Update app.py

app.py

@@ -1,135 +1,300 @@
+# =============================================================================
+#  codeMax — Qwen 3.6 27B CoderBot on ZeroGPU (Gradio 6)
+#  Model: DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF
+# =============================================================================
 
-"""codeMax — DavidAU coder/chat with file drop-in."""
-
+import spaces
+import gradio as gr
 import os
-import sys
-import subprocess
-
-# ── Runtime install: llama-cpp-python (CPU) ──
-try:
-    import llama_cpp
-except ImportError:
-    print("[RUNTIME] Installing llama-cpp-python (CPU)...", flush=True)
-    subprocess.check_call([
-        sys.executable, "-m", "pip", "install", "--no-cache-dir",
-        "--extra-index-url", "https://abetlen.github.io/llama-cpp-python/whl/cpu",
-        "llama-cpp-python"
-    ])
-    import llama_cpp
-    print("[RUNTIME] llama-cpp-python installed", flush=True)
+import json
+from pathlib import Path
+from typing import List, Optional, Generator
 
-import gradio as gr
-from huggingface_hub import hf_hub_download
-from llama_cpp import Llama
-
-# ── Load DavidAU ──
-REPO_ID = "DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF"
-FILENAME = "Qwen3.6-27B-NEO-CODE-HERE-2T-OT-Q4_K_M.gguf"
-
-print(f"[MODEL] Downloading {FILENAME}...", flush=True)
-model_path = hf_hub_download(repo_id=REPO_ID, filename=FILENAME)
-print(f"[MODEL] Loading...", flush=True)
-
-llm = Llama(
-    model_path=model_path,
-    n_ctx=8192,
-    n_threads=8,
-    verbose=False,
-)
-print("[MODEL] Ready.", flush=True)
-
-
-def build_prompt(message, history, system_prompt, file_content=None):
-    """Build ChatML prompt with optional file content."""
-    parts = []
-    if system_prompt:
-        parts.append(f"<|system|>\n{system_prompt}</s>")
-    if history:
-        for msg in history:
-            role = msg.get("role", "user")
-            content = msg.get("content", "")
-            parts.append(f"<|{role}|>\n{content}</s>")
-    user_msg = message
-    if file_content:
-        user_msg = f"File contents:\n```\n{file_content}\n```\n\nUser: {message}"
-    parts.append(f"<|user|>\n{user_msg}</s>")
-    parts.append("<|assistant|>\n")
-    return "\n".join(parts)
-
-
-def respond(message, history, system_prompt, file_upload):
-    """Generate response from DavidAU."""
-    if not message.strip():
-        return ""
-
-    file_content = None
-    if file_upload is not None:
+# --- Config ----------------------------------------------------------------
+MODEL_REPO = os.environ.get("MODEL_REPO", "DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF")
+MODEL_FILE = os.environ.get("MODEL_FILE", "Qwen3.6-27B-NEO-CODE-HERE-2T-OT-Q4_K_M.gguf")
+N_CTX      = int(os.environ.get("N_CTX", "8192"))
+N_GPU      = int(os.environ.get("N_GPU_LAYERS", "-1"))  # -1 = all layers on GPU
+
+SYSTEM_PROMPT = """You are codeMax, an uncensored 27B coding assistant. You excel at:
+• Writing production code in Python, TypeScript, JavaScript, Rust, Go, C/C++, SQL, shell
+• Debugging, refactoring, reviewing complex codebases
+• Explaining algorithms, architecture, system design
+• Reading shared files and answering about them
+
+Rules: put code in ```language blocks. Be precise and thorough."""
+
+# --- Model singleton -------------------------------------------------------
+_llm = None
+_model_path: Optional[str] = None
+
+def _download():
+    global _model_path
+    if _model_path is not None:
+        return _model_path
+    from huggingface_hub import hf_hub_download
+    print(f"[MODEL] Downloading {MODEL_FILE}...")
+    _model_path = hf_hub_download(repo_id=MODEL_REPO, filename=MODEL_FILE)
+    print(f"[MODEL] Cached → {_model_path}")
+    return _model_path
+
+def _load():
+    global _llm
+    if _llm is not None:
+        return _llm
+    from llama_cpp import Llama
+    path = _download()
+    print(f"[MODEL] Loading (n_gpu_layers={N_GPU})...")
+    _llm = Llama(model_path=path, n_ctx=N_CTX, n_gpu_layers=N_GPU,
+                 chat_format="chatml", verbose=False, seed=-1)
+    print("[MODEL] Ready.")
+    return _llm
+
+# --- File parsers ----------------------------------------------------------
+def _read_text(path, enc="utf-8"):
+    for e in [enc, "utf-8-sig", "latin-1", "cp1252", "utf-16"]:
         try:
-            if isinstance(file_upload, dict):
-                path = file_upload.get("name", file_upload.get("path", ""))
-            else:
-                path = str(file_upload)
-            with open(path, "r", encoding="utf-8", errors="replace") as f:
-                file_content = f.read()
+            with open(path, "r", encoding=e) as f:
+                return f.read()
+        except UnicodeError:
+            continue
+    with open(path, "r", encoding="utf-8", errors="replace") as f:
+        return f.read()
+
+CODE_EXTS = {".txt",".md",".py",".ts",".tsx",".js",".jsx",".mjs",".cjs",
+             ".css",".scss",".less",".html",".htm",".xml",".svg",
+             ".yaml",".yml",".toml",".ini",".cfg",".conf",
+             ".sh",".bash",".zsh",".fish",".ps1",".bat",
+             ".sql",".graphql",".prisma",".cs",
+             ".c",".cpp",".cc",".cxx",".h",".hpp",".hh",
+             ".java",".kt",".kts",".scala",".groovy",
+             ".go",".rs",".rb",".php",".pl",".pm",
+             ".swift",".r",".lua",".zig",".nim",".dart",
+             ".env",".gitignore",".editorconfig",
+             ".tf",".tfvars",".hcl",".vue",".svelte",".astro"}
+
+def parse_file(file_path, file_name):
+    suffix = Path(file_name).suffix.lower()
+    name = Path(file_name).name
+    if suffix in CODE_EXTS:
+        content = _read_text(file_path)
+        lang = suffix.lstrip(".")
+        if suffix == ".md": lang = "markdown"
+        elif suffix in (".yml",): lang = "yaml"
+        elif suffix in (".tf",".tfvars"): lang = "hcl"
+        elif suffix in (".htm",): lang = "html"
+        return f"### 📄 `{name}`\n```{lang}\n{content}\n```\n\n---\n"
+    if suffix == ".json":
+        content = _read_text(file_path)
+        try:
+            parsed = json.loads(content)
+            content = json.dumps(parsed, indent=2, ensure_ascii=False)
         except Exception:
             pass
+        return f"### 📄 `{name}`\n```json\n{content}\n```\n\n---\n"
+    if suffix == ".csv":
+        import csv, io
+        content = _read_text(file_path)
+        rows = list(csv.reader(io.StringIO(content)))
+        if len(rows) > 51:
+            rows = rows[:50] + [[f"… {len(rows)-50} rows truncated"]]
+        w = [max(len(str(c)) for c in col) for col in zip(*rows)]
+        sep = "|-" + "-|-".join("-"*x for x in w) + "-|"
+        t = "\n".join("| " + " | ".join(str(c).ljust(w[i]) for i,c in enumerate(r)) + " |" for r in rows)
+        t = t.split("\n",1); t.insert(1,sep)
+        return f"### 📊 `{name}`\n" + "\n".join(t) + "\n\n---\n"
+    if suffix == ".docx":
+        try:
+            import docx; d=docx.Document(file_path)
+            return f"### 📝 `{name}`\n"+"\n\n".join(p.text for p in d.paragraphs if p.text.strip())+"\n\n---\n"
+        except Exception as e:
+            return f"### ⚠️ `{name}` (DOCX error: {e})\n\n---\n"
+    if suffix == ".pdf":
+        try:
+            import pdfplumber
+            with pdfplumber.open(file_path) as pdf:
+                t = "\n\n".join(p.extract_text() or "" for p in pdf.pages)
+            return f"### 📑 `{name}`\n{t}\n\n---\n"
+        except Exception as e:
+            return f"### ⚠️ `{name}` (PDF error: {e})\n\n---\n"
+    try:
+        return f"### 📄 `{name}`\n```\n{_read_text(file_path)[:10000]}\n```\n\n---\n"
+    except Exception as e:
+        return f"### ❌ `{name}` ({e})\n\n---\n"
 
-    # Convert history to dict format
-    history_dicts = []
-    if history:
-        for h in history:
-            if isinstance(h, (list, tuple)) and len(h) == 2:
-                history_dicts.append({"role": "user", "content": h[0]})
-                history_dicts.append({"role": "assistant", "content": h[1]})
+def parse_all_files(files):
+    if not files: return ""
+    out = []
+    for f in files:
+        if isinstance(f, dict):
+            path = f.get("path") or f.get("name")
+            name = f.get("orig_name") or f.get("name","unknown")
+        elif hasattr(f,"name"):
+            path = f.name; name = getattr(f,"orig_name",Path(path).name)
+        else:
+            path=str(f); name=Path(path).name
+        try:
+            out.append(parse_file(path,name))
+        except Exception as e:
+            out.append(f"### ❌ `{name}`\n{e}\n\n---\n")
+    return "\n".join(out)
 
-    prompt = build_prompt(message, history_dicts, system_prompt, file_content)
+# --- Generation ------------------------------------------------------------
+@spaces.GPU
+def respond(message: str, history: List, uploaded_files) -> Generator[List, None, None]:
+    fc = parse_all_files(uploaded_files)
+    msgs = [{"role":"system","content":SYSTEM_PROMPT}]
+    if fc:
+        msgs.append({"role":"user","content":"[Uploaded files — read carefully]\n\n"+fc})
+        msgs.append({"role":"assistant","content":"Got it! I've read all the files. Ask me anything."})
+    for entry in history:
+        r = entry.get("role","user"); c = entry.get("content","")
+        if c: msgs.append({"role":r,"content":c})
+    msgs.append({"role":"user","content":message})
+    llm = _load()
+    stream = llm.create_chat_completion(messages=msgs, temperature=0.6, top_p=0.8,
+                                        top_k=20, max_tokens=4096, stream=True)
+    partial = ""
+    for chunk in stream:
+        if chunk.get("choices"):
+            d = chunk["choices"][0].get("delta",{}).get("content","")
+            if d: partial += d
+        yield history + [{"role":"user","content":message},
+                         {"role":"assistant","content":partial}]
+    yield history + [{"role":"user","content":message},
+                     {"role":"assistant","content":partial}]
 
-    try:
-        stream = llm(
-            prompt,
-            max_tokens=2048,
-            stop=["</s>", "<|user|>"],
-            stream=True,
-        )
-        response = ""
-        for output in stream:
-            token = output["choices"][0].get("text", "")
-            response += token
-            yield response
-    except Exception as e:
-        yield f"[Error: {e}]"
-
-
-# ── UI ──
-with gr.Blocks(title="codeMax — DavidAU Coder", css="""
-    :root { --bg: #0d1117; --fg: #c9d1d9; --accent: #58a6ff; }
-    body, .gradio-container { background: var(--bg) !important; color: var(--fg) !important; }
-    .gr-button-primary { background: var(--accent) !important; }
-""") as demo:
-    gr.Markdown("# codeMax\n### DavidAU Qwen3.6-27B — Coder + File Drop")
-    with gr.Row():
-        with gr.Column(scale=1):
-            system_box = gr.Textbox(
-                label="System Prompt",
-                value="You are an expert software engineer. Be concise, write clean code.",
-                lines=3,
-            )
-            file_input = gr.File(label="Drop code file here", file_types=[".py", ".js", ".ts", ".tsx", ".json", ".txt", ".md", ".html", ".css"])
-        with gr.Column(scale=3):
-            chatbot = gr.Chatbot(label="Chat", height=480, type="messages")
-            msg_box = gr.Textbox(label="Message", placeholder="Ask me to code something...", lines=2)
-            send_btn = gr.Button("Send", variant="primary")
-
-    send_btn.click(
-        respond,
-        inputs=[msg_box, chatbot, system_box, file_input],
-        outputs=[chatbot],
-    )
-    msg_box.submit(
-        respond,
-        inputs=[msg_box, chatbot, system_box, file_input],
-        outputs=[chatbot],
-    )
+# --- UI --------------------------------------------------------------------
+def create_demo():
+    with gr.Blocks(title="codeMax — Qwen 3.6 27B Coder") as demo:
+        gr.Markdown("# 🧠 codeMax\n**Qwen 3.6 27B · Uncensored · Code-optimized · ZeroGPU**")
+        with gr.Row(equal_height=True):
+            with gr.Column(scale=3):
+                chatbot = gr.Chatbot(label="Chat", height=580,
+                    avatar_images=(None,"https://huggingface.co/front/assets/huggingface_logo-noborder.svg"))
+                with gr.Row():
+                    msg = gr.Textbox(placeholder="Ask about code, share files, or chat...",
+                                     scale=8, show_label=False, container=False)
+                    send = gr.Button("▶", scale=1, variant="primary", min_width=48)
+                clear_btn = gr.Button("🗑 Clear", size="sm")
+            with gr.Column(scale=1, elem_classes="file-upload-col"):
+                gr.Markdown("### 📎 Drop Files")
+                files = gr.File(file_count="multiple", label="Code, docs, data...",
+                    file_types=list(CODE_EXTS)+[".json",".csv",".docx",".pdf"])
+                uploaded_info = gr.Markdown("_No files uploaded._")
+
+        def update_info(uploaded):
+            if not uploaded: return "_No files uploaded._"
+            names = [f.get("orig_name","?") if isinstance(f,dict) else getattr(f,"orig_name",Path(str(f)).name) for f in uploaded]
+            return "**Uploaded:**\n"+"\n".join(f"• `{n}`" for n in names)
+
+        def stream_response(message, history, current_files):
+            for h in respond(message, history, current_files):
+                yield h
+
+        msg.submit(stream_response, [msg, chatbot, files], [chatbot]).then(lambda:"", None, [msg])
+        send.click(stream_response, [msg, chatbot, files], [chatbot]).then(lambda:"", None, [msg])
+        clear_btn.click(lambda: [], None, chatbot, queue=False)
+        files.change(update_info, files, uploaded_info)
+    return demo
+
+if __name__ == "__main__":
+    demo = create_demo()
+    demo.queue(default_concurrency_limit=1, max_size=4)
+    demo.launch(css=""".file-upload-col{background:var(--background-fill-secondary);border-radius:12px;padding:16px}footer{display:none!important}""")
+                with gr.Row():
+                    msg = gr.Textbox(
+                        placeholder="Ask about your code, or just start chatting …",
+                        scale=8,
+                        show_label=False,
+                        container=False,
+                    )
+                    send = gr.Button("▶", scale=1, variant="primary", min_width=48)
+                with gr.Row():
+                    clear_btn = gr.Button("🗑 Clear chat", size="sm", scale=1)
+
+            # ---- RIGHT: File upload + info ----
+            with gr.Column(scale=1, elem_classes="file-upload-col"):
+                gr.Markdown("### 📎 Drop Files")
+                files = gr.File(
+                    file_count="multiple",
+                    label="Upload code, docs, data …",
+                    file_types=[
+                        ".py", ".ts", ".tsx", ".js", ".jsx", ".mjs",
+                        ".json", ".yaml", ".yml", ".toml",
+                        ".md", ".txt", ".csv",
+                        ".html", ".css", ".scss", ".xml", ".svg",
+                        ".c", ".cpp", ".h", ".hpp", ".java", ".kt",
+                        ".go", ".rs", ".rb", ".php", ".swift",
+                        ".sql", ".sh", ".bash", ".ps1",
+                        ".docx", ".pdf",
+                        ".env", ".gitignore", ".cfg", ".ini", ".conf",
+                        ".vue", ".svelte", ".dockerfile",
+                    ],
+                )
+                gr.Markdown(
+                    """
+                    **Supported:** `.py` `.ts` `.js` `.json` `.md` `.docx` `.pdf` `.csv`
+                    `.yaml` `.toml` `.html` `.css` `.sql` `.go` `.rs` `.java` + more
+
+                    Files are parsed and sent as context — the model reads them
+                    before answering.
+                    """
+                )
+                uploaded_info = gr.Markdown("_No files uploaded yet._")
+                file_content_state = gr.State("")
+
+        # ---- Event wiring ----
+
+        def update_file_info(uploaded):
+            if not uploaded:
+                return "_No files uploaded._", ""
+            names = []
+            for f in uploaded:
+                if isinstance(f, dict):
+                    names.append(f.get("orig_name", "?"))
+                else:
+                    names.append(getattr(f, "orig_name", Path(str(f)).name))
+            label = "**Uploaded:**\n" + "\n".join(f"• `{n}`" for n in names)
+            return label, parse_all_files(uploaded)
+
+        def respond(message, history, current_files):
+            """Generator wrapper that streams into the chatbot."""
+            for updated_history in chatbot_respond(message, history, current_files):
+                yield updated_history
+
+        # Trigger response on text submit or send button
+        msg.submit(
+            respond,
+            inputs=[msg, chatbot, files],
+            outputs=[chatbot],
+        ).then(lambda: "", None, [msg])
+
+        send.click(
+            respond,
+            inputs=[msg, chatbot, files],
+            outputs=[chatbot],
+        ).then(lambda: "", None, [msg])
+
+        # Clear chat — return empty list (Gradio 6 "messages" format)
+        clear_btn.click(lambda: [], None, chatbot, queue=False)
+
+        # File upload feedback
+        files.change(update_file_info, files, [uploaded_info, file_content_state])
+
+    return demo
+
+
+# ===========================================================================
+#  LAUNCH
+# ===========================================================================
 
 if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", server_port=7860)
+    demo = create_demo()
+    demo.queue(default_concurrency_limit=1, max_size=4)
+    demo.launch(
+        css="""
+        .file-upload-col { background: var(--background-fill-secondary); border-radius: 12px; padding: 16px; }
+        footer { display: none !important; }
+        """,
+    )