
RP Fiber Power – Simulation and Design Software for Fiber Optics, Amplifiers and Fiber Lasers
| Overview | Interface | AI Assistant | Physics |
| Speed | Demos | Power Forms | Versions |
The AI Assistant
Since version V9, RP Fiber Power contains an AI assistant: an intelligent advisor built into the software. You enter your prompts in a chat window, in your own words and in your own language. It does not just give general advice: It works with the program on your behalf. It reads the settings of your simulation model, runs simulations and fetches computed values. On request, it can also change field values of a form or the code of a script – always in a way that you can see, and for anything lasting only with your consent.
For high quality of the outputs, the assistant is equipped with a detailed knowledge base, informing it on both the underlying physics and the details of this software, e.g. the syntax details of the script language. In addition, it consults our detailed help pages where necessary, and can refer you to them. With these means, we achieve far more competent and reliable advice than you would get e.g. by consulting a general-purpose AI chatbot. Nevertheless, like any AI system, it can make mistakes; so check important results before you rely on them.
What You Gain
If you are new to the software or to fiber simulations, the assistant helps you to get going quickly:
- Ask what a Power Form simulates and what its settings mean, instead of first searching the documentation.
- Let it run the simulation and explain the results, also in terms of the underlying physics – with links to articles of the RP Photonics Encyclopedia where background knowledge helps.
- Ask why a result is as it is – for example, why the power conversion efficiency is so low – and get a diagnosis based on the simulation of your design.
- When dealing with script code, let it explain how it works or how it could be improved.
So you will learn far more quickly both about the physics and how to handle this software.
If you are an experienced user, it efficiently takes routine work off your hands:
- It does tedious series of simulation runs for you, and fetches results which the form does not display – without you writing any script code.
- For a design target, e.g. a certain output power, it presents the possible routes and then carries out the one you choose.
- In the script editors, it tracks down errors, extends your scripts, and writes new scripts for your own devices, based on our demo scripts.
- It can also equip a script with a custom form, so that input data can be conveniently edited in the form instead of using the script editor, and multiple data sets can be saved.
- It compares the fiber datasets delivered with the software and suggests suitable fibers for your task.
So you gain amazing productivity: Get your results more easily and faster, and focus on the interesting aspects!
In all cases, you stay in control: Every change of your design is shown together with its reason, and nothing is saved without your consent.
The Chat Window
You open the chat window with the menu item View | AI assistant, with Ctrl+Shift+A or with a toolbar button.
- The answers may contain tables, and the assistant can put a diagram into the conversation – the program's own drawing of that diagram, not a picture made by the AI.
- Each form and each script has its own conversation: When you switch to another form or editor page, the chat window continues the conversation belonging to it.
- While the assistant works, you see each action it takes – reading the settings, running a simulation, changing a field – as a short line in the conversation, and you can stop it at any time.
- Simple commands such as
/diagramor/againare executed by the chat window itself – instantly, and without any cost. - If you wish, all conversations are recorded automatically in your work folder, beside the project files they discuss, with the diagrams they contain.
Carefully Worked Out Details
Conversations: The assistant remembers everything said in a conversation, so that you can refer to earlier questions and answers. A conversation keeps the AI service and model with which it was started; when you change them, a new conversation is offered.
Commands: A line starting with a slash is not sent to the assistant, but executed by the chat window at once; it costs nothing and does not become part of the conversation which the assistant sees.
/diagram Nputs diagram N into the conversation, so that it stays with the current answer even after the next run has redrawn the diagram window./commentadds a remark of your own to the record, and/stateputs the current form settings into it – optionally only the fields matching a word./againasks the last question once more, e.g. after you have changed some setting./promptshows which model the conversation runs on and where its record is kept,/configopens the settings, and/helplists the commands.
Records: With automatic records switched on (off by default), every conversation is saved after each answer in the subfolder AI conversations of your work folder – so it is archived and backed up together with your project files: as a readable record in Markdown and HTML, with the diagrams it contains, plus a technical transcript. In any case, you can save a conversation as a Markdown or HTML file, print it (also as a PDF file), or copy the last answer as Markdown text, with tables staying intact when pasted.
Always keeps you informed: The status bar says what the assistant is doing at the moment, and which AI service and model it uses; a click on the model name opens a menu for switching the model. A line shows the amount of text (tokens) processed so far in the conversation and, with Claude via RP Photonics, how much of your usage budget is left. In long conversations, you can go to the previous or next question or answer, search the whole conversation, and make the type smaller or larger.
Robust operation:
- Temporary problems of the AI service, such as overload or network problems, are handled with automatic retries.
- If the assistant ends without an answer, the program first asks it once more for one. If a form's settings had already been changed in that request, you are offered to restore the settings from before.
- After 40 steps within one answer, the program asks whether it should carry on. With a weaker model, it may then recommend a stronger one, which typically needs far fewer steps.
Details on Provided AI Assistance
In Power Forms and Custom Forms
With a Power Form loaded, you can ask the assistant, for example:
- What does this form simulate, and what are the main settings?
- Run the simulation and explain the results.
- How could I get a higher output power? Which parameters would you change?
- Is the pump absorption in this design adequate?
- Which fiber length gives the highest gain? Save the best design under a new name.
To answer such questions, the assistant can:
- read the current settings, the meanings of the fields, the output fields and the log area, and it knows whether the displayed results still belong to the current inputs
- run the simulation itself
- fetch further results which the form does not display, e.g. the powers along the fiber or the data behind a diagram curve – with small read-only scripts, which cannot change anything in the model
- compute a series of variants in one step, e.g. the output power for five fiber lengths
Beyond its general knowledge base, the assistant has specific background knowledge on each Power Form: what the form models and where its limits are, typical pitfalls, and how to obtain quantities which the form does not display. The assistant can therefore tell you, for example, when a question goes beyond what the form can simulate, and suggest the nearest thing it can do. With a custom form, it relies on its general knowledge of the software and of the physics, and on what the form shows – the labels of the fields, their hint texts, and descriptions written for the assistant.
Your Design Stays Under Your Control
The assistant can change field values; that is the core of its assistance. The program makes sure that you keep control:
- Before changing values, the assistant says what it is about to change and why. Each change appears in the log area with the old value, the new value and the reason.
- Changes affect the form content only. Your parameter file is written only when you save it – or when you confirm a request of the assistant to save, which the program shows in a dialog. The assistant can also save a design under a new file name; it never overwrites another existing file.
- Reloading the parameter set discards all unsaved changes, those of the assistant included. If a request ends without an answer after the assistant has already changed settings, the program offers to restore the settings from before.
- Fields containing script code are changed only after your confirmation in a dialog, which shows the old and the new text.
In addition, the assistant has been instructed to follow certain working rules. Design decisions are yours: It is to propose changes of device parameters such as powers, wavelengths or fiber data, and to act only after your go-ahead. If you name a target, e.g. a certain output power, without saying what may be changed, it presents the possible routes instead of silently picking one. Questions which simulation runs of a few seconds can settle are answered with such runs; explorations which would take minutes are proposed, and carried out when you agree. Like any AI system, the assistant does not always follow its instructions perfectly – that is why the program itself lists every change and saves nothing without your consent.
In the Script Editors
In the script editors, the assistant helps you with the script on the active editor page. It can explain what the script does, track down errors, answer questions about the script language, run the script and interpret the results, and change or extend the script. If you select some lines before asking, you can ask about just these – for example, “Explain this” or “Equip this with a sanity check for the numerical parameters”.
Every change to your script needs your approval: A dialog shows the text to be replaced and the new text, which you can apply, reject or edit before applying. (Within one conversation, you can allow further changes without asking.) An applied change can be undone with Ctrl+Z, like your own typing, and before the first change to a script file, the program saves a backup copy of it.
The assistant can also write new scripts for you, e.g. for a fiber amplifier or laser of your own. It first asks for the data of your device which it needs and which you have not given – it does not invent such data. It then builds the script, often based on one of our demo scripts, and shows it to you with a proposed file name; you can still change both. An existing file is never overwritten. The assistant then runs the new script and reports the results.
The assistant knows the demo scripts delivered with the software and the code snippets library. It can point you to a demo which matches your task, explain it, and use demos and snippets as patterns for the code it proposes.
Free Choice of the AI Service
The assistant is based on a large language model (LLM) – an AI service running on a server outside the software. Before the first use, you choose one of three options:
- Claude via RP Photonics: Your easiest option. We use the Claude models of Anthropic; your requests are forwarded to Anthropic by a relay server which RP Photonics operates on the infrastructure of the Swiss hosting provider Cyon. You need no account with an AI provider and no API key. The usage is charged against a budget included in your software license, which should be sufficient for typical use over an extended period; the assistant shows you how much is left. You can order a top-up from RP Photonics; we charge the token costs according to Anthropic's published prices (not adding any commission).
- Your own AI service: You configure an endpoint and API key of your own, e.g. of a corporate AI gateway or a local model. Your requests then go directly to that service, without involving any infrastructure of RP Photonics, and your provider bills you directly.
- No AI assistant: You do not use the assistant at all; nothing is then ever sent anywhere, and all other functions of the software are unaffected.
With Claude via RP Photonics, you can currently choose between two models:
- Claude Opus 5.5 (default) – the strongest model, recommended in particular for design and optimization tasks,
- Claude Sonnet 5.5 – cheaper and faster, good for simpler questions.
AI providers regularly release new models and retire older ones, and we adapt our selection accordingly – offering a new model only after testing the assistant with it. The list of models is updated from our server, without a software update.
Note that Claude via RP Photonics is not available in countries and regions where Anthropic does not permit the use of its services (e.g. currently China) or where export control limitations apply. The option to use your own AI service remains available there.
Getting the AI Assistant
The AI assistant is part of RP Fiber Power V9. Existing customers can upgrade to V9.
And of course, our diligent technical support is there for you as before – the AI assistant nicely complements it, and will often provide the quick answer you need.
If you have any further questions or need a quotation: Contact us
