Manually finding the deadlines of conferences of interest is always a hassle, especially since there is no centralized channel to keep all conference deadlines updated. This project leverages agentic AIs to retrieve all conference deadlines for us, and then renders a clean HTML table for browsing and sorting.

General workflow of this project:

1. Specify your conferences in conference_list.csv.
2. Set up your environment variables, including your OpenAI API key.
3. Run the provided commands to retrieve deadlines and official conference webpages.
4. Run another command to generate a nice HTML file for visualization.

An example generated HTML file is placed in the root directory, using our conference_list.csv.

Note: Running one round of conference information retrieval costs around 3–5 Australian dollars if using the GPT-5 API. We found that using other, less advanced agentic models may lead to inaccurate or failed information retrieval.

We recommend the uv workflow.

1. Install uv (macOS/Linux)

curl -LsSf https://astral.sh/uv/install.sh | sh

2. Create and sync the project environment

cd /Users/xu/system-conference
uv python install 3.11
uv sync

3. Activate (optional)

uv venv activate

4. Configure OpenAI

export OPENAI_API_KEY=sk-...
# or place it in .env at the repo root

Use a single input file: conference_list.csv (UTF-8).

Required columns:

• Acronym
• Title (the official or common conference title)

Optional columns:

• core_year (used as a hint for the target edition year; otherwise current year is used)

Example:

Acronym,Title
UCC,Utility and Cloud Computing
ICDCS,International Conference on Distributed Computing Systems

Runs the agent to resolve each conference to an official homepage and the most relevant submission deadline. We found gpt-5 is needed to retrieve more accurate conference information.

uv run deadlines collect \
  --input-csv /Users/xu/system-conference/conference_list.csv \
  --output-csv /Users/xu/system-conference/output/metadata.csv \
  --model gpt-5.1

Flags you may care about:

• --workers 20: increase concurrency of retreving conference information.
• --verbose: print prompts and raw model output for debugging

Output columns(simplified):

• name, acronym, year
• homepage_url
• deadline (final chosen date: submission preferred, else abstract)
• notification_date (if present)
• location (if present)

Turns the collected CSV into a responsive HTML table with sorting buttons and theme toggle. The page shows a single Date column and hides notification_date to keep the layout clean.

uv run deadlines render-html \
  --input-csv /Users/xu/system-conference/output/metadata.csv \
  --output-html /Users/xu/system-conference/output/metadata.html \
  --title "Conference Deadlines"
open /Users/xu/system-conference/output/metadata.html

Features:

• Sort by name or by date (upcoming first; past entries follow).
• Dark/Light theme with persistent preference.
• GitHub link in the footer: System-Conference-Deadlines.

• If you hit rate limits, re-run with fewer --workers or simply re-run later; the agent backs off automatically.
• For maximum accuracy on tricky websites, enable --verbose and check the prompts/output to refine the watchlist terms.

MIT