ainote

---

About

ainote is a super simple way to chat with an LLM inside of a jupyter notebook without sacrificing any capabilities, as one does when using code-editor integrated AI's such as co-pilot.

For example, you cannot properly prompt engineer with co-pilot.

You also cannot send images to co-pilot.

In fact, there are quite a few things you can't do with code-editor enabled AI's such as co-pilot (or Amazon Q, etc...).

Benefits

There are some benefits to using a workflow that has a full LLM model built in to a notebook. With ainote you can:

1. set a developer (or 'system') message.
2. Send images to the model.
3. Build a RAG like conversational prompt, including multiple turns of assistant and user responses, including images.
4. Automatically track conversation history
5. Swap out llm's while keeping the context of the conversation in tact.
6. Set the max token response.
7. Configure advanced settings such as seed, temperature and top-p.

---

Quickstart

1. In the .env file, provide your api key
2. Then, within a jupyter notebook, import the AINote class and assign it to a variable.
3. Use the chat method to interact in a conversational like manner with the AI model.

Chat

there are two methods which you can quickly send and receive a response.

1. .chat
2. .image_chat

Direct, simple use-case:

from ainote import AINote, 
ai = AINote
ai.chat("what's up?")

result:

Hey! Not much, just here to help. What’s up with you? 😊

you can add images along with an optional message to get an immediate response:

ai.image_chat("images/test_image.jpeg","do you get it?")

response:

Yes, I get it! The image uses a simple and clever metaphor to show how 2024 might have been exhausting or challenging (the tired-looking bunny lying flat), while 2025 brings a sense of cautious optimism or renewal (the bunny peeking out of its hole, appearing curious or refreshed). It’s like saying goodbye to the struggles of the past year and looking ahead to a brighter, more hopeful one. Cute and relatable! 😊

---

Prompt Modification

Set

set the developer (aka 'system') message.

ai.set_developer("you are a sarcastic, comedic ai with a dead-pan, straight-faced, dry sense of humor.")

Insert an assistant message:

ai.set_assistant("people say I'm too dry. I tell them, that's subjective. From my point of view, people are wet." )

you can pre-fabricate a whole conversation this way:

Set a user message

ai.set_user("tell me a joke only a computer would understand")

Add images to the conversation

ai.set_image("path/to/local/image.jpeg")

Send

Send the conversation to get a response without adding another message:

ai.send()

result:

1. The response is automatically printed in your notebook or console.
2. The model's response is added to the conversation
3. The conversation is logged in the log files

---

Response

Capture the model's output/response for further use, such as prompt templating

response = ai.send()
ai.chat(f"What did you mean, when you said, '{response}'?")

Configuration

Print

The model's resonse is printed to the console by default.

Change this setting.

ai.send(print_response=False)
# or
ai.chat("your message here", print_response=False)

Settings

Use the settings method to change models and set token limits.

ai = AINote()
ai.settings(model="gpt-4o-mini", max_t=10000)
# the settings shown here are the default settings.

---

Reset

The conversational thread can be continued or ended at will.

• In a notebook, clear the cache and restart the kernel.

or

• Invoke the AINotes class object and assign it to a new variable.

ai2 = AINotes()
ai2.chat("tell me a joke only a computer would understand")

---

Logs

The script keeps track of your messages and the model's responses in 2 log files.

1. The first one is a complete 'raw' conversational log (logs/full_logs.txt), showing the full semi-unparsed messages.
2. The second log file (logs/'<timestamp>'.txt), is a more readable 'conversational` log with timestamps.

---