Update app.py

app.py

@@ -1,201 +1,93 @@
-# app.py
+# app.py ── SHASHA AI “Hybrid” (FastAPI + Gradio + Static UI)
 
-"""
-Main application file for SHASHA AI, a Gradio-based AI code generation tool.
-
-Provides a UI for generating code in many languages using various AI models.
-Supports text prompts, file uploads, website scraping, optional web search,
-and live previews of HTML output.
-"""
+from __future__ import annotations
+from typing import Any, Dict, List, Optional, Tuple
 
+import asyncio
 import gradio as gr
-from typing import Optional, Dict, List, Tuple, Any
+from fastapi import FastAPI
+from fastapi.staticfiles import StaticFiles
 
-# --- Local module imports ---
+# ────────────────────────────────────────────────────────────────
+# internal helpers (unchanged)
+# ────────────────────────────────────────────────────────────────
 from constants import (
     HTML_SYSTEM_PROMPT,
     TRANSFORMERS_JS_SYSTEM_PROMPT,
     AVAILABLE_MODELS,
-    DEMO_LIST,
-)
-from hf_client import get_inference_client
-from tavily_search import enhance_query_with_search
-from utils import (
-    extract_text_from_file,
-    extract_website_content,
-    apply_search_replace_changes,
-    history_to_messages,
-    history_to_chatbot_messages,
-    remove_code_block,
-    parse_transformers_js_output,
-    format_transformers_js_output,
 )
-from deploy import send_to_sandbox
+from inference import generation_code
 
-# --- Type aliases ---
 History = List[Tuple[str, str]]
-Model = Dict[str, Any]
-
-# --- Supported languages for dropdown ---
-SUPPORTED_LANGUAGES = [
-    "python", "c", "cpp", "markdown", "latex", "json", "html", "css",
-    "javascript", "jinja2", "typescript", "yaml", "dockerfile", "shell",
-    "r", "sql", "sql-msSQL", "sql-mySQL", "sql-mariaDB", "sql-sqlite",
-    "sql-cassandra", "sql-plSQL", "sql-hive", "sql-pgSQL", "sql-gql",
-    "sql-gpSQL", "sql-sparkSQL", "sql-esper"
-]
-
-def get_model_details(name: str) -> Optional[Model]:
-    for m in AVAILABLE_MODELS:
-        if m["name"] == name:
-            return m
-    return None
-
-def generation_code(
-    query: Optional[str],
-    file: Optional[str],
-    website_url: Optional[str],
-    current_model: Model,
-    enable_search: bool,
-    language: str,
-    history: Optional[History],
-) -> Tuple[str, History, str, List[Dict[str, str]]]:
-    query = query or ""
-    history = history or []
-    try:
-        # Choose system prompt based on language
-        if language == "html":
-            system_prompt = HTML_SYSTEM_PROMPT
-        elif language == "transformers.js":
-            system_prompt = TRANSFORMERS_JS_SYSTEM_PROMPT
-        else:
-            # Generic fallback prompt
-            system_prompt = (
-                f"You are an expert {language} developer. "
-                f"Write clean, idiomatic {language} code based on the user's request."
-            )
-
-        model_id = current_model["id"]
-        # Determine provider
-        if model_id.startswith("openai/") or model_id in {"gpt-4", "gpt-3.5-turbo"}:
-            provider = "openai"
-        elif model_id.startswith("gemini/") or model_id.startswith("google/"):
-            provider = "gemini"
-        elif model_id.startswith("fireworks-ai/"):
-            provider = "fireworks-ai"
-        else:
-            provider = "auto"
-
-        # Build message history
-        msgs = history_to_messages(history, system_prompt)
-        context = query
-        if file:
-            ftext = extract_text_from_file(file)
-            context += f"\n\n[Attached file]\n{ftext[:5000]}"
-        if website_url:
-            wtext = extract_website_content(website_url)
-            if not wtext.startswith("Error"):
-                context += f"\n\n[Website content]\n{wtext[:8000]}"
-        final_q = enhance_query_with_search(context, enable_search)
-        msgs.append({"role": "user", "content": final_q})
-
-        # Call the model
-        client = get_inference_client(model_id, provider)
-        resp = client.chat.completions.create(
-            model=model_id,
-            messages=msgs,
-            max_tokens=16000,
-            temperature=0.1
-        )
-        content = resp.choices[0].message.content
-
-    except Exception as e:
-        err = f"❌ **Error:**\n```\n{e}\n```"
-        history.append((query, err))
-        return "", history, "", history_to_chatbot_messages(history)
-
-    # Process model output
-    if language == "transformers.js":
-        files = parse_transformers_js_output(content)
-        code = format_transformers_js_output(files)
-        preview = send_to_sandbox(files.get("index.html", ""))
-    else:
-        cleaned = remove_code_block(content)
-        if history and history[-1][1] and not history[-1][1].startswith("❌"):
-            code = apply_search_replace_changes(history[-1][1], cleaned)
-        else:
-            code = cleaned
-        preview = send_to_sandbox(code) if language == "html" else ""
-
-    new_hist = history + [(query, code)]
-    chat = history_to_chatbot_messages(new_hist)
-    return code, new_hist, preview, chat
-
-# --- Custom CSS ---
-CUSTOM_CSS = """
-body { font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; }
-#main_title { text-align: center; font-size: 2.5rem; margin-top: 1.5rem; }
-#subtitle { text-align: center; color: #4a5568; margin-bottom: 2.5rem; }
-.gradio-container { background-color: #f7fafc; }
-#gen_btn { box-shadow: 0 4px 6px rgba(0,0,0,0.1); }
-"""
 
-with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue"), css=CUSTOM_CSS, title="Shasha AI") as demo:
-    history_state = gr.State([])
-    initial_model = AVAILABLE_MODELS[0]
-    model_state = gr.State(initial_model)
-
-    gr.Markdown("# 🚀 Shasha AI", elem_id="main_title")
-    gr.Markdown("Your AI partner for generating, modifying, and understanding code.", elem_id="subtitle")
-
-    with gr.Row():
-        with gr.Column(scale=1):
-            gr.Markdown("### 1. Select Model")
-            model_dd = gr.Dropdown(
-                choices=[m["name"] for m in AVAILABLE_MODELS],
-                value=initial_model["name"],
-                label="AI Model"
-            )
-
-            gr.Markdown("### 2. Provide Context")
-            with gr.Tabs():
-                with gr.Tab("📝 Prompt"):
-                    prompt_in = gr.Textbox(lines=7, placeholder="Describe your request...", show_label=False)
-                with gr.Tab("📄 File"):
-                    file_in = gr.File(type="filepath")
-                with gr.Tab("🌐 Website"):
-                    url_in = gr.Textbox(placeholder="https://example.com")
-
-            gr.Markdown("### 3. Configure Output")
-            lang_dd = gr.Dropdown(SUPPORTED_LANGUAGES, value="html", label="Target Language")
-            search_chk = gr.Checkbox(label="Enable Web Search")
-
-            with gr.Row():
-                clr_btn = gr.Button("Clear Session", variant="secondary")
-                gen_btn = gr.Button("Generate Code", variant="primary", elem_id="gen_btn")
-
-        with gr.Column(scale=2):
-            with gr.Tabs():
-                with gr.Tab("💻 Code"):
-                    code_out = gr.Code(language="html", interactive=True)
-                with gr.Tab("👁️ Live Preview"):
-                    preview_out = gr.HTML()
-                with gr.Tab("📜 History"):
-                    chat_out = gr.Chatbot(type="messages")
+# ────────────────────────────────────────────────────────────────
+# 1.  Blocks‑only “headless” API  (no UI, just /api/predict JSON)
+# ────────────────────────────────────────────────────────────────
+with gr.Blocks(css="body{display:none}") as api_demo:  # invisible
+    prompt_in     = gr.Textbox()
+    file_in       = gr.File()
+    url_in        = gr.Textbox()
+    model_state   = gr.State(AVAILABLE_MODELS[0])
+    search_chk    = gr.Checkbox()
+    lang_dd       = gr.Dropdown(choices=["html", "python"], value="html")
+    hist_state    = gr.State([])
+
+    code_out      = gr.Textbox()       # plain JSON
+    hist_out      = gr.State()
+    preview_out   = gr.Textbox()
+    chat_out      = gr.State()
+
+    api_demo.load(
+        generation_code,
+        inputs=[prompt_in, file_in, url_in, model_state,
+                search_chk, lang_dd, hist_state],
+        outputs=[code_out, hist_out, preview_out, chat_out],
+    )
 
-    model_dd.change(lambda n: get_model_details(n) or initial_model, inputs=[model_dd], outputs=[model_state])
+# ────────────────────────────────────────────────────────────────
+# 2.  Hybrid FastAPI server mounts:
+#      • /           → static/  (index.html, style.css, index.js …)
+#      • /api/*      → Gradio JSON (& websocket queue)
+# ────────────────────────────────────────────────────────────────
+app = FastAPI(title="SHASHA AI hybrid server")
 
-    gen_btn.click(
-        fn=generation_code,
-        inputs=[prompt_in, file_in, url_in, model_state, search_chk, lang_dd, history_state],
-        outputs=[code_out, history_state, preview_out, chat_out],
-    )
+# static assets
+app.mount(
+    "/", StaticFiles(directory="static", html=True), name="static"
+)
 
-    clr_btn.click(
-        lambda: ("", None, "", [], "", "", []),
-        outputs=[prompt_in, file_in, url_in, history_state, code_out, preview_out, chat_out],
-        queue=False,
-    )
+# gradio API
+app.mount(
+    "/api",
+    gr.mount_gradio_app(app, api_demo, path="/predict"),   # POST /api/predict
+    name="gradio-api",
+)
 
-if __name__ == "__main__":
-    demo.queue().launch()
+# ────────────────────────────────────────────────────────────────
+# 3.  Bonus: Web‑socket streamer for lightning‑fast preview
+# ────────────────────────────────────────────────────────────────
+@app.websocket("/api/stream")
+async def stream(websocket):
+    """
+    Front‑end connects, sends the same JSON as /predict,
+    and receives chunks (string tokens) as they arrive.
+    """
+    await websocket.accept()
+    payload = await websocket.receive_json()
+    queue: asyncio.Queue[str] = asyncio.Queue()
+
+    # spawn background generation
+    async def _run() -> None:
+        async for token in generation_code.stream(**payload):  # type: ignore
+            await queue.put(token)
+        await queue.put("__END__")
+
+    asyncio.create_task(_run())
+
+    while True:
+        item = await queue.get()
+        if item == "__END__":
+            break
+        await websocket.send_text(item)
+
+    await websocket.close()