OpenAI Agent Q&A API (Render-ready)

<<<<<<< codex/fix-fetch-request-error-qv346p A minimal Flask API that answers math questions using OpenAI Responses API with code interpreter.

A minimal Flask API that answers questions using an OpenAI agent-style workflow with tools enabled:

• Web search (web_search)
• Code interpreter (code_interpreter)

main

Endpoints

• GET / → healthcheck ({"status":"ok"})
• GET /hello?name=Alice → returns {"message":"Hello Alice"}
• POST /ask → asks the agent a question

Example request:

curl -X POST http://localhost:5000/ask \
  -H "Content-Type: application/json" \
  -d '{"question": "Find the latest Python release and compare 2^10 vs 10^2"}'

Example response:

{
  "answer": "...",
<<<<<<< codex/fix-fetch-request-error-qv346p
  "model": "gpt-4.1"
=======
  "model": "gpt-5-mini"
>>>>>>> main
}

If question is missing/empty, API returns HTTP 400.

Environment variables

• OPENAI_API_KEY (required) — your OpenAI API key. <<<<<<< codex/fix-fetch-request-error-qv346p
• OPENAI_MODEL (optional) — defaults to gpt-4.1.
• Uses OpenAI Responses API with code_interpreter configured with container: {"type": "auto", "memory_limit": "4g"} (requires modern openai SDK). =======
• OPENAI_MODEL (optional) — defaults to gpt-5-mini.
• Uses OpenAI Responses API with web_search and code_interpreter tools (requires modern openai SDK).

main

Local development

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
export OPENAI_API_KEY="your_api_key_here"
# optional:
<<<<<<< codex/fix-fetch-request-error-qv346p
# export OPENAI_MODEL="gpt-4.1"
=======
# export OPENAI_MODEL="gpt-5-mini"
>>>>>>> main
python app.py

Deploy on Render

This repository includes render.yaml, so the easiest option is Render Blueprint deploy:

1. Push this repo to GitHub.
2. In Render, click New + → Blueprint.
3. Select your repository.
4. Add OPENAI_API_KEY in the Render service environment variables.
5. Deploy.