app.py

# app.py
"""
ShashaCode Builder – AI code‑generation playground.

• Hugging Face Spaces + Gradio front‑end
• Supports prompts, file upload, web‑site scraping, optional web search
• Streams code back, shows live HTML preview, can deploy to a user Space
"""

# ─────────────────────────────────────────  Imports
import gradio as gr
from pathlib import Path
from typing import Dict, List, Optional, Tuple, Any

from constants import (                               # ← all constants live here
    HTML_SYSTEM_PROMPT,
    TRANSFORMERS_JS_SYSTEM_PROMPT,
    SYSTEM_PROMPTS,
    AVAILABLE_MODELS,
    DEMO_LIST,
    GRADIO_SUPPORTED_LANGUAGES,                      # ← new import
    SEARCH_START, DIVIDER, REPLACE_END,
)

from hf_client      import get_inference_client
from tavily_search  import enhance_query_with_search
from utils          import (                          # helpers split into utils.py
    history_to_messages,
    history_to_chatbot_messages,
    remove_code_block,
    parse_transformers_js_output,
    format_transformers_js_output,
    parse_svelte_output,
    format_svelte_output,
    apply_search_replace_changes,
    apply_transformers_js_search_replace_changes,
    extract_text_from_file,
    extract_website_content,
    get_gradio_language,
)
from deploy         import send_to_sandbox

# ─────────────────────────────────────────  Type Aliases
History = List[Tuple[str, str]]
ModelInfo = Dict[str, Any]

# ─────────────────────────────────────────  Core Function
def generate_code(
    query: str,
    file_path: Optional[str],
    website_url: Optional[str],
    model: ModelInfo,
    enable_search: bool,
    language: str,
    history: Optional[History],
) -> Tuple[str, History, str, List[Dict[str, str]]]:
    """Main inference pipeline: build prompt → call model → post‑process."""
    query = query or ""
    history = history or []

    # 1. pick system prompt
    if   language == "html":             system = HTML_SYSTEM_PROMPT
    elif language == "transformers.js":  system = TRANSFORMERS_JS_SYSTEM_PROMPT
    else:                                system = SYSTEM_PROMPTS.get(language, HTML_SYSTEM_PROMPT)

    # 2. build message list
    messages = history_to_messages(history, system)

    ctx_parts = [query.strip()]

    if file_path:    ctx_parts += ["[File]",    extract_text_from_file(file_path)[:5000]]
    if website_url:
        html = extract_website_content(website_url)
        if not html.startswith("Error"):
            ctx_parts += ["[Website]", html[:8000]]

    user_query = "\n\n".join(ctx_parts)
    user_query = enhance_query_with_search(user_query, enable_search)
    messages.append({"role": "user", "content": user_query})

    # 3. call model
    client = get_inference_client(model["id"])
    resp   = client.chat.completions.create(
        model=model["id"],
        messages=messages,
        max_tokens=16000,
        temperature=0.15,
    )
    answer = resp.choices[0].message.content

    # 4. post‑process
    if language == "transformers.js":
        files = parse_transformers_js_output(answer)
        code  = format_transformers_js_output(files)
        preview = send_to_sandbox(files.get("index.html", ""))
    else:
        clean  = remove_code_block(answer)
        if history and not history[-1][1].startswith("❌"):
            clean = apply_search_replace_changes(history[-1][1], clean)
        code    = clean
        preview = send_to_sandbox(code) if language == "html" else ""

    history.append((query, code))
    chat_msgs = history_to_chatbot_messages(history)
    return code, history, preview, chat_msgs


# ─────────────────────────────────────────  UI
LOGO_PATH = "assets/logo.png"                       # ensure this file exists

CUSTOM_CSS = """
body {font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;}
#logo {max-height:64px;margin:auto;}
"""

with gr.Blocks(css=CUSTOM_CSS, title="ShashaCode Builder") as demo:
    state_history = gr.State([])
    state_model   = gr.State(AVAILABLE_MODELS[0])

    # Header
    with gr.Row():
        gr.Image(LOGO_PATH, elem_id="logo", show_label=False, height=64)
        gr.Markdown("## **AnyCoder AI**\nYour AI partner for generating, modifying & understanding code.")

    # Sidebar (inputs)
    with gr.Row():
        with gr.Column(scale=1, min_width=300):
            # Model
            dd_model = gr.Dropdown(
                label="AI Model",
                choices=[m["name"] for m in AVAILABLE_MODELS],
                value=AVAILABLE_MODELS[0]["name"],
            )

            # Prompt / File / Website tabs
            with gr.Tabs():
                with gr.Tab("Prompt"):
                    tb_prompt = gr.Textbox(label="Describe what you'd like to build…", lines=6)
                with gr.Tab("File"):
                    inp_file  = gr.File(label="Reference file", type="filepath")
                with gr.Tab("Website"):
                    tb_url    = gr.Textbox(label="URL to redesign")

            # Output config
            dd_lang = gr.Dropdown(
                label="Target language",
                choices=[l for l in GRADIO_SUPPORTED_LANGUAGES if l],  # ← fixed list
                value="html",
            )
            chk_search = gr.Checkbox(label="Enable Tavily Web Search")

            # Buttons
            btn_generate = gr.Button("Generate Code", variant="primary")
            btn_clear    = gr.Button("Clear Session",  variant="secondary")

        # Main panel (outputs)
        with gr.Column(scale=2):
            with gr.Tabs():
                with gr.Tab("Code"):
                    out_code = gr.Code(language="html", show_label=False)
                with gr.Tab("Preview"):
                    out_prev = gr.HTML()
                with gr.Tab("History"):
                    out_hist = gr.Chatbot(type="messages")

    # ─── Callbacks ─────────────────────────────────────────────
    def _model_from_name(name):
        return next((m for m in AVAILABLE_MODELS if m["name"] == name), AVAILABLE_MODELS[0])

    dd_model.change(lambda n: _model_from_name(n), inputs=dd_model, outputs=state_model)

    btn_generate.click(
        fn=generate_code,
        inputs=[tb_prompt, inp_file, tb_url, state_model, chk_search, dd_lang, state_history],
        outputs=[out_code, state_history, out_prev, out_hist],
    )

    btn_clear.click(lambda: ("", None, "", [], "", []),
                    outputs=[tb_prompt, inp_file, tb_url, state_history, out_code, out_prev])

if __name__ == "__main__":
    demo.queue().launch()