|
|
|
|
|
""" |
|
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. |
|
""" |
|
|
|
import gradio as gr |
|
from typing import Optional, Dict, List, Tuple, Any |
|
|
|
|
|
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 |
|
|
|
|
|
History = List[Tuple[str, str]] |
|
Model = Dict[str, Any] |
|
|
|
|
|
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: |
|
|
|
if language == "html": |
|
system_prompt = HTML_SYSTEM_PROMPT |
|
elif language == "transformers.js": |
|
system_prompt = TRANSFORMERS_JS_SYSTEM_PROMPT |
|
else: |
|
|
|
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"] |
|
|
|
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" |
|
|
|
|
|
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}) |
|
|
|
|
|
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"β **Error:**\n```\n{e}\n```" |
|
history.append((query, err)) |
|
return "", history, "", history_to_chatbot_messages(history) |
|
|
|
|
|
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("β"): |
|
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 = """ |
|
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("# π 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("π Prompt"): |
|
prompt_in = gr.Textbox(lines=7, placeholder="Describe your request...", show_label=False) |
|
with gr.Tab("π File"): |
|
file_in = gr.File(type="filepath") |
|
with gr.Tab("π 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("π» Code"): |
|
code_out = gr.Code(language="html", interactive=True) |
|
with gr.Tab("ποΈ Live Preview"): |
|
preview_out = gr.HTML() |
|
with gr.Tab("π History"): |
|
chat_out = gr.Chatbot(type="messages") |
|
|
|
model_dd.change(lambda n: get_model_details(n) or initial_model, inputs=[model_dd], outputs=[model_state]) |
|
|
|
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], |
|
) |
|
|
|
clr_btn.click( |
|
lambda: ("", None, "", [], "", "", []), |
|
outputs=[prompt_in, file_in, url_in, history_state, code_out, preview_out, chat_out], |
|
queue=False, |
|
) |
|
|
|
if __name__ == "__main__": |
|
demo.queue().launch() |
|
|