hg_chat_with_model

A small educational Python project for asynchronous chat inference using Hugging Face models.

The project reads user queries from a CSV file, sends them concurrently to a chat model, and writes the results to a new CSV file that fully reproduces the input data with an additional model response column.

---

How It Works

1. Read user queries from a CSV file.
2. Convert each query into a single-turn chat message.
3. Send all chats concurrently to a Hugging Face model.
4. Save a copy of the input CSV with an added model response column.

---

Configuration

The application is configured via environment variables (.env file):

hg_token=YOUR_HUGGINGFACE_TOKEN
model=HuggingFaceModelName
busket_file_name=input.csv
result_file_name=output.csv

---

Data Format

Input CSV (located in data/) must contain at least:

id,query

The output CSV:

• contains all original columns
• adds an additional model_answer column

---

Project Structure

.
├── src/
│   └── hg_chat_with_model/      
        ├── __main__.py          # Entry point and pipeline orchestration
        ├── hg_api_calls.py      # Asynchronous Hugging Face chat requests
        └── data_processing.py   # CSV input and output processing
├── tests/                       
├── .gitignore
├── poetry.lock
├── pyproject.toml               # Depenedcies
└── README.md

---

Run

poetry run python -m src.hg_chat_with_model

---

Notes

• Each query is sent as a single user message (no conversation history).