This project demonstrates a creative‑writing agent built with LangGraph and Google Gemini. The underlying graph strings together several LLM powered nodes so a story can evolve from an initial idea to a polished draft.

1. Query rewriter – interprets the user prompt and expands it into a detailed description.
2. Story ideas generator – proposes multiple plot options.
3. Human feedback handler – lets you pick an idea or provide guidance, looping back for more ideas if needed.
4. Character & scene generation – creates casts and outlines while a supervisor node rates their quality.
5. Draft writer – produces a first draft from the approved outline.
6. Macro & micro editors – give structural and line‑level feedback.
7. Revision orchestrator – decides whether to loop back for another draft or finish the story. ======= The graph coordinates multiple LLM driven nodes to iterate from a user query all the way to a revised story draft.

• ReAct style reasoning powers decision making between nodes.
• Human‑in‑the‑loop feedback influences story ideas and outlines.
• A Supervisor node evaluates outlines before drafting.

The graph is rendered below and can also be explored in the writer.ipynb notebook.

1. Install Python 3.10+.
2. Install dependencies:

   pip install -e .

3. Set environment variables required by langchain-google-genai (e.g. GOOGLE_API_KEY). You can use a .env file which is loaded automatically.

Run the graph to compile and visualize it:

python graph.py

This will build the StateGraph and display the workflow image. From here you can experiment by calling individual nodes or adapting the graph for your own projects.