Welcome to OctiAI v4!
OctiAI v4 is a state-of-the-art prompt engineering application capable of generating, editing, and enhancing AI prompts for various models using an assortment of powerful tools.Overview:
Generation:
1. Select the AI tool you want to create a prompt for. 2. Enter your task or a description of what you need done. It can be as detailed or brief as you feel necessary. 3. If your prompt requires custom information, provide it within the "Contextual data" box (see more). 4. Press "Enter" or click the arrow to generate your improved prompt. - Try using "Voice Mode" to share ideas quickly!Iteration:
1. Once you have a prompt, you can select a section of text to modify. 2. Type any improvements you want to make into the main task bar. 3. Press "Enter" or click the arrow for OctiAI to regenerate the selected section based on your instructions. 4. Finally, choose one of the generated options that best fits your needs and press "Enter" to apply it to the prompt.Switch Models & Customize Your Experience:
1. Click on the tool button near the top left corner. 2. A menu will open with a list of dozens of available AI tools. 2. If the tool you need doesn’t exist, open the tools menu and select "+ Create A New Model" (Pro Users Only). 3. Enter the name and description of your tool, then Fill in the meta-prompt details of your desired AI tool. 4. Press the save button to ensure your tool is saved to your account. PS. You can also copy/edit an existing model by clicking the small gear icon, this allows you to work with preset templates.Miscellaneous:
- Right-click on a prompt to open the [Actions] menu, which provides access to a set of useful commands. - Click the (Advanced v4) button to toggle between V3 & V4 functionality. (V3 is a legacy model.) - Control+Z/Control+Y to undo or redo any changes made to the prompt; History Is preserved throughout sessions. - The limit displayed in the top right corner accounts for the total actions available to you each month. - Click the "Export" button to copy your prompt to the clipboard—this ensures all relevant data is copied and formatted correctly. ---Please note: V4 is still in beta.
We hope you find this guide useful and informative! Contact support at: projectloomie@gmail.com OctiAI Team <3Prompting Inputs Guide
What to write in the chat bar?
Here's where you can outline your project goals and immediate task instructions!Using Voice:
This is one of our users' favorite features! - Simply activate the microphone button and talk to OctiAI as if you are speaking to a colleague or friend. - Deliver a monologue! You can pause and resume, taking as many breaks as you'd like. - This makes the information and context delivery process feel more natural, swift, and empowering. - Speak for as long as you'd like until you feel you have covered the task. - Once ready, click the enhance button to generate your prompt! Remember: - Your voice input will be treated as "chat bar input" by default. - If you want to provide context via voice, simply select/focus the "contextual details" container and start speaking! ---Using Text:
The process is very simple. Don't worry about spelling or grammar—we'll handle that for you! OctiAI v4 is more than capable of understanding even the most atrociously written text. Provide OctiAI with: - A description of the task you want to accomplish. - Any relevant ideas you have. - Your raw, unfiltered thoughts. - As many details as possible. Don'ts: - When generating a new prompt, avoid directly saying "improve my prompt" or "I want a prompt that..." - This disrupts the AI system, as it is already trained to interpret all of your input text as the prompt itself needing improvement. ---During Iteration
When editing a prompt, you can either select a section of text or the entire prompt. OctiAI v4 will then treat the selected region as a canvas, and anything you input into the chat bar will be used to modify the selected region. During iteration, OctiAI's algorithm is very section-specific, ensuring it only regenerates the text in the immediate section. You should focus only on the changes you want made to the selected region. If you want to make changes to the entire prompt, press escape to remove the highlight and press enter to generate a new iteration of the prompt."Provide Contextual Data" Input
When your task requirements rely on knowledge that is not accessible via the internet, it is essential to provide your AI model with this information directly. This documentation explains how to supply and label contextual data for optimal performance. Such context helps the AI understand the nuances of the task, ensuring that its response is both accurate and tailored to your needs. When providing contextual data, consider including the following:- Large text examples.
- Code snippets & developer logs.
- Unknown background knowledge.
- Secondary priority & bulk data, large text, etc.
- Big-picture project goals.
Contextual data should not include:
- Direct instructions.
- Information such as tone, style, format, and audience.
- Directions on how to complete the task.
---
Below are several examples and use cases where contextual data is beneficial:
Programming Tasks:
Provide details such as the function name or code block where the new code will be integrated, dependencies, or related modules.
Creative Writing:
Supply the AI with an example of your writing style or a text with a similar tone and energy to your current requirements.
Academic Assignments:
Supply the assignment prompt, rubric, grading criteria, and any specific instructions provided by the educator.
Business Articles:
Include background information on the product, target audience details, market research data, and key messaging points.
Data Analysis Projects:
Provide sample datasets, data schemas, variable definitions, and any relevant analysis criteria.
Legal Documents:
Include contract clauses, jurisdiction details, and relevant case references.
Marketing Materials:
Provide brand guidelines, insights from previous campaigns, target demographic information, and market trends.
Technical Documentation:
Supply existing documentation, style guides, audience information, and technical specifications.
Financial Reporting:
Include company financial data, metric definitions, reporting requirements, and relevant fiscal periods.
Content Localization:
Provide target language guidelines, cultural considerations, and audience preferences for accurate localization.
Adhering to this structured, clear approach ensures that the AI model receives all the necessary information, resulting in more accurate and efficient task execution.
Editing & Iteration
Selecting A Section
When editing a prompt, you can either select a section of text or the entire prompt. OctiAI v4 will then treat the selected region as a canvas, and any instructions you input into the chat bar will be used to modify the selected region. During iteration, OctiAI's algorithm is very boundary-specific, ensuring that it only re-generates the text in the selected section. This means you can focus on making changes to just one part of the prompt without worrying about affecting the rest of the text. ---Create a new iteration of your prompt
1.) To de-select a previously highlighted section press the escape key. 2.) Simply type your desired changes into the chatbar and hit enter. 3.) OctiAI will then create a new prompt below, which is a completely overhauled to fit your changes ---Tune a section of your prompt
After every generation OctiAI breaks down the full prompt into anywhere between three and a dozen sections, each split at a logic point, each representing some functional componenet of the prompt.1.) Selecting The Section
You have several options on how to select the section of text you wish to alter: a) using the dissection menui.) after generating a new prompt or loading an past one, you will see a menu
ii.) click one of those section to select its text for iteration
iii.) to break the prompt down into new sections:
- without selecting any text, right-click on it to open the action menu
- press the [Edit With AI] buttons
- the menu will provide a set of section suggestion to edit
b.) using the actions menu
i) select a chunk of text and right-click on it to open the action menu,
ii) click the [Edit With AI] button
iii) this will result in the text color changing to a cyan blue indicating it has been selected.
2.) Finishing the Process
i.) After the text has been selected type any improvements you want made to the prompt, into the main task bar.
ii.) Hit "enter" or click the arrow for OctiAI to regenerate the selected section of text according to your instructions
iii.) On completion, OctiAI will generate a set of up to 5 choices to pick from.
iv.) Select whichever fit your needs the best and hit "enter" to confirm in your choice.
Changes & Version Control
Best Pratices
More Documentation Coming Soon
Access to following the features currently limited to "pro" users only...OctiAI V4 - NLP Code, Database & Process
The OctiAI system utilizes a structured database to store all developer and system prompts, referred to as NLP Code. These prompts modify the AI generation algorithm and can be viewed, edited, or deleted by the user. Each AI tool model is assigned a set of modular NLP Code groups, which act as instructional components that guide the AI’s output. Each NLP Code group follows a sequential execution process. In each call, the active NLP groups will remain the same with only the "sections" parameter changing incrementally. The first API call triggers the "section-1" NLP Code for the relevant model, generating an initial response based on the instructions provided. As the system continues its execution, it processes subsequent "section-n" NLP Code groups, building the new output prompt in multiple stages.NLP Code Group Variable Naming
All NLP Code group IDs use the following format:
1. "system" - naming convention to indicate the variable holds a system-level group of NLP Code.
2. "[parameter-key]" - represents the parameter the NLP Code group controls.
2.1 "[section-n]" - a single integer value representing which output is desired at the current stage of the prompt's generation.
- This is only present in the NLP group variable name if the [parameter-key] is "section."
3. "[parameter-value]" - represents the model's value, allowing it to alter the parameter's behavior.
Each "part" is separated by dashes ("-").
Example:
1.) system-job-defaultText
[parameter-key] = "job"
[parameter-value] = "defaultText"
2.) system-section-3-image
[parameter-key] = "section"
[section-n] = "3"
[parameter-value] = "image"
Finding Relevant NLP Code Keys for a Model
To identify the correct NLP Code values associated with an AI model:1. Use "Inspect Element" on any model in the "Select an AI Tool to Prompt" dropdown menu.
2. Look at the model's HTML element, find the innerHTML dataset properties, and locate the "data-envdata" object.
3. This object contains the parameters whose values determine the NLP Code that will be executed.
4. Find the function containing both the key and value you need.
5. Look under "data-itercount" properties to see the total number of sections outputted to format your improved prompt.
- This is the highest value for [section-n] which could be used by this AI Tool/Model.
- It will not exceed that value, even if a higher [section-n] value NLP Code group exists for the same [parameter-value].
Formatting of NLP Code (System Prompts) for OctiAI
All internal system prompts undergo pre-processing before being sent to the NLP query completions API. This ensures: - The prompt contains only essential instructions for clear AI execution priorities. - Standard prompt engineering techniques are applied dynamically in the backend, improving efficiency and readability. - Adhering to conventional prompt structuring remains beneficial, as it provides additional control over pre-processing interpretation.Deleting NLP Code Groups
- Only existing NLP Code tabs under "internal system prompts" override default settings. - If an NLP Code group is deleted, the system automatically reverts to its default functionality. - Deleting a custom instruction follows the same process as deleting a standard prompt. - Users can re-create custom instructions by loading the NLP Code group from "+ Edit an Internal Function."API Intergation
Currently unavailable for public use due to OctiAI V4 being in beta Additional Notes:Different models may reuse the same NLP Code groups, e.g., for structuring "input data."
Some models reuse NLP Code groups, especially for shared functions like structuring input data.
Models run for n iterations (n = itercount) to fully generate a response.
Certain models may include additional output sections depending on their iteration count.
Unfortunately, Users cannot create new NLP Code variables, modify "itercount," or "envdata" properties & values
Instead, they can adjust the NLP Code itself assoicated with said values to influence AI behavior.
Adjusting said NLP Code allows users to customize & control the system’s output format and style.
This allows users to hardcode built-in custom commands, add precision to how "generation" & "iteration" functions behave.
Remember: OctiAI v4 is still in beta
- Some features may not work as intended. - Several of the current processes listed above may later be depericated in future updates. - Documentation as of march 18th, 2025 is rather limited. More information will be appended shortly. - OctiAI is a work in progress, and we're constantly working to improve our AI tools and services. Please contact support with any bugs or feature requests at: projectloomie@gmail.comOctiAI V4 currently in beta. Please report all bugs to support.
OctiAI V4 is not intended as a free service, access to features is limited.