The purpose of this repo is to help you to enhance your prompting skills
A prompt is the input and/or instruction given to a AI Model in order to guide Its response.
Refining your prompts will enhance your deliverables and productivity by leveraging the AI. The better you guide the AI Model's responses, the more your work will be boosted.
- Shape the behaviour of the AI Model.
- Format the output / response from the AI Model.
- Get accurate responses.
- Minimize incorrect responses.
- https://huggingface.co/playground
- https://aistudio.google.com/prompts/new_chat
- https://chat.mistral.ai/chat
- https://chatgpt.com/
- https://chat.deepseek.com/a/chat
- https://gemini.google.com/app
- Define your objective clearly: Be specific about the purpose and desired output.
- Provide context: Include relevant information or background.
- Use clear and simple language: Look for the right wording, don't be ambiguous.
- Specify the format or style.
- Ask direct questions or give explicit instructions.
- Iterate and Refine: Test your prompt and modify It in order to achieve your goal.
This technique do not provide context either examples in the query or prompt. As a result, the AI Model will entirely rely on the pre-trained data.
Recommended for:
- Simple tasks that do not require dinamic information.
- Experimentation with LLM.
- General knowledge tasks.
In this technique you give to the AI Model a single example of how you want the response, this guides the AI Model on how the response should be.
Recommended for:
- Limited training data.
- Easy tasks that require output formatting.
- Not complex enough to require several examples.
You provide in the prompt a few examples of the task you want the model to execute. This technique helps the AI Model to response as expected.
Recommended for:
- Task definition.
- Tasks for a niche that the AI Model was not train on.
- When the AI Model needs to adapt quickly having no datataset to be trained.
The CoT technique involves in sending in the prompt a series of steps to follow by the AI Model, in such a way that the response is the result of following a procedure, making easier to resolve complex tasks and reducing the risk of incorrect assumptions.
Recommended for:
- Complex problem solving
- When a problem requires several steps to achieve a solution
- Reduce errores in ambiguos or unclear contexts
- Teach the model a procedure or process to deal with specific areas
Using this technique you send direct and clear requests to the model, for example: resume a text in 200 words, make a list of 10 points, etc.
Recommended for:
- Performing specific tasks
- Chatbots that response to users in specific way
- Generating content from text or input
In this technique, you assign a role or persona for the model to assume, so the model has context and boundaries in order to elaborate answers to the inquires.
Recommended for:
- Education content, for example assuming the role of teacher.
- Customer service and support.
- Professional in a specific area.