ChatGPT API

- Posted in Uncategorized by

I signed up for some kind of ChatGPT API service. OpenAI offers API code in a number of languages like Go, Python, Javascript... not C. I was looking at invoking Python scripts from C. It is possible through a library, or by fork + exec a shell script. But, populating some JSON templates and contacting ChatGPT through libcurl looks like the easiest thing to do.

Here is an example from the commandline:

curl -v https://api.openai.com/v1/chat/completions -H "Content-Type: application/json" \
-H "Authorization: Bearer <arcane authorization here>
-d '{
  "model": "gpt-5-nano",
  "messages": [{"role": "user", "content": "In a few words, answer this question: does mario love the princess"}]
}'

The trouble is, the answer comes back: "Yes, he loves Princess Peach." (!)

Oh? Do we know that?

Now to embedding it in Brainhat. libcurl is available from https://curl.se/libcurl/ with little license restriction. There are non-blocking calls, which means I can ask ChatGPT a question and go away. It takes a long time to answer, typically.

What Next?

- Posted in Uncategorized by

Fixed a bug from May 4th. Was not recording multiple conditions in the result of an inference.

What next? GUI pages for vocab, communications, memories or scripts? Glue into ChatGPT? Integrate with MQTT? Try to create an application? Maybe Brainhat agents that follow network traffic looking for a reason?