fix: added

server.js

@@ -81,6 +81,38 @@ async function runInference(imageBuffer, prompt, maxTokens) {
   return decoded[0];
 }
 
+async function runTextInference(prompt, maxTokens) {
+  const conversation = [
+    {
+      role: "user",
+      content: [{ type: "text", text: prompt }],
+    },
+  ];
+
+  const text = processor.apply_chat_template(conversation, {
+    add_generation_prompt: true,
+  });
+
+  const inputs = await processor(text);
+  const output = await model.generate({
+    ...inputs,
+    max_new_tokens: maxTokens,
+  });
+
+  const promptLength = inputs.input_ids.dims.at(-1);
+  const decoded = processor.batch_decode(
+    output.slice(null, [promptLength, null]),
+    { skip_special_tokens: true },
+  );
+  return decoded[0];
+}
+
+function queueTextInference(prompt, maxTokens) {
+  const task = inferenceQueue.then(() => runTextInference(prompt, maxTokens));
+  inferenceQueue = task.catch(() => {});
+  return task;
+}
+
 function queueInference(imageBuffer, prompt, maxTokens) {
   const task = inferenceQueue.then(() => runInference(imageBuffer, prompt, maxTokens));
   inferenceQueue = task.catch(() => {});
@@ -133,6 +165,31 @@ const swaggerDoc = {
         },
       },
     },
+    "/prompt": {
+      post: {
+        summary: "Text prompt inference (no image)",
+        requestBody: {
+          required: true,
+          content: {
+            "application/json": {
+              schema: {
+                type: "object",
+                required: ["prompt"],
+                properties: {
+                  prompt: { type: "string", description: "Text prompt to send to the model" },
+                  max_tokens: { type: "integer", default: 256 },
+                },
+              },
+            },
+          },
+        },
+        responses: {
+          200: { description: "Inference result" },
+          400: { description: "Invalid input" },
+          503: { description: "Model not loaded" },
+        },
+      },
+    },
     "/inference/base64": {
       post: {
         summary: "Image inference (base64)",
@@ -199,6 +256,46 @@ app.get("/health", (req, res) => {
   res.json({ status: "healthy", model_loaded: model !== null });
 });
 
+app.post("/prompt", express.json(), async (req, res) => {
+  const prompt = req.body.prompt;
+  const maxTokens = parseInt(req.body.max_tokens) || 256;
+  log("info", "prompt_request_received", {
+    request_id: req.requestId,
+    prompt_chars: prompt?.length ?? 0,
+    max_tokens: maxTokens,
+  });
+
+  if (!model || !processor) {
+    log("error", "prompt_model_unavailable", { request_id: req.requestId });
+    return res.status(503).json({ detail: "Model not loaded yet." });
+  }
+  if (!prompt) {
+    log("error", "prompt_validation_failed", {
+      request_id: req.requestId,
+      reason: "missing_prompt",
+    });
+    return res.status(400).json({ detail: "No prompt provided." });
+  }
+
+  try {
+    const start = Date.now();
+    const response = await queueTextInference(prompt, maxTokens);
+    log("info", "prompt_completed", {
+      request_id: req.requestId,
+      duration_ms: Date.now() - start,
+      response_chars: response?.length ?? 0,
+    });
+    res.json({ response });
+  } catch (err) {
+    log("error", "prompt_failed", {
+      request_id: req.requestId,
+      error: err.message,
+      stack: err.stack,
+    });
+    res.status(500).json({ detail: "Inference failed.", error: err.message });
+  }
+});
+
 app.post("/inference", upload.single("file"), async (req, res) => {
   const prompt = req.body.prompt || "Describe this image in detail.";
   const maxTokens = parseInt(req.body.max_tokens) || 256;