from __future__ import annotations
import os
import gradio as gr
import health
from layout import CELL_CSS, cell
from problem_cell import render_problem_cell
from solution_cell import render_solution_cell
from setup_cell import render_setup_cell
from context_biased_transcription_cell import render_context_biased_transcription_cell
from media_analysis_cell import render_media_analysis_cell
from translation_cell import render_translation_cell
from wrap_up_cell import render_wrap_up_cell
def render_health_panel(gemini_api_key: str | None = None) -> str:
return health.render_health_notice(gemini_api_key)
def create_app() -> gr.Blocks:
"""Create the Gradio Blocks application used in the Hugging Face Space.
The layout is intentionally notebook-like: each conceptual unit
(problem, health, demos, wrap-up) is encapsulated in its own module
and rendered as a "cell" to keep the main app glue straightforward.
"""
with gr.Blocks(title="Aileen3 Demo") as demo:
gr.HTML(f"")
with cell("๐ Introduction"):
gr.Markdown(
"""
# Aileen 3 Core
Large Language Models (LLMs) rely on tools โ sometimes provided by MCP servers โ to interact with the outside world.
Aileen 3 Core is an MCP server focused on **Information Foraging**: mining high-noise sources for novel insights and turning
them into dense briefings you can consume quickly. Grounded in cognitive science, Aileen 3 models novelty as prediction error against
explicit priors such as user expectations, facts from an AI Memory Bank, or media context. To that end, the Aileen 3 Core
MCP server provides media access and analysis services backed by Google Gemini.
This Space shows a small, notebook-style slice of that system so you can verify the core MCP server is reachable and working.
Today, Aileen 3 Core is a contender in the **[MCP's 1st Birthday โ Hosted by Anthropic and Gradio](https://huggingface.co/MCP-1st-Birthday)**
hackathon. It also powers the **[Aileen 3 Agent](https://github.com/ndurner/aileen3-agent)** project, a
[capstone project](https://ndurner.de/links/aileen3-kaggle-writeup) for the
[*AI Agents Intensive Course with Google*](https://www.kaggle.com/learn-guide/5-day-agents). Its predecessor, Aileen 2, explored autonomous
personalized summarization of German parliament proceedings and won Honorable Mention in the
[Generative AI Agents Developer Contest by NVIDIA and LangChain](https://www.nvidia.com/en-in/ai-data-science/generative-ai/developer-contest-with-langchain/).
For more details on setup, usage, and background, see the full project README. Aileen 3 Core has been tested with Claude Desktop and Aileen 3 Agent.
"""
)
with cell("๐งญ How this demo is organized"):
gr.Markdown(
"""
Think of this interface as a lightweight Jupyter notebook: instead of code cells, you get a sequence of interactive โcellsโ that walk through key aspects of the Aileen 3 Core MCP server.
- The **Introduction** cell gives you the high-level context from the README.
- The **Health check** cell below connects to the MCP server used by the Space and shows whether it can be started successfully.
- Future cells will demonstrate the problem statement and core information-foraging capabilities.
"""
)
render_problem_cell()
render_solution_cell()
gemini_key_box = render_setup_cell()
with cell("๐ฉ๐ปโโ๏ธ Health check"):
health_panel = gr.HTML(value=health.render_placeholder_notice())
run_button = gr.Button("Run health check", variant="primary")
run_button.click(
fn=render_health_panel,
inputs=gemini_key_box,
outputs=health_panel,
queue=False,
)
render_context_biased_transcription_cell(gemini_key_box)
render_media_analysis_cell(gemini_key_box)
render_translation_cell(gemini_key_box)
render_wrap_up_cell()
return demo
def main() -> None:
app = create_app()
server_name = os.environ.get("GRADIO_SERVER_NAME") or os.environ.get("HOST") or "0.0.0.0"
port_value = os.environ.get("GRADIO_SERVER_PORT") or os.environ.get("PORT") or "7860"
try:
server_port = int(port_value)
except ValueError:
server_port = 7860
app.launch(
server_name=server_name,
server_port=server_port,
)
if __name__ == "__main__":
main()
