𝗢𝗣𝗲𝗻𝗔𝗜 𝗜𝗻 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻: 𝗦𝘁𝗼𝗽 𝗪𝗮𝘀𝘁𝗶𝗻𝗴 𝗠𝗼𝗻𝗲𝘆

𝗢𝗣𝗲𝗻𝗔𝗜 𝗜𝗻 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻: 𝗦𝘁𝗼𝗽 𝗪𝗮𝘀𝘁𝗶𝗻𝗴 𝗠𝗼𝗻𝗲𝘆

I helped a startup add a chat feature. It took one day. Three days later, the bill was $340. They had 60 users.

Tutorials show you how to get a response. They skip the hard parts.

Production code needs guards. Without them, you pay too much.

Here is how to build it right:

• Set a timeout. Do not let requests hang.
• Count tokens before the API call. Reject messages too long for your budget.
• Use rate limits. Stop one user from hitting the API 30 times a minute.
• Cap output tokens. Stop the AI from writing books on your dime.
• Handle API errors. Give users clear messages when the service is busy.
• Use streaming for chat. It makes the app feel instant.

Testing is also different. Do not mock the API. Use cheap models like gpt-4o-mini for your tests. It costs almost nothing.

Build for scale. Build for cost.

Source: https://dev.to/harshdeepsingh13/how-to-integrate-the-openai-api-into-a-production-express-app-2mff

Optional learning community: https://t.me/GyaanSetuAi