Remove image upload; switch to text-only coding assistant

The model is agentic-coding focused and does not accept images. Drop the
image upload path from both the frontend and backend and use the text-only
serving recipe from the model card: AutoTokenizer + AutoModelForCausalLM
with tokenizer.apply_chat_template(..., tokenize=False).

Also keeps the terminal-style UI and Ornith logo header.

Co-Authored-By: GLM-5.2 <noreply@z.ai>

app.py

@@ -1,7 +1,7 @@
 """
 Ornith-1.0-9B — a gradio.Server chat app.
 
-This Space demonstrates `deepreinforce-ai/Ornith-1.0-9B`, a multimodal
+This Space demonstrates `deepreinforce-ai/Ornith-1.0-9B`, an agentic coding
 (qwen3.5-based) reasoning model, served through `gradio.Server`.
 
 `gradio.Server` extends FastAPI and adds Gradio's API engine on top
@@ -21,16 +21,14 @@ from threading import Thread
 from typing import Optional
 
 import torch
-from PIL import Image
 from transformers import (
-    AutoModelForMultimodalLM,
-    AutoProcessor,
+    AutoModelForCausalLM,
+    AutoTokenizer,
     TextIteratorStreamer,
 )
 
 from fastapi.responses import HTMLResponse
 from gradio import Server
-from gradio.data_classes import FileData
 
 # `spaces` is only available on Hugging Face Spaces. Guard the import so the
 # file still parses/imports on a plain machine (the GPU decorator becomes a
@@ -74,8 +72,8 @@ THINK_CLOSE = None  # resolved lazily from tokenizer config in _split_think
 # and call .to("cuda") once - the CUDA emulation handles that safely. Real
 # CUDA then becomes available inside the @spaces.GPU-decorated function.
 print(f"[ornith] loading {MODEL_ID} ...", flush=True)
-_processor = AutoProcessor.from_pretrained(MODEL_ID)
-_model = AutoModelForMultimodalLM.from_pretrained(
+_tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
+_model = AutoModelForCausalLM.from_pretrained(
     MODEL_ID,
     torch_dtype=torch.bfloat16,
     low_cpu_mem_usage=True,
@@ -84,16 +82,6 @@ _model.to("cuda")
 _model.eval()
 print("[ornith] model ready on cuda", flush=True)
 
-_tokenizer = _processor.tokenizer if hasattr(_processor, "tokenizer") else _processor
-
-
-def _file_path(image):
-    if image is None:
-        return None
-    if isinstance(image, dict):
-        return image.get("path")
-    return getattr(image, "path", None)
-
 
 def _resolve_markers():
     open_tok = None
@@ -127,27 +115,6 @@ def _split_think(text, open_tok, close_tok):
     return text.replace(open_tok or "", "").strip(), ""
 
 
-def _build_messages(message, image_path, history):
-    messages = []
-    for turn in history or []:
-        role = turn.get("role")
-        content = turn.get("content")
-        if role in ("system", "user", "assistant") and content:
-            messages.append({"role": role, "content": content})
-
-    parts = []
-    if image_path:
-        try:
-            img = Image.open(image_path).convert("RGB")
-            parts.append({"type": "image", "image": img})
-        except Exception:
-            parts.append({"type": "image", "image": image_path})
-    parts.append({"type": "text", "text": message or ""})
-
-    messages.append({"role": "user", "content": parts})
-    return messages
-
-
 # --------------------------------------------------------------------------- #
 # App + streaming endpoint
 # --------------------------------------------------------------------------- #
@@ -158,7 +125,6 @@ app = Server()
 @GPU(duration=300)
 def generate(
     message: str,
-    image: Optional[FileData] = None,
     history: list = None,
     max_new_tokens: int = DEFAULT_MAX_NEW_TOKENS,
     temperature: float = DEFAULT_TEMPERATURE,
@@ -167,23 +133,23 @@ def generate(
     # {reasoning, answer, status, error}.
     if history is None:
         history = []
-    image_path = _file_path(image)
     open_tok, close_tok = _resolve_markers()
 
     try:
-        messages = _build_messages(message, image_path, history)
-
-        inputs = _processor.apply_chat_template(
+        messages = []
+        for turn in history:
+            role = turn.get("role")
+            content = turn.get("content")
+            if role in ("system", "user", "assistant") and content:
+                messages.append({"role": role, "content": content})
+        messages.append({"role": "user", "content": message or ""})
+
+        prompt = _tokenizer.apply_chat_template(
             messages,
+            tokenize=False,
             add_generation_prompt=True,
-            tokenize=True,
-            return_dict=True,
-            return_tensors="pt",
         )
-        inputs = {
-            k: (v.to("cuda") if hasattr(v, "to") else v)
-            for k, v in inputs.items()
-        }
+        inputs = _tokenizer(prompt, return_tensors="pt").to("cuda")
 
         streamer = TextIteratorStreamer(
             _tokenizer,

index.html

@@ -3,7 +3,7 @@
 <head>
 <meta charset="UTF-8" />
 <meta name="viewport" content="width=device-width, initial-scale=1.0" />
-<title>Ornith-1.0-9B · Claude Code</title>
+<title>Ornith-1.0-9B · Terminal</title>
 <style>
   :root {
     --bg: #0a0a0a;
@@ -65,7 +65,6 @@
   .content { flex: 1; min-width: 0; }
 
   .user-input { color: var(--text); white-space: pre-wrap; }
-  .user-image { margin-top: 8px; max-width: 240px; border: 1px solid var(--border); border-radius: 4px; display: block; }
 
   .thinking { margin: 6px 0 10px; }
   .thinking summary {
@@ -113,15 +112,6 @@
   #settings label { display: flex; flex-direction: column; gap: 4px; }
   #settings input[type=range] { width: 150px; accent-color: var(--green); }
 
-  .previews { display: flex; gap: 8px; flex-wrap: wrap; margin-bottom: 8px; }
-  .previews .thumb { position: relative; }
-  .previews .thumb img { height: 48px; border-radius: 4px; border: 1px solid var(--border); }
-  .previews .thumb .x {
-    position: absolute; top: -6px; right: -6px; cursor: pointer;
-    background: var(--panel-2); border: 1px solid var(--border); border-radius: 50%;
-    width: 16px; height: 16px; text-align: center; line-height: 14px; font-size: 11px;
-  }
-
   .input-row { display: flex; align-items: flex-end; gap: 8px; }
   #text {
     flex: 1; resize: none; min-height: 42px; max-height: 160px;
@@ -157,7 +147,7 @@
 <div id="chat">
   <div class="welcome">
     <h2>Welcome to Ornith-1.0-9B</h2>
-    <p>Agentic coding assistant. Ask me to write, debug, or explain code. You can also attach an image for multimodal questions.</p>
+    <p>Agentic coding assistant. Ask me to write, debug, or explain code.</p>
   </div>
 </div>
 <div class="status" id="status"></div>
@@ -168,32 +158,25 @@
       <label>Temperature <input type="range" id="temp" min="0.1" max="1.5" step="0.1" value="0.6" /><span id="temp-v">0.6</span></label>
       <label>Max tokens <input type="range" id="maxtok" min="256" max="8192" step="256" value="2048" /><span id="maxtok-v">2048</span></label>
     </div>
-    <div class="previews" id="previews"></div>
     <div class="input-row">
-      <button class="icon-btn" id="attach" title="Attach image">📎</button>
       <textarea id="text" placeholder="Ask Ornith to write code, explain, or debug…"></textarea>
       <button id="send">Send</button>
       <button id="stop">Stop</button>
     </div>
-    <div class="hint">Shift+Enter for newline · Enter to send · Image-text-to-text supported</div>
+    <div class="hint">Shift+Enter for newline · Enter to send</div>
   </div>
 </div>
 
-<input type="file" id="file" accept="image/*" hidden />
-
 <script src="https://cdn.jsdelivr.net/npm/marked/marked.min.js"></script>
 <script src="https://cdn.jsdelivr.net/npm/dompurify@3/dist/purify.min.js"></script>
 <script type="module">
-import { Client, handle_file } from "https://cdn.jsdelivr.net/npm/@gradio/client/dist/index.min.js";
+import { Client } from "https://cdn.jsdelivr.net/npm/@gradio/client/dist/index.min.js";
 
 const chatEl   = document.getElementById("chat");
 const statusEl = document.getElementById("status");
 const textEl   = document.getElementById("text");
 const sendBtn  = document.getElementById("send");
 const stopBtn  = document.getElementById("stop");
-const attachBtn= document.getElementById("attach");
-const fileEl   = document.getElementById("file");
-const prevEl   = document.getElementById("previews");
 const setToggle= document.getElementById("settings-toggle");
 const settings = document.getElementById("settings");
 const tempEl   = document.getElementById("temp");
@@ -202,7 +185,6 @@ const maxEl    = document.getElementById("maxtok");
 const maxV     = document.getElementById("maxtok-v");
 
 let conversation = [];
-let pendingImage = null;
 let client = null;
 let job = null;
 
@@ -215,26 +197,6 @@ setToggle.addEventListener("click", () => settings.classList.toggle("open"));
 tempEl.addEventListener("input", () => tempV.textContent = tempEl.value);
 maxEl.addEventListener("input", () => maxV.textContent = maxEl.value);
 
-attachBtn.addEventListener("click", () => fileEl.click());
-fileEl.addEventListener("change", () => {
-  if (fileEl.files && fileEl.files[0]) {
-    pendingImage = fileEl.files[0];
-    renderPreviews();
-  }
-  fileEl.value = "";
-});
-
-function renderPreviews() {
-  prevEl.innerHTML = "";
-  if (!pendingImage) return;
-  const url = URL.createObjectURL(pendingImage);
-  const wrap = document.createElement("div");
-  wrap.className = "thumb";
-  wrap.innerHTML = `<img src="${url}" alt="attachment"/><div class="x" title="remove">×</div>`;
-  wrap.querySelector(".x").addEventListener("click", () => { pendingImage = null; renderPreviews(); });
-  prevEl.appendChild(wrap);
-}
-
 textEl.addEventListener("input", () => {
   textEl.style.height = "auto";
   textEl.style.height = Math.min(textEl.scrollHeight, 160) + "px";
@@ -273,19 +235,11 @@ function addTurn(role) {
 
 async function send() {
   const message = textEl.value.trim();
-  if (!message && !pendingImage) return;
+  if (!message) return;
   if (job) return;
 
   const userContent = addTurn("user");
-  let html = `<div class="user-input">${escapeHtml(message)}</div>`;
-  if (pendingImage) {
-    const url = URL.createObjectURL(pendingImage);
-    html += `<img class="user-image" src="${url}" alt="attachment"/>`;
-  }
-  userContent.innerHTML = html;
-
-  const sentImage = pendingImage;
-  pendingImage = null; renderPreviews();
+  userContent.innerHTML = `<div class="user-input">${escapeHtml(message)}</div>`;
   textEl.value = ""; textEl.style.height = "auto";
 
   const history = conversation.slice();
@@ -301,7 +255,6 @@ async function send() {
     const c = await connect();
     const payload = {
       message,
-      image: sentImage ? handle_file(sentImage) : null,
       history,
       max_new_tokens: parseInt(maxEl.value, 10),
       temperature: parseFloat(tempEl.value),