feat(release): v2.0.8 — custom image model handler resolution

fix: pass resolved provider to image handler for custom models (#238)

CHANGELOG.md

@@ -7,6 +7,70 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ---
 
+## [2.0.8] — 2026-03-07
+
+> ### 🐛 Bug Fix — Custom Image Model Handler Resolution
+
+### 🐛 Bug Fixes
+
+- **#238 — Custom image models still fail in handler layer** — v2.0.7 fixed the route-layer validation, but the handler (`handleImageGeneration()`) called `parseImageModel()` again internally, rejecting custom models a second time. Fix: handler now accepts an optional `resolvedProvider` parameter; when provided, it skips re-validation and routes custom models to the OpenAI-compatible handler with a synthetic config. PR #239
+
+### 📁 Files Changed
+
+| File                                         | Change                                                                           |
+| -------------------------------------------- | -------------------------------------------------------------------------------- |
+| `open-sse/handlers/imageGeneration.ts`       | Added `resolvedProvider` param + custom model fallback                           |
+| `src/app/api/v1/images/generations/route.ts` | Tracks `isCustomModel`, passes `resolvedProvider`, credentials for custom models |
+
+---
+
+## [2.0.7] — 2026-03-07
+
+> ### 🐛 Bug Fixes — Custom Image Models + Codex OAuth Workspace Isolation
+
+### 🐛 Bug Fixes
+
+- **#232 — Custom Gemini image models fail on `/v1/images/generations`** — Custom models tagged with `supportedEndpoints: ["images"]` appeared in the model listing (GET) but were rejected by the POST handler. `parseImageModel()` only checked the built-in `IMAGE_PROVIDERS` registry. Fix: added a custom model DB fallback for models with the `images` endpoint tag. PR #237
+- **#236 — Codex OAuth overwrites existing connection when same email added to another workspace** — The OAuth callback route had 3 upsert blocks matching connections by email-only, bypassing the workspace-aware logic in `createProviderConnection()`. When the same email authenticated to a new workspace, the existing connection's `workspaceId` was silently overwritten. Fix: for Codex, the match now also checks `providerSpecificData.workspaceId`, allowing separate connections per workspace. PR #237
+
+### 📁 Files Changed
+
+| File                                             | Change                                               |
+| ------------------------------------------------ | ---------------------------------------------------- |
+| `src/app/api/v1/images/generations/route.ts`     | Custom model DB fallback in POST handler             |
+| `src/app/api/oauth/[provider]/[action]/route.ts` | Workspace-aware Codex matching in 3 upsert locations |
+
+### ⏭️ Issues Triaged
+
+- **#234** — Playground feature request — Acknowledged, added to roadmap
+- **#235** — ACP support feature request — Acknowledged, added to roadmap
+
+---
+
+## [2.0.6] — 2026-03-07
+
+> ### 🐛 Bug Fix — Custom Model API Format Routing
+
+### 🐛 Bug Fixes
+
+- **#204 — Custom model `apiFormat` not used in routing** — Custom models configured with `apiFormat: "responses"` in the dashboard were still being routed through the Chat Completions translator. The `apiFormat` field was stored in the DB and displayed in the UI, but never consumed by the routing layer. Fix: `getModelInfo()` now returns `apiFormat` from the custom model DB, and both `resolveModelOrError()` functions override `targetFormat` to `openai-responses` when set. PR #233
+
+### ✅ Issues Closed
+
+- **#205** — Combo endpoint support — Already implemented in v2.0.2
+- **#206** — Manual model→endpoint mapping — Already implemented in v2.0.2
+- **#223** — CLI fingerprint parity — Responded with 4-phase roadmap
+
+### 📁 Files Changed
+
+| File                              | Change                                                                 |
+| --------------------------------- | ---------------------------------------------------------------------- |
+| `src/sse/services/model.ts`       | Added `lookupCustomModelApiFormat()`, enriched `getModelInfo()` return |
+| `src/sse/handlers/chat.ts`        | Override `targetFormat` when `apiFormat === "responses"`               |
+| `src/sse/handlers/chatHelpers.ts` | Same override in duplicate `resolveModelOrError()`                     |
+
+---
+
 ## [2.0.5] — 2026-03-06
 
 > ### 🐛 Bug Fix, Electron Auto-Update & Dependency Bumps

open-sse/handlers/imageGeneration.ts

@@ -30,9 +30,23 @@ import {
  * @param {object} options.body - Request body
  * @param {object} options.credentials - Provider credentials { apiKey, accessToken }
  * @param {object} options.log - Logger
+ * @param {string} [options.resolvedProvider] - Pre-resolved provider ID (from route layer custom model resolution)
  */
-export async function handleImageGeneration({ body, credentials, log }) {
-  const { provider, model } = parseImageModel(body.model);
+export async function handleImageGeneration({ body, credentials, log, resolvedProvider = null }) {
+  let provider, model;
+
+  if (resolvedProvider) {
+    // Provider was already resolved by the route layer (custom model from DB)
+    // Extract model name from the full "provider/model" string
+    provider = resolvedProvider;
+    const modelStr = body.model || "";
+    model = modelStr.startsWith(provider + "/") ? modelStr.slice(provider.length + 1) : modelStr;
+  } else {
+    // Standard path: resolve from built-in image registry
+    const parsed = parseImageModel(body.model);
+    provider = parsed.provider;
+    model = parsed.model;
+  }
 
   if (!provider) {
     return {
@@ -43,12 +57,42 @@ export async function handleImageGeneration({ body, credentials, log }) {
   }
 
   const providerConfig = getImageProvider(provider);
+
+  // For custom models without a built-in provider config, use OpenAI-compatible handler
+  // with a synthetic config based on the provider's credentials
   if (!providerConfig) {
-    return {
-      success: false,
-      status: 400,
-      error: `Unknown image provider: ${provider}`,
+    if (!resolvedProvider) {
+      return {
+        success: false,
+        status: 400,
+        error: `Unknown image provider: ${provider}`,
+      };
+    }
+
+    // Custom model: use OpenAI-compatible format with provider's base URL
+    // The credentials were already resolved by the route layer
+    if (log) {
+      log.info("IMAGE", `Custom model ${provider}/${model} — using OpenAI-compatible handler`);
+    }
+
+    const syntheticConfig = {
+      id: provider,
+      baseUrl:
+        credentials?.baseUrl ||
+        `https://generativelanguage.googleapis.com/v1beta/openai/images/generations`,
+      authType: "apikey",
+      authHeader: "bearer",
+      format: "openai",
     };
+
+    return handleOpenAIImageGeneration({
+      model,
+      provider,
+      providerConfig: syntheticConfig,
+      body,
+      credentials,
+      log,
+    });
   }
 
   // Route to format-specific handler

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "omniroute",
-  "version": "2.0.4",
+  "version": "2.0.7",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omniroute",
-      "version": "2.0.4",
+      "version": "2.0.7",
       "hasInstallScript": true,
       "license": "MIT",
       "workspaces": [
@@ -6905,6 +6905,7 @@
       "version": "2.3.2",
       "resolved": "https://registry.npmjs.org/fsevents/-/fsevents-2.3.2.tgz",
       "integrity": "sha512-xiqMQR4xAeHTuB9uWm+fFRcIOgKBMiOBP+eXiyT7jsgVCq1bkVygt00oASowB7EdtpOHaaPgKt812P9ab+DDKA==",
+      "dev": true,
       "hasInstallScript": true,
       "license": "MIT",
       "optional": true,
@@ -9012,6 +9013,17 @@
         }
       }
     },
+    "node_modules/next-intl/node_modules/@swc/helpers": {
+      "version": "0.5.19",
+      "resolved": "https://registry.npmjs.org/@swc/helpers/-/helpers-0.5.19.tgz",
+      "integrity": "sha512-QamiFeIK3txNjgUTNppE6MiG3p7TdninpZu0E0PbqVh1a9FNLT2FRhisaa4NcaX52XVhA5l7Pk58Ft7Sqi/2sA==",
+      "license": "Apache-2.0",
+      "optional": true,
+      "peer": true,
+      "dependencies": {
+        "tslib": "^2.8.0"
+      }
+    },
     "node_modules/next/node_modules/postcss": {
       "version": "8.4.31",
       "resolved": "https://registry.npmjs.org/postcss/-/postcss-8.4.31.tgz",

package.json

@@ -1,6 +1,6 @@
 {
   "name": "omniroute",
-  "version": "2.0.5",
+  "version": "2.0.8",
   "description": "Smart AI Router with auto fallback — route to FREE & cheap models, zero downtime. Works with Cursor, Cline, Claude Desktop, Codex, and any OpenAI-compatible tool.",
   "type": "module",
   "bin": {

src/app/api/oauth/[provider]/[action]/route.ts

@@ -221,9 +221,15 @@ export async function POST(
       let connection: any;
       if (tokenData.email) {
         const existing = await getProviderConnections({ provider });
-        const match = existing.find(
-          (c: any) => c.email === tokenData.email && c.authType === "oauth"
-        );
+        const match = existing.find((c: any) => {
+          if (c.email !== tokenData.email || c.authType !== "oauth") return false;
+          // For Codex, also check workspaceId to avoid overwriting different workspace connections
+          if (provider === "codex" && tokenData.providerSpecificData?.workspaceId) {
+            const existingWorkspace = c.providerSpecificData?.workspaceId;
+            return existingWorkspace === tokenData.providerSpecificData.workspaceId;
+          }
+          return true;
+        });
         const matchId = typeof match?.id === "string" ? match.id : null;
         if (matchId) {
           connection = await updateProviderConnection(matchId, {
@@ -285,9 +291,15 @@ export async function POST(
         let connection: any;
         if (result.tokens.email) {
           const existing = await getProviderConnections({ provider });
-          const match = existing.find(
-            (c: any) => c.email === result.tokens.email && c.authType === "oauth"
-          );
+          const match = existing.find((c: any) => {
+            if (c.email !== result.tokens.email || c.authType !== "oauth") return false;
+            // For Codex, also check workspaceId to avoid overwriting different workspace connections
+            if (provider === "codex" && result.tokens.providerSpecificData?.workspaceId) {
+              const existingWorkspace = c.providerSpecificData?.workspaceId;
+              return existingWorkspace === result.tokens.providerSpecificData.workspaceId;
+            }
+            return true;
+          });
           const matchId = typeof match?.id === "string" ? match.id : null;
           if (matchId) {
             connection = await updateProviderConnection(matchId, {
@@ -399,9 +411,15 @@ export async function POST(
         let connection: any;
         if (tokenData.email) {
           const existing = await getProviderConnections({ provider });
-          const match = existing.find(
-            (c: any) => c.email === tokenData.email && c.authType === "oauth"
-          );
+          const match = existing.find((c: any) => {
+            if (c.email !== tokenData.email || c.authType !== "oauth") return false;
+            // For Codex, also check workspaceId to avoid overwriting different workspace connections
+            if (provider === "codex" && tokenData.providerSpecificData?.workspaceId) {
+              const existingWorkspace = c.providerSpecificData?.workspaceId;
+              return existingWorkspace === tokenData.providerSpecificData.workspaceId;
+            }
+            return true;
+          });
           const matchId = typeof match?.id === "string" ? match.id : null;
           if (matchId) {
             connection = await updateProviderConnection(matchId, {

src/app/api/v1/images/generations/route.ts

@@ -107,7 +107,30 @@ export async function POST(request) {
   if (policy.rejection) return policy.rejection;
 
   // Parse model to get provider
-  const { provider } = parseImageModel(body.model);
+  let { provider } = parseImageModel(body.model);
+  let isCustomModel = false;
+
+  // If not in built-in registry, check custom models tagged for images
+  if (!provider) {
+    try {
+      const customModelsMap = (await getAllCustomModels()) as Record<string, any>;
+      for (const [providerId, models] of Object.entries(customModelsMap)) {
+        if (!Array.isArray(models)) continue;
+        for (const model of models) {
+          if (!model?.id || !Array.isArray(model.supportedEndpoints)) continue;
+          if (!model.supportedEndpoints.includes("images")) continue;
+          const fullId = `${providerId}/${model.id}`;
+          if (fullId === body.model) {
+            provider = providerId;
+            isCustomModel = true;
+            break;
+          }
+        }
+        if (provider) break;
+      }
+    } catch {}
+  }
+
   if (!provider) {
     return errorResponse(
       HTTP_STATUS.BAD_REQUEST,
@@ -128,9 +151,23 @@ export async function POST(request) {
         `No credentials for image provider: ${provider}`
       );
     }
+  } else if (isCustomModel) {
+    // Custom models need credentials from the provider connection
+    credentials = await getProviderCredentials(provider);
+    if (!credentials) {
+      return errorResponse(
+        HTTP_STATUS.BAD_REQUEST,
+        `No credentials for custom image provider: ${provider}`
+      );
+    }
   }
 
-  const result = await handleImageGeneration({ body, credentials, log });
+  const result = await handleImageGeneration({
+    body,
+    credentials,
+    log,
+    ...(isCustomModel && { resolvedProvider: provider }),
+  });
 
   if (result.success) {
     return new Response(JSON.stringify((result as any).data), {

src/sse/handlers/chat.ts

@@ -369,7 +369,14 @@ async function resolveModelOrError(modelStr: string, body: any) {
   const { provider, model } = modelInfo;
   const sourceFormat = detectFormat(body);
   const providerAlias = PROVIDER_ID_TO_ALIAS[provider] || provider;
-  const targetFormat = getModelTargetFormat(providerAlias, model) || getTargetFormat(provider);
+
+  // If the custom model specifies apiFormat="responses", override targetFormat
+  // to route through the Responses API translator instead of Chat Completions
+  let targetFormat = getModelTargetFormat(providerAlias, model) || getTargetFormat(provider);
+  if ((modelInfo as any).apiFormat === "responses") {
+    targetFormat = "openai-responses";
+    log.info("ROUTING", `Custom model apiFormat=responses → targetFormat=openai-responses`);
+  }
 
   if (modelStr !== `${provider}/${model}`) {
     log.info("ROUTING", `${modelStr} → ${provider}/${model}`);

src/sse/handlers/chatHelpers.ts

@@ -34,7 +34,12 @@ const HTTP_STATUS = {
  * @param {Function} errorResponse - Error response factory
  * @returns {Promise<{ error?: Response, provider: string, model: string, sourceFormat: string, targetFormat: string }>}
  */
-export async function resolveModelOrError(modelStr: string, body: any, log: any, errorResponse: Function) {
+export async function resolveModelOrError(
+  modelStr: string,
+  body: any,
+  log: any,
+  errorResponse: Function
+) {
   const modelInfo = await getModelInfo(modelStr);
 
   if (!modelInfo.provider) {
@@ -44,7 +49,8 @@ export async function resolveModelOrError(modelStr: string, body: any, log: any,
         `Ambiguous model '${modelStr}'. Use provider/model prefix (ex: gh/${modelStr} or cc/${modelStr}).`;
       log.warn("CHAT", message, {
         model: modelStr,
-        candidates: (modelInfo as any).candidateAliases || (modelInfo as any).candidateProviders || [],
+        candidates:
+          (modelInfo as any).candidateAliases || (modelInfo as any).candidateProviders || [],
       });
       return { error: errorResponse(HTTP_STATUS.BAD_REQUEST, message) };
     }
@@ -56,7 +62,14 @@ export async function resolveModelOrError(modelStr: string, body: any, log: any,
   const { provider, model } = modelInfo;
   const sourceFormat = detectFormat(body);
   const providerAlias = PROVIDER_ID_TO_ALIAS[provider] || provider;
-  const targetFormat = getModelTargetFormat(providerAlias, model) || getTargetFormat(provider);
+
+  // If the custom model specifies apiFormat="responses", override targetFormat
+  // to route through the Responses API translator instead of Chat Completions
+  let targetFormat = getModelTargetFormat(providerAlias, model) || getTargetFormat(provider);
+  if ((modelInfo as any).apiFormat === "responses") {
+    targetFormat = "openai-responses";
+    log.info("ROUTING", `Custom model apiFormat=responses → targetFormat=openai-responses`);
+  }
 
   // Log routing
   if (modelStr !== `${provider}/${model}`) {

src/sse/services/model.ts

@@ -1,5 +1,5 @@
 // Re-export from open-sse with localDb integration
-import { getModelAliases, getComboByName, getProviderNodes } from "@/lib/localDb";
+import { getModelAliases, getComboByName, getProviderNodes, getCustomModels } from "@/lib/localDb";
 import {
   parseModel,
   resolveModelAliasFromMap,
@@ -16,13 +16,30 @@ export async function resolveModelAlias(alias) {
   return resolveModelAliasFromMap(alias, aliases);
 }
 
+/**
+ * Look up the apiFormat for a custom model from the DB.
+ * Returns "responses" if the model is configured for the Responses API, otherwise undefined.
+ */
+async function lookupCustomModelApiFormat(
+  providerId: string,
+  modelId: string
+): Promise<string | undefined> {
+  try {
+    const models = await getCustomModels(providerId);
+    if (!Array.isArray(models)) return undefined;
+    const match = models.find((m: any) => m.id === modelId);
+    return match?.apiFormat === "responses" ? "responses" : undefined;
+  } catch {
+    return undefined;
+  }
+}
+
 /**
  * Get full model info (parse or resolve)
  */
 export async function getModelInfo(modelStr) {
   const parsed = parseModel(modelStr);
 
-  // Check custom provider nodes first (for both alias and non-alias formats)
   // Check custom provider nodes first (for both alias and non-alias formats)
   if (parsed.providerAlias || parsed.provider) {
     // Ensure prefixToCheck is always a concise identifier, not a full model string
@@ -32,14 +49,26 @@ export async function getModelInfo(modelStr) {
     const openaiNodes = await getProviderNodes({ type: "openai-compatible" });
     const matchedOpenAI = openaiNodes.find((node) => node.prefix === prefixToCheck);
     if (matchedOpenAI) {
-      return { provider: matchedOpenAI.id, model: parsed.model };
+      const apiFormat = await lookupCustomModelApiFormat(
+        matchedOpenAI.id as string,
+        parsed.model as string
+      );
+      return { provider: matchedOpenAI.id, model: parsed.model, ...(apiFormat && { apiFormat }) };
     }
 
     // Check Anthropic Compatible nodes
     const anthropicNodes = await getProviderNodes({ type: "anthropic-compatible" });
     const matchedAnthropic = anthropicNodes.find((node) => node.prefix === prefixToCheck);
     if (matchedAnthropic) {
-      return { provider: matchedAnthropic.id, model: parsed.model };
+      const apiFormat = await lookupCustomModelApiFormat(
+        matchedAnthropic.id as string,
+        parsed.model as string
+      );
+      return {
+        provider: matchedAnthropic.id,
+        model: parsed.model,
+        ...(apiFormat && { apiFormat }),
+      };
     }
   }