Dutique is a lighthearted, AI-powered task creator focused on home organization and cleanliness. Whether you're tidying up the kitchen, refreshing your living space, or getting the whole family involved, Dutique conjures up practical and motivating household tasks to keep things running smoothly. Designed to be simple, cheerful, and helpful — it's your go-to tool for turning chores into small wins and tidy moments. 🧹🪴
flowchart LR
id1["**Function Step:**
Task Generation"] -->
id2["**LLM Step:**
Description Generation"] -->
id3{"**LLM Step:**
Quality Check"}
id3 -->|"Fail & attempts < 5"| id4["**LLM Step:**
Repair Process"]
id4 --> id3
id3 -->|"Fail & attempts ≥ 5"| id5["Failed"]
id3 -->|"Pass"| id6["Successful"]
id6 --> id7["**Storage:**
ChromaDB"]
%% Styling
classDef functionStep fill:#d4f1f9,stroke:#05a3d6,stroke-width:2px,color:black
classDef llmStep fill:#ffe6cc,stroke:#ff9933,stroke-width:2px,color:black
classDef decision fill:#fff2cc,stroke:#ffcc00,stroke-width:2px,shape:diamond,color:black
classDef success fill:#d5e8d4,stroke:#82b366,stroke-width:2px,color:black
classDef failure fill:#f8cecc,stroke:#b85450,stroke-width:2px,color:black
classDef storage fill:#e1d5e7,stroke:#9673a6,stroke-width:2px,color:black
class id1 functionStep
class id2,id4 llmStep
class id3 decision
class id6 success
class id5 failure
class id7 storage
- Task Generation:
- Initial task creation through code functions (difficulty level, time required, category)
- Description Generation:
- LLM creates engaging descriptions based on generated task parameters and RAG database
- Quality Check:
- LLM checks the task quality criteria
- Repair Process:
- Tasks failing quality checks undergo repair (up to 5 attempts) by an LLM that tries to improve the task
- Storage:
- Successfully verified tasks are stored for retrieval in ChromaDB
To get Dutique up and running, please install the following software:
- Ollama
- Ensure that Ollama is properly configured and reachable from outside. For more information on configuring Ollama, refer to the Ollama FAQ.
- Docker & Docker Compose
Then run:
# Download LLM Modell
ollama pull llama3.2:latest
# Clone the repo
git clone https://github.com/wmneco/Dutique
cd Dutique
# Start up the project
docker compose up