yc-software · importantitemzl-tech · Aug 30, 2026
diff --git a/plugins/admin/public/index.html b/plugins/admin/public/index.html
@@ -3979,11 +3979,11 @@ <h2>Custom providers</h2>
                     placeholder="Write-only; blank on edit keeps the stored key"
                 /></label>
                 <label
-                  >Models (one per line: id | name | context | maxTokens)
+                  >Models (one per line: id | name | context | maxTokens | input modalities | input $/M | output $/M)
                   <textarea
                     id="custom-provider-models"
                     rows="3"
-                    placeholder="deepseek-chat | DeepSeek V3.2 | 128000 | 8192"
+                    placeholder="deepseek-chat | DeepSeek V3.2 | 128000 | 8192&#10;vision-model | Vision Model | 128000 | 8192 | text,image | 0 | 0"
                   ></textarea>
                 </label>
                 <label class="inline"
@@ -7345,14 +7345,36 @@ <h2 id="governance-review-title">Confirm governance change</h2>
           .map((line) => line.trim())
           .filter(Boolean)
           .map((line) => {
-            const [id, name, contextWindow, maxTokens] = line.split("|").map((part) => part.trim());
+            const [id, name, contextWindow, maxTokens, modalities, input, output] = line
+              .split("|")
+              .map((part) => part.trim());
             const model = { id };
             if (name) model.name = name;
             if (contextWindow) model.contextWindow = Number(contextWindow);
             if (maxTokens) model.maxTokens = Number(maxTokens);
+            if (modalities) model.modalities = modalities.split(",").map((part) => part.trim());
+            if (input) model.input = Number(input);
+            if (output) model.output = Number(output);
             return model;
           });
       }
+      function formatCustomModels(models) {
+        return models
+          .map((model) => {
+            const parts = [
+              model.id,
+              model.name,
+              model.contextWindow,
+              model.maxTokens,
+              Array.isArray(model.modalities) ? model.modalities.join(",") : "",
+              model.input,
+              model.output,
+            ];
+            while (parts.length > 1 && (parts[parts.length - 1] == null || parts[parts.length - 1] === "")) parts.pop();
+            return parts.map((part) => part ?? "").join(" | ");
+          })
+          .join("\n");
+      }
       async function loadCustomProviders() {
         const res = await api("GET", "/api/custom-providers");
         if (!res.ok) return;
@@ -7383,11 +7405,7 @@ <h2 id="governance-review-title">Confirm governance change</h2>
             $("custom-provider-protocol").value = provider.protocol;
             $("custom-provider-url").value = provider.baseUrl;
             $("custom-provider-key").value = "";
-            $("custom-provider-models").value = provider.models
-              .map((model) =>
-                [model.id, model.name, model.contextWindow, model.maxTokens].filter((part) => part != null).join(" | "),
-              )
-              .join("\n");
+            $("custom-provider-models").value = formatCustomModels(provider.models);
           };
           const remove = document.createElement("button");
           remove.className = "danger";

diff --git a/plugins/admin/test/onboarding-view.test.ts b/plugins/admin/test/onboarding-view.test.ts
@@ -28,6 +28,15 @@ function resolveView(pathname: string, search: string): string {
   return vm.runInContext(src, context);
 }
 
+function customModelFields(): {
+  parseCustomModels(text: string): unknown[];
+  formatCustomModels(models: unknown[]): string;
+} {
+  const source = slice("function parseCustomModels(text) {", "async function loadCustomProviders()");
+  const context = vm.createContext({});
+  return vm.runInContext(`${source}; ({ parseCustomModels, formatCustomModels })`, context);
+}
+
 test("onboarding is a navigable view", () => {
   assert.match(html, /\{ label: "Admin", views: \["onboarding",/);
 });
@@ -43,3 +52,23 @@ test("?view=onboarding resolves to the onboarding view", () => {
 test("unknown views still fall back to the default view", () => {
   assert.equal(resolveView("/admin/no-such-view", ""), "history");
 });
+
+test("custom model fields preserve optional positions, modalities, and pricing", () => {
+  const { parseCustomModels, formatCustomModels } = customModelFields();
+  const models = [
+    {
+      id: "vision-model",
+      contextWindow: 128000,
+      maxTokens: 8192,
+      modalities: ["text", "image"],
+      input: 1.25,
+      output: 4.5,
+    },
+  ];
+  const formatted = formatCustomModels(models);
+  assert.equal(formatted, "vision-model |  | 128000 | 8192 | text,image | 1.25 | 4.5");
+  assert.deepEqual(JSON.parse(JSON.stringify(parseCustomModels(formatted))), models);
+  assert.deepEqual(JSON.parse(JSON.stringify(parseCustomModels("legacy | Legacy | 64000 | 4096"))), [
+    { id: "legacy", name: "Legacy", contextWindow: 64000, maxTokens: 4096 },
+  ]);
+});
diff --git a/src/harness/opencode-harness.ts b/src/harness/opencode-harness.ts
@@ -8,7 +8,7 @@ import { pathToFileURL } from "node:url";
 import { spawn, type ChildProcess } from "node:child_process";
 import { createOpencodeClient, type OpencodeClient } from "@opencode-ai/sdk";
 import { CONFIG_DEFAULTS, type Config } from "../config.ts";
-import { isCustomModelId } from "../model/custom-providers.ts";
+import { customModelInputModalities, isCustomModelId } from "../model/custom-providers.ts";
 import type { CustomProviderSpec } from "../model/custom-providers.ts";
 import { DEFAULT_AGENT_MODEL_ID, resolveModel } from "../model/pi-models.ts";
 import { startSignalPoll, type RunSignalStore } from "../runs/run-signal-store.ts";
@@ -670,6 +670,7 @@ export function createOpenCodeHarness(opts: OpenCodeHarnessOptions = {}): Harnes
                   m.id,
                   {
                     name: m.name ?? m.id,
+                    modalities: { input: customModelInputModalities(m), output: ["text"] },
                     ...(m.contextWindow || m.maxTokens
                       ? {
                           limit: {

diff --git a/src/model/custom-provider-store.ts b/src/model/custom-provider-store.ts
@@ -9,7 +9,7 @@
 
 import { decryptSecret, deriveConnectorKey, encryptSecret } from "../connectors/connector-client-store.ts";
 import type { DurableMap } from "../persistence/durable-map.ts";
-import { validateCustomProviderSpec, type CustomProviderSpec } from "./custom-providers.ts";
+import { customModelInputModalities, validateCustomProviderSpec, type CustomProviderSpec } from "./custom-providers.ts";
 
 export interface StoredCustomProvider extends CustomProviderSpec {
   apiKeyEnc?: string;
@@ -42,7 +42,12 @@ function strip(saved: StoredCustomProvider): CustomProviderSpec {
     name: saved.name,
     protocol: saved.protocol,
     baseUrl: saved.baseUrl,
-    models: saved.models,
+    models: saved.models.map((model) => {
+      const stored = model as typeof model & { modalities?: unknown };
+      return stored.modalities === undefined
+        ? { ...model }
+        : { ...model, modalities: customModelInputModalities(stored) };
+    }),
   };
 }
 

diff --git a/src/model/custom-providers.ts b/src/model/custom-providers.ts
@@ -23,6 +23,7 @@ interface CustomModelSpec {
   name?: string;
   contextWindow?: number;
   maxTokens?: number;
+  modalities?: ("text" | "image")[];
   /** USD per million input tokens. Defaults to 0 (unknown / not metered). */
   input?: number;
   /** USD per million output tokens. Defaults to 0. */
@@ -63,6 +64,9 @@ export function validateCustomProviderSpec(spec: CustomProviderSpec): void {
       throw new Error(`model "${m.id}": name must be a string of 200 chars or fewer`);
     if (seen.has(m.id)) throw new Error(`duplicate model id "${m.id}"`);
     seen.add(m.id);
+    if (m.modalities !== undefined && !isCustomModelInputModalities(m.modalities)) {
+      throw new Error(`model "${m.id}": modalities must contain text and optional image exactly once`);
+    }
     for (const [field, v] of [
       ["contextWindow", m.contextWindow],
       ["maxTokens", m.maxTokens],
@@ -97,6 +101,23 @@ export interface CustomRuntimeModel {
 const DEFAULT_CONTEXT_WINDOW = 128_000;
 const DEFAULT_MAX_TOKENS = 8_192;
 
+function isCustomModelInputModalities(value: unknown): value is ("text" | "image")[] {
+  return (
+    Array.isArray(value) &&
+    value.includes("text") &&
+    new Set(value).size === value.length &&
+    value.every((modality) => modality === "text" || modality === "image")
+  );
+}
+
+export function customModelInputModalities(model: { modalities?: unknown }): ("text" | "image")[] {
+  const value =
+    model.modalities && typeof model.modalities === "object" && !Array.isArray(model.modalities)
+      ? (model.modalities as { input?: unknown }).input
+      : model.modalities;
+  return isCustomModelInputModalities(value) ? [...value] : ["text"];
+}
+
 function toRuntimeModel(provider: CustomProviderSpec, m: CustomModelSpec): CustomRuntimeModel {
   return {
     id: m.id,
@@ -105,7 +126,7 @@ function toRuntimeModel(provider: CustomProviderSpec, m: CustomModelSpec): Custo
     api: provider.protocol === "anthropic" ? "anthropic-messages" : "openai-completions",
     baseUrl: provider.baseUrl,
     reasoning: false,
-    input: ["text"],
+    input: customModelInputModalities(m),
     cost: { input: m.input ?? 0, output: m.output ?? 0, cacheRead: 0, cacheWrite: 0 },
     contextWindow: m.contextWindow ?? DEFAULT_CONTEXT_WINDOW,
     maxTokens: m.maxTokens ?? DEFAULT_MAX_TOKENS,
@@ -173,6 +194,7 @@ export function customModelsJson(): { providers: Record<string, unknown> } | und
             name: m.name ?? m.id,
             contextWindow: m.contextWindow ?? 128_000,
             maxTokens: m.maxTokens ?? 8_192,
+            input: customModelInputModalities(m),
             cost: { input: m.input ?? 0, output: m.output ?? 0, cacheRead: 0, cacheWrite: 0 },
           })),
         },

diff --git a/test/custom-provider-e2e.test.ts b/test/custom-provider-e2e.test.ts
@@ -13,7 +13,8 @@ import { test } from "node:test";
 import { createInsecureTestServer } from "../src/api/server.ts";
 import { buildApp } from "../src/wiring.ts";
 import { testConfig } from "./support/test-config.ts";
-import { oneShot } from "../src/harness/pi-harness.ts";
+import { createPiHarness, oneShot } from "../src/harness/pi-harness.ts";
+import type { HarnessTurnInput } from "../src/harness/harness.ts";
 import { resolveModel, modelSupportedByHarness, modelServiceable } from "../src/model/pi-models.ts";
 import { setCustomProviders } from "../src/model/custom-providers.ts";
 import { createCustomProviderStore } from "../src/model/custom-provider-store.ts";
@@ -24,7 +25,7 @@ const ADMIN = { "content-type": "application/json", "x-admin-actor": "admin-alic
 
 test("QA: full custom-provider lifecycle against a live fake upstream", async () => {
   // --- fake OpenAI-compatible upstream ---
-  const seen: Array<{ path: string; auth: string | undefined; model?: string }> = [];
+  const seen: Array<{ path: string; auth: string | undefined; model?: string; body?: unknown }> = [];
   const upstream = createServer((req, res) => {
     let body = "";
     req.on("data", (c) => (body += c));
@@ -40,7 +41,8 @@ test("QA: full custom-provider lifecycle against a live fake upstream", async ()
         return res.end(JSON.stringify({ data: [{ id: "qa-chat" }] }));
       }
       if (req.url?.endsWith("/chat/completions")) {
-        record.model = (JSON.parse(body) as { model?: string }).model;
+        record.body = JSON.parse(body);
+        record.model = (record.body as { model?: string }).model;
         seen.push(record);
         res.writeHead(200, { "content-type": "text/event-stream" });
         const chunk = (delta: object, finish: string | null) =>
@@ -127,7 +129,17 @@ test("QA: full custom-provider lifecycle against a live fake upstream", async ()
         protocol: "openai",
         baseUrl: upstreamUrl,
         apiKey: "sk-qa-good",
-        models: [{ id: "qa-chat", name: "QA Chat", contextWindow: 64000, maxTokens: 4096 }],
+        models: [
+          {
+            id: "qa-chat",
+            name: "QA Chat",
+            contextWindow: 64000,
+            maxTokens: 4096,
+            modalities: ["text", "image"],
+            input: 1.5,
+            output: 6,
+          },
+        ],
       }),
     });
     assert.equal(r.status, 200);
@@ -166,14 +178,44 @@ test("QA: full custom-provider lifecycle against a live fake upstream", async ()
     assert.equal(call!.model, "qa-chat");
     assert.equal(call!.auth, "Bearer sk-qa-good", "stored key was sent to the custom endpoint");
 
+    const imageHarness = createPiHarness({
+      modelId: "qa-chat",
+      resolveProviderKeys: async () => ({ qa: "sk-qa-good" }),
+      captureRequests: false,
+      turnWallClockMs: 5_000,
+    });
+    const imageData = "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mNk+A8AAQUBAScY42YAAAAASUVORK5CYII=";
+    const imageTurn = {
+      session: { id: "qa-image" } as HarnessTurnInput["session"],
+      input: "describe this image",
+      images: [{ mimeType: "image/png", dataBase64: imageData }],
+      model: "qa-chat",
+      systemPrompt: "be concise",
+      history: [],
+      tools: {} as HarnessTurnInput["tools"],
+      scopeLabel: "org:test" as HarnessTurnInput["scopeLabel"],
+      orgScopeId: "org:test" as HarnessTurnInput["orgScopeId"],
+      emit: async (entry) =>
+        ({ ...entry, seq: 1, sessionId: "qa-image", createdAt: Date.now() }) as Awaited<
+          ReturnType<HarnessTurnInput["emit"]>
+        >,
+      recordModelCall: () => {},
+    } satisfies HarnessTurnInput;
+    const imageReply = await imageHarness.turns.runTurn(imageTurn);
+    assert.equal(imageReply.reply, "QA UPSTREAM REPLY");
+    const imageCall = seen.filter((item) => item.path.endsWith("/chat/completions")).at(-1);
+    const imagePayload = JSON.stringify(imageCall?.body);
+    assert.ok(imagePayload.includes(`data:image/png;base64,${imageData}`));
+    assert.doesNotMatch(imagePayload, /image omitted: model does not support images/);
+
     // 7. edit WITHOUT key keeps the stored key
     r = await api("/v1/admin/custom-providers/qa", {
       method: "PUT",
       body: JSON.stringify({
         name: "QA Provider v2",
         protocol: "openai",
         baseUrl: upstreamUrl,
-        models: [{ id: "qa-chat" }],
+        models: [{ id: "qa-chat", modalities: ["text", "image"], input: 1.5, output: 6 }],
       }),
     });
     assert.equal(r.status, 200);