(image?.b64), expectation: "image", request,
+            response: {
+                httpStatus: 200, contentType: "application/json", upstreamRequestId: null, durationMs: Date.now() - start,
+                eventCount: result.eventCount, eventTypes: result.eventTypes, webSearchCalls: result.webSearchCalls,
+                textOutputChars: result.text?.length ?? 0, imageResultCount: result.images.length,
+                firstImageChars: image?.b64.length ?? 0, diagnostics: result.diagnostics,
+            },
+            error: null,
+        };
+        return { ...partial, diagnosticReason: reasonFrom(partial) };
+    }
+    catch (error) {
+        const err = error;
+        const partial = {
+            id, ok: false, expectation: "image", request,
+            response: {
+                httpStatus: typeof err.status === "number" ? err.status : null, contentType: null, upstreamRequestId: null,
+                durationMs: Date.now() - start, eventCount: 0, eventTypes: {}, webSearchCalls: 0, textOutputChars: 0,
+                imageResultCount: 0, firstImageChars: 0, diagnostics: null,
+            },
+            error: { code: err.code ?? null, type: null, param: null, message: sanitizeProbeErrorMessage(err.message ?? "OAuth lane probe failed") },
+        };
+        return { ...partial, diagnosticReason: reasonFrom(partial) };
+    }
+}
 export async function runImageDoctorProbe(options = {}) {
     const provider = options.provider || "oauth";
-    const model = options.model || defaultConfig.imageModels?.default || "gpt-5.6-luna";
+    const requested = options.model || defaultConfig.imageModels?.default || "gpt-6-luna";
+    const model = provider === "oauth" ? migrateOAuthImageModel(requested) : requested;
     const size = options.size || "1024x1024";
     const quality = options.quality || "low";
     const moderation = options.moderation || "low";
@@ -364,8 +394,16 @@ export async function runImageDoctorProbe(options = {}) {
         matrix: options.matrix === true,
     });
     const probes = [];
-    for (const spec of specs)
-        probes.push(await runSingleProbe(provider, options, spec));
+    if (provider === "api") {
+        for (const spec of specs)
+            probes.push(await runSingleProbe(provider, options, spec));
+    }
+    else {
+        const settings = { model, size, quality, prompt };
+        probes.push(await runOAuthLaneProbe(options, "oauth_plan_render", "auto", settings));
+        if (options.matrix === true)
+            probes.push(await runOAuthLaneProbe(options, "oauth_direct_render", "direct", settings));
+    }
     return {
         provider,
         endpointKind: provider === "api" ? "api" : "oauth",

package/lib/responsesFallback.js

@@ -1,6 +1,7 @@
 import { logEvent } from "./logger.js";
 import { imageToolChoice, tools } from "./responsesTools.js";
 import { emptyResponseError } from "./responsesErrors.js";
+import { thrownFields } from "./errInfo.js";
 import { GENERATE_DEVELOPER_PROMPT, GENERATE_NO_SEARCH_DEVELOPER_PROMPT, buildUserTextPrompt, } from "./oauthProxy.js";
 const MAX_RETRIES = 2;
 export async function retryPromptOnlyJsonImage({ postResponses, ctx, provider, prompt, mode, model, quality, size, moderation, requestId, signal, initial, referenceInputs = [], webSearchDroppedOnRetry, reasoningEffort, background, outputFormat, }) {
@@ -82,7 +83,7 @@ export async function retryPromptOnlyJsonImage({ postResponses, ctx, provider, p
                     Object.assign(e, retryMeta);
                 throw e;
             }
-            logEvent("oauth", "retry_error", { requestId, attempt, error: e.message, status: e.status, code: e.code });
+            logEvent("oauth", "retry_error", { requestId, attempt, error: e.message, status: thrownFields(e).status, code: thrownFields(e).code });
             continue;
         }
         const image = lastRetry.images[0];

package/lib/responsesParse.js

@@ -5,6 +5,7 @@ export { safeDiagnosticLabel };
 function createState() {
     return {
         images: [],
+        functionCalls: [],
         eventTypes: {},
         outputItemSummary: [],
         usage: null,
@@ -134,6 +135,17 @@ function recordOutputItem(state, eventType, item) {
         state.webSearchCallSeen = true;
     if (item.type === "message")
         state.messageOutputSeen = true;
+    if (item.type === "function_call" && typeof item.name === "string") {
+        const callId = typeof item.call_id === "string" ? item.call_id : null;
+        const duplicate = callId !== null && state.functionCalls.some((call) => call.callId === callId);
+        if (!duplicate) {
+            state.functionCalls.push({
+                name: item.name,
+                arguments: typeof item.arguments === "string" ? item.arguments : "",
+                callId,
+            });
+        }
+    }
 }
 function diagnosticsFromState(state) {
     const outputTextChars = (state.finalTextOutput ?? cleanTextOutput(state.textOutput) ?? "").length;
@@ -171,6 +183,7 @@ function resultFromState(state) {
         extraIgnored: state.extraIgnored,
         text,
         diagnostics: diagnosticsFromState(state),
+        ...(state.functionCalls.length ? { functionCalls: state.functionCalls } : {}),
     };
 }
 function makeStreamError(message, code, eventCount, eventType) {

package/lib/responsesTransport.js

@@ -4,6 +4,7 @@ import { errInfo } from "./errInfo.js";
 import { setJobPhase } from "./inflight.js";
 import { parseJson, parseStream, safeDiagnosticLabel, } from "./responsesParse.js";
 import { waitForOAuthReady } from "./oauthProxy.js";
+import { oauthFetch } from "./codexBackend/index.js";
 const RESPONSES_ERROR_MARKER = "ima2ResponsesError";
 function makeError(message, { status = 500, code = "RESPONSES_IMAGE_ERROR", cause, ...rest } = {}) {
     const err = new Error(message);
@@ -31,6 +32,9 @@ function parseOpenAIErrorBody(text) {
     }
 }
 function normalizedCode(upstream) {
+    // The in-process GPT OAuth client reports a missing or unrefreshable ChatGPT session this way.
+    if (upstream?.code === "OAUTH_SESSION_REQUIRED")
+        return "AUTH_CHATGPT_EXPIRED";
     const byCode = classifyUpstreamErrorCode(upstream?.code);
     if (byCode !== "UNKNOWN")
         return byCode;
@@ -42,6 +46,8 @@ function normalizedCode(upstream) {
 }
 function safeUpstreamClientMessage(upstream, status) {
     const code = normalizedCode(upstream);
+    if (code === "AUTH_CHATGPT_EXPIRED")
+        return "ChatGPT sign-in is missing or expired. Run `ima2 gpt login`.";
     if (code === "AUTH_API_KEY_INVALID")
         return "API key is invalid or unavailable.";
     if (code === "MODERATION_REFUSED")
@@ -57,17 +63,6 @@ function safeUpstreamClientMessage(upstream, status) {
         return "OpenAI rate limited the image request.";
     return "OpenAI rejected the image request.";
 }
-function safeBaseUrl(value) {
-    try {
-        const parsed = new URL(value);
-        parsed.username = "";
-        parsed.password = "";
-        return parsed.toString().replace(/\/$/, "");
-    }
-    catch {
-        return value.replace(/\/$/, "");
-    }
-}
 function apiAuthorizationHeader(apiKey) {
     const key = typeof apiKey === "string" ? apiKey.trim() : "";
     if (!key) {
@@ -101,12 +96,15 @@ async function getEndpoint(ctx, provider, _scope) {
         };
     }
     await waitForOAuthReady(ctx);
-    const port = ctx?.config?.oauth?.proxyPort || 10531;
+    // GPT OAuth: path only; oauthFetch picks the in-process Codex client or the configured proxy.
     return {
-        url: `${safeBaseUrl(ctx?.oauthUrl || `http://127.0.0.1:${port}`)}/v1/responses`,
+        url: null,
         headers: { "Content-Type": "application/json", Accept: "text/event-stream" },
     };
 }
+function requestResponses(ctx, url, init) {
+    return url ? fetch(url, init) : oauthFetch(ctx, "/v1/responses", init);
+}
 function combineAbortSignals(signals) {
     const only = signals[0];
     if (signals.length === 1 && only)
@@ -130,7 +128,7 @@ export async function postResponses({ ctx, provider, scope, payload, requestId,
         ? combineAbortSignals([controller.signal, signal])
         : controller.signal;
     try {
-        const res = await fetch(url, {
+        const res = await requestResponses(ctx, url, {
             method: "POST",
             headers: headers,
             signal: fetchSignal,
@@ -188,3 +186,76 @@ export async function postResponses({ ctx, provider, scope, payload, requestId,
         clearTimeout(timer);
     }
 }
+/**
+ * Render through the GPT OAuth Images API (gpt-image-2 on the ChatGPT backend, called in
+ * process by lib/codexBackend). Shares postResponses' endpoint readiness, timeout, cancellation and upstream
+ * error classification so OAuth failures surface with the same codes as before.
+ */
+export async function postOAuthImages({ ctx, scope, kind, json, form, requestId, signal = null, }) {
+    await waitForOAuthReady(ctx);
+    const timeoutMs = ctx?.config?.oauth?.generationTimeoutMs || 400 * 1000;
+    const controller = new AbortController();
+    const timer = setTimeout(() => controller.abort(), timeoutMs);
+    const fetchSignal = signal ? combineAbortSignals([controller.signal, signal]) : controller.signal;
+    try {
+        const res = await oauthFetch(ctx, `/v1/images/${kind}`, {
+            method: "POST",
+            ...(json ? { headers: { "Content-Type": "application/json" }, body: JSON.stringify(json) } : { body: form ?? null }),
+            signal: fetchSignal,
+        });
+        logEvent(scope, "images_response", { requestId, kind, status: res.status });
+        const text = await res.text();
+        if (!res.ok) {
+            const upstream = parseOpenAIErrorBody(text);
+            if (res.status >= 400 && res.status < 500 && upstream?.message) {
+                throw makeError(safeUpstreamClientMessage(upstream, res.status), {
+                    status: res.status,
+                    code: normalizedCode(upstream),
+                    upstreamBodyChars: text.length,
+                    upstreamCode: upstream.code,
+                    upstreamType: upstream.type,
+                    upstreamParam: upstream.param,
+                    upstreamMessageRedacted: true,
+                });
+            }
+            throw makeError(`GPT OAuth returned ${res.status}`, { status: res.status, upstreamBodyChars: text.length });
+        }
+        let parsed;
+        try {
+            parsed = JSON.parse(text);
+        }
+        catch {
+            throw makeError("GPT OAuth returned a non-JSON image response", { status: 502, upstreamBodyChars: text.length });
+        }
+        const images = (Array.isArray(parsed.data) ? parsed.data : [])
+            .map((item) => (typeof item?.b64_json === "string" && item.b64_json ? { b64: item.b64_json } : null))
+            .filter((item) => item !== null);
+        logEvent(scope, "images_end", { requestId, kind, imageCount: images.length });
+        return {
+            images,
+            usage: parsed.usage && typeof parsed.usage === "object" ? parsed.usage : null,
+            background: typeof parsed.background === "string" ? parsed.background : null,
+            outputFormat: typeof parsed.output_format === "string" ? parsed.output_format : null,
+        };
+    }
+    catch (e) {
+        const err = errInfo(e);
+        if (err.name === "AbortError") {
+            if (signal?.aborted) {
+                throw makeError("Generation canceled", { status: 499, code: "GENERATION_CANCELED", cause: err.raw });
+            }
+            throw makeError("OAuth image generation timed out", { status: 504, code: "RESPONSES_IMAGE_TIMEOUT", cause: err.raw });
+        }
+        if (isKnownResponsesError(err.raw))
+            throw err.raw;
+        throw makeError("OAuth image request failed before receiving a response", {
+            status: 502,
+            code: "NETWORK_FAILED",
+            errorName: err.name,
+            upstreamMessageRedacted: true,
+        });
+    }
+    finally {
+        clearTimeout(timer);
+    }
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ima2-gen",
-  "version": "3.20.0",
+  "version": "3.21.0",
   "packageManager": "npm@11.18.0",
   "description": "Local-first visual generation runtime and studio for people and coding agents, with reproducible image and video workflows across multiple providers.",
   "type": "module",
@@ -29,7 +29,8 @@
     "setup": "node bin/ima2.js setup",
     "prepack": "npm run ui:build && npm run build:server && npm run build:cli",
     "prepublishOnly": "node scripts/release-contract.mjs assert-publish-context && npm run verify:release",
-    "lint:pkg": "node -e \"const p=require('./package.json'); const req=['name','version','bin']; for(const k of req){if(!p[k])throw new Error('missing '+k)} const mustInclude=['bin/**/*.js','lib/**/*.js','routes/**/*.js','skills/','integrations/comfyui/ima2_gen_bridge/__init__.py','integrations/comfyui/ima2_gen_bridge/nodes.py','integrations/comfyui/ima2_gen_bridge/README.md','assets/card-news/templates/','assets/mcp-snapshots/','vendor/','server.js','LICENSE']; for(const f of mustInclude){if(!p.files.includes(f))throw new Error('files[] must include '+f)}\"",
+    "lint": "eslint",
+    "lint:pkg": "node -e \"const p=require('./package.json'); const req=['name','version','bin']; for(const k of req){if(!p[k])throw new Error('missing '+k)} const mustInclude=['bin/**/*.js','lib/**/*.js','routes/**/*.js','skills/','integrations/comfyui/ima2_gen_bridge/__init__.py','integrations/comfyui/ima2_gen_bridge/nodes.py','integrations/comfyui/ima2_gen_bridge/README.md','assets/card-news/templates/','assets/mcp-snapshots/','server.js','LICENSE']; for(const f of mustInclude){if(!p.files.includes(f))throw new Error('files[] must include '+f)}\"",
     "docs:refresh-line-counts": "node scripts/refresh-structure-line-counts.mjs",
     "docs:runtime": "node --import tsx scripts/generate-runtime-install-contract.mjs",
     "docs:runtime:check": "node --import tsx scripts/generate-runtime-install-contract.mjs --check",
@@ -61,10 +62,10 @@
     "gemini"
   ],
   "license": "MIT",
-  "homepage": "https://lidge-jun.github.io/ima2-gen/",
+  "homepage": "https://lidge-ai.github.io/ima2-gen/",
   "repository": {
     "type": "git",
-    "url": "git+https://github.com/lidge-jun/ima2-gen.git"
+    "url": "git+https://github.com/lidge-ai/ima2-gen.git"
   },
   "files": [
     "bin/**/*.js",
@@ -76,7 +77,6 @@
     "integrations/comfyui/ima2_gen_bridge/README.md",
     "ui/dist/",
     "docs/",
-    "vendor/",
     "assets/card-news/templates/",
     "assets/mcp-snapshots/",
     ".env.example",
@@ -99,7 +99,6 @@
     "express": "^5.1.0",
     "google-auth-library": "^11.0.1",
     "openai": "^7.4.0",
-    "openai-oauth": "file:vendor/openai-oauth-1.0.2-ima2.2.tgz",
     "sharp": "^0.35.3",
     "trash": "^10.1.1",
     "ulid": "^3.0.2",
@@ -109,7 +108,6 @@
     "fast-uri": "3.1.7"
   },
   "bundleDependencies": [
-    "openai-oauth",
     "zod"
   ],
   "allowScripts": {
@@ -118,13 +116,18 @@
     "fsevents": true
   },
   "devDependencies": {
+    "@eslint/js": "10.0.1",
     "@types/better-sqlite3": "^9.6.0",
     "@types/express": "^5.0.6",
     "@types/node": "^22.20.1",
+    "eslint": "10.10.0",
+    "eslint-plugin-react-hooks": "7.1.1",
+    "globals": "17.12.0",
     "node-gyp": "13.0.2",
     "tsx": "^4.23.12",
     "typescript": "^5.9.3",
+    "typescript-eslint": "8.70.0",
     "yaml": "2.9.1"
   },
-  "gitHead": "e78bbbaa6955c872c578e1ac063caa8c721c44a5"
+  "gitHead": "b502a70fb703fd700ae1788c232cb6411db0c323"
 }

package/routes/auth.js

@@ -148,7 +148,7 @@ export function registerAuthRoutes(app, ctx) {
     });
     /** CLI logins write the session file directly; this lets them tell a running server. */
     app.post("/api/oauth/restart", (_req, res) => {
-        const result = ctx?.restartOAuthProxy?.() ?? { restarted: false, reason: "oauth proxy is not managed by this server" };
+        const result = ctx?.restartOAuthProxy?.() ?? { restarted: false, reason: "GPT OAuth is not managed by this server" };
         res.json(result);
     });
 }