Cod3r

An agentic engineering assistant that plans, architects, and scaffolds projects from a natural-language prompt. It uses LangGraph with a Planner → Architect → Coder flow and writes output files under a dynamically named project directory: gen/<slug-of-plan-or-prompt>.

Features

• Planner agent converts your prompt into a structured Plan (name, description, techstack, features, files).
• Architect agent produces an ordered TaskPlan of implementation steps.
• Coder agent executes steps, reading/writing files using constrained tools.
• Streamlit UI for entering prompts and previewing generated files.
• Dynamic project directory naming based on the plan name or the prompt.

Requirements

• Python 3.10+
• A Groq API key with access to the configured model

Installation

cd /root/Projects/Agents/Coder
python -m pip install -e .

Environment

Create a .env file in the project root:

GROQ_API_KEY=your_groq_api_key

You can also provide GROQ_API_KEY at runtime via environment variable or in the Streamlit sidebar.

Usage

Run via CLI

streamlit run main.py

• You will be prompted for your project description.
• The agent will plan, architect, and implement files.
• Output is written under gen/<slug> where <slug> is derived from the plan name, falling back to your prompt.

Run the Streamlit UI

python -m streamlit run main.py --server.headless=true --server.port=8501
# If you kept a separate UI script, use:
# python -m streamlit run streamlit_app.py --server.headless=true --server.port=8501

• Open http://localhost:8501 in your browser.
• Enter a prompt, run the agent, and view the generated project tree.

Project Output Directory

• The project root is set dynamically to gen/<slug> where <slug> is a URL-friendly version of the Plan name.
• If the Plan name is missing, the slug is derived from your original prompt.
• All tool-based file operations (read/write/list) are constrained to this directory.

Architecture

• agent/states.py: Pydantic models (Plan, TaskPlan, ImplementationTask, CoderState).
• agent/prompts.py: Orchestrator prompts for Planner/Architect/Coder.
• agent/tools.py: Tooling utilities and I/O tools exposed to the agent:
  • write_file(path, content)
  • read_file(path)
  • list_files(directory)
  • get_current_directory()
  • run_cmd(cmd, cwd, timeout)
  • set_project_root_from_name(name) and init_project_root()
• agent/graph.py: LangGraph setup with nodes:
  • Planner → Architect → Coder; loops coder until done
  • Sets project directory based on the Plan name (fallback prompt)

Workflow

Configuration

• Model is configured in agent/graph.py via ChatGroq:
  • ChatGroq(model="openai/gpt-oss-120b", api_key=os.getenv("GROQ_API_KEY"))
• Adjust verbosity with LangChain globals if needed.

Development

• Linting is handled by your editor; ensure no syntax errors.
• The tools enforce a safe project root; writing outside gen/<slug> will error.

Notes

• Streamlit should be launched with streamlit run ... (not python ...).
• If running in WSL2, access the UI at http://localhost:8501 from Windows.

Sample

## License MIT (or your preferred license)