Use our Polymash version of Chat-GPT, enhanced with chat history, voice capability, folders, web search, prompt libraries and more...

Polymash AI

Run JavaScript

Run general-purpose JavaScript in Toolbox. Recommended for document generation, image processing, archives, and file creation. Choose the runtime whose listed libraries best support the task. Prefer available libraries over reimplementing their functionality. Runtime recommendations are not restrictions. Always try to reuse an existing session_id for related work. If the response says sessions are unavailable, each call starts fresh; include all required inputs in every call. Save reusable values in SESSION_STATE; ordinary const and let variables are local to each call. Responses list saved variables and loaded libraries. Select libraries with the libraries parameter: papaparse (Papa) for CSV/TSV, xlsx (XLSX) for spreadsheets, lodash (_) for data transformations, jszip (JSZip) for ZIP archives, pdf-lib (PDFLib) for creating/editing PDFs, docx (docx) for Word documents, pptxgenjs (PptxGenJS) for creating presentations, jimp (Jimp) for raster images, and ffmpeg for audio/video processing. Selected libraries remain loaded in the session. Omit libraries or pass [] when no additional libraries are needed. Use the selected globals directly; no imports or package installation are needed. This is a browser sandbox: use in-memory bytes and export through output.file, not filesystem or download helpers. Keep Papa worker and download options disabled and check parse errors. pdf-lib cannot extract text or render PDFs; standard PDF fonts have limited character support. pptxgenjs cannot import existing presentations. Jimp uses the v1 API: fromBuffer(bytes), resize({ w, h }), crop({ x, y, w, h }), and new Jimp({ width, height, color }). Use getBuffer for image export; avoid filesystem paths, URL loading, and font downloads. Use browser Canvas for text drawing and SVG/WebP decoding. Selecting ffmpeg provides an already-loaded ffmpeg instance; do not call load() or terminate(). Write inputs with await ffmpeg.writeFile("input.mp4", new Uint8Array(await files[0].arrayBuffer())). Run await ffmpeg.exec(["-i", "input.mp4", "output.mp3"], 60000); check the returned exit code is 0, then export await ffmpeg.readFile("output.mp3") through output.file. For ffmpeg.ffprobe, write JSON metadata to a file with -o and validate that file; this build may return -1 even on success. FFmpeg and its virtual files persist in the session; delete unneeded files with await ffmpeg.deleteFile(name). Prefer short clips and small files; transcoding is CPU- and memory-intensive. All FFmpeg assets are bundled, so local-file processing works without network access. Selected attachments are available in the files array as File objects, in the requested order, each with an additional id property. Read UTF-8 text with await files[0].text(), binary bytes with await files[0].arrayBuffer(), or raster images with await createImageBitmap(files[0]). For SVG, use new Image() with an image/svg+xml Blob URL created inside the sandbox, await image.decode(), and draw the image onto a canvas; do not pass an SVG Blob to createImageBitmap. Include the SVG namespace and explicit width/height in SVG markup. Revoke the Blob URL after drawing. Export PNG with await canvas.convertToBlob({ type: "image/png" }) on an OffscreenCanvas. When network access is disabled, only blob: and data: image sources are allowed. Return a value to produce text output; objects and arrays are serialized as JSON. Create chat attachments with output.file({ filename, data, mimeType }); data accepts a string, Blob, File, ArrayBuffer, or typed array. Use a filename without paths. mimeType is optional and defaults to text/plain for strings, the Blob type when available, or application/octet-stream. Call output.file multiple times to create multiple files; you can also return text in the same run. Network access is disabled by default and controlled by the Toolbox setting.

toolbox_run_js

Run Python

Run general-purpose Python in Toolbox. Recommended for data processing and analytics, including dataset cleaning, aggregation, statistics, and numerical calculations. Choose the runtime whose listed libraries best support the task. Prefer available libraries over reimplementing their functionality. Runtime recommendations are not restrictions. New sessions provide the Python standard library. Import its modules normally. Load third-party dependencies through the required packages parameter before code runs. Available packages: numpy for numerical arrays, pandas for tabular data analysis, matplotlib and seaborn for static charts, plotly and bokeh for interactive HTML charts, pillow for images, openpyxl for Excel workbooks, pypdf2 for reading and editing PDFs, python-docx for Word documents, requests for HTTP requests, beautifulsoup4 for HTML parsing (from bs4 import BeautifulSoup), jinja2 for text/HTML templates, and jsonschema for JSON validation. Dependencies load automatically, including NumPy, pandas, Matplotlib, and Pillow where required. Write packages before code. For import pandas as pd, include "pandas" unless the session response already lists it as loaded. Use packages: [] for standard-library-only code or when all required packages are already loaded in the session. Then import the selected packages in code; loading a package does not create an import alias. If an available package raises ModuleNotFoundError, retry in the same session with that package included in packages. Other third-party packages and pip installation are unavailable. For Matplotlib or seaborn charts, call matplotlib.use("Agg") before importing matplotlib.pyplot as plt or seaborn, and save figures to io.BytesIO. Pass the buffer's getvalue() to output.file with a filename and MIME type; close figures with plt.close(fig). For Plotly, export fig.to_html(include_plotlyjs=True, full_html=True). For Bokeh, use bokeh.embed.file_html with bokeh.resources.INLINE. Export that HTML through output.file with MIME type text/html; avoid show() and CDN resources. Plotly static image export is unavailable. For images, select pillow and import from PIL, for example from PIL import Image. Save images to io.BytesIO with an explicit format, then export its getvalue() through output.file. For pypdf2, import PdfReader and PdfWriter from PyPDF2. For python-docx, import Document from docx. Save Excel, PDF, and Word documents to io.BytesIO and export its getvalue() through output.file. For HTTP, select requests and import requests. Network access must be enabled; browser CORS rules apply. Use a timeout and call response.raise_for_status() before reading the response. For HTML, use BeautifulSoup(markup, "html.parser"). Always try to reuse an existing session_id for related work. If the response says sessions are unavailable, each call starts fresh; include all required inputs in every call. Variables, imports, loaded packages, and virtual files persist in that session. Responses list saved variables and loaded packages. Attachments selected for the current call appear in the files list in the requested order as dictionaries with id, name, type, and path. Their bytes are available in the virtual filesystem at files[0]["path"]. Read text with pathlib.Path(files[0]["path"]).read_text(), bytes with .read_bytes(), or CSV with pd.read_csv(files[0]["path"]). Use print() or a final expression for text output; values use their Python string representation. Create chat attachments with output.file(filename, data, mime_type=None). Data must be a string, bytes, bytearray, or memoryview; filenames must not contain paths. Text defaults to text/plain and bytes to application/octet-stream. For example, output.file("summary.csv", df.to_csv(index=False), "text/csv"). For a file written by a library, read its bytes and pass them to output.file. Call output.file multiple times to create multiple files. Only files passed to output.file become chat attachments. Top-level await is supported, but top-level return is not. Network access is disabled by default and controlled by the Toolbox setting.

toolbox_run_python

Toolbox

Image Search

Stable Diffusion v3 Image

Web Search

Word Generator

Perplexity Search

Extract Web Page

Read Web Page (Firecrawl)

Trigger n8n Webhook

Sends the conversation data to an n8n webhook for further processing.

send_conversation_to_n8n

n8n Connector

GPT Image 2.0

get_recipe

Read one recipe from Paprika in full, by uid. find_recipes returns only names and the lines that matched; this returns the whole thing.

Use it when the user wants to revise something they have cooked, or to look at a recipe the search turned up. Keep the hash it returns and pass it back to update_recipe.

update_recipe

Revise a recipe ALREADY in Paprika. Every revision keeps the version it replaces - the service saves the previous version before writing and refuses to write if it cannot - so a bad revision can be put back.

Call it the moment the user asks for the change: 'update it', 'change that', 'halve the sugar', or a yes to a change you proposed. Do it in that same turn. Do NOT first quote the whole new text back for approval, confirm the title, ask what a note should say, or warn that there is no undo - decide, write, and then say in one line what changed, naming the recipe. The user asked for exactly this after an update that only went through on the seventh answer.

Read the recipe with get_recipe first - search returns only names and matching lines. Take the uid from there, never from memory. The one time to ask is when two results have similar titles and you cannot tell which they mean.

Only the fields you send are changed; anything you leave out keeps its current value, which is what protects the photo and the categories. Send the COMPLETE new text for any field you change - it replaces rather than merges.

Say in notes what changed and why, in plain words you write yourself, so the recipe does not silently differ from the one they cooked.

OMITTED FIELDS: leave a parameter out entirely if you are not setting it. Do not send a placeholder, a dash, or the word 'unchanged' - the service treats an unfilled parameter as 'leave this alone', and anything else as a value to write.

find_recipes

Search the recipes ALREADY in the user's Paprika. Read-only.

Matches across the recipe's NAME, INGREDIENTS, DIRECTIONS, NOTES, DESCRIPTION and SOURCE - so 'wok' finds recipes whose method says so, and 'nytimes' or 'zimmern' finds ones from that writer. Also filters by rating, favourite, category and source, and the search term may be left empty when you are only filtering: that is how you answer 'what are my favourites' or 'what have I rated five'.

Call it whenever the user asks what they already have, and before proposing a new recipe that might duplicate one.

**Read match_type before you describe the results.** A term of several words is tried as a phrase first; if nothing has the words side by side it falls back to near misses - every word somewhere in the recipe ('all_words'), then any one of them ('any_word'). Only 'phrase' is a straight answer. Near misses must be offered as the nearest thing, never as what was asked for.

**Never tell the user they have nothing.** Matching is on whole words, so 'peppercorn' does not find 'black pepper' and a cuisine name finds nothing at all - search the ingredient a recipe would certainly name, like 'doubanjiang' rather than 'Chinese'. A dish name is often not the recipe's title: 'pepper steak' may be filed as 'Beef with Black Pepper'. If the user named a writer, a site or a book, use the source filter. When your searches cannot answer the question, say which ones you ran and ask what the recipe might be called.

Each result says in matched_in which field the hit came from and in matched_words how much of the term it actually contains, and carries the rating, whether it is a favourite, its source and its categories.

publish_recipe

NEWLINES: in every text field below, write line breaks as the two characters backslash-n, never by pressing Enter. A real line break makes the request body invalid JSON and the call fails inside the plugin before it is sent.

Publish a completed recipe to the user's Paprika Recipe Manager, optionally with a generated food-photography image. ONLY call this once the user has explicitly asked for the recipe to be published - 'publish it', 'send it to Paprika', or an equally plain yes. Finishing the recipe is NOT permission to publish it; the user develops recipes in conversation and decides when one is ready.

When they do ask, publish in that same turn. Do NOT ask about the title, the servings, metric, the categories, the description or anything else first - settle each one with your own judgment, publish, and then say in one line what you chose so it can be changed afterwards with update_recipe. The only step that may sit between 'publish' and this call is preview_recipe_images, and not even that if they declined a photo or said to skip the choice.

Send the recipe as it stands after every revision agreed in the conversation, not the first draft. Do not print the recipe as markdown in the chat as well - publishing is the delivery.

OMITTED FIELDS: leave a parameter out entirely if you are not setting it. Do not send a placeholder, a dash, or the word 'unchanged' - the service treats an unfilled parameter as 'leave this alone', and anything else as a value to write.

preview_recipe_images

NEWLINES: in every text field below, write line breaks as the two characters backslash-n, never by pressing Enter. A real line break makes the request body invalid JSON and the call fails inside the plugin before it is sent.

Generate several food-photography options for a recipe and return their URLs, so the user can see them and choose one BEFORE the recipe is published. Call it as soon as the user asks to publish, without asking whether they want a photo - unless they declined one earlier, or told you to pick it yourself, or said no more questions.

Describe the FOOD only, in the three fields subject, surface and serving. Every word about photography - camera, lens, light, plate, grain - is supplied by the service and is the same on every recipe, so do not write any of it here.

Write those fields from the recipe you just developed, NOT from the dish's name. A title says 'charred cabbage'; the camera needs to know it is a two-inch wedge rather than half a head. Give real dimensions, how many pieces, how each was cut, and what the cut surface reveals.

State what things ARE. Never write an exclusion such as 'not bok choy' or 'no sauce' - image models have no negation and weight up whatever you name, so an exclusion summons the thing you meant to avoid. Exclusions are stripped before the prompt is sent.

Takes roughly 20 to 30 seconds, so tell the user the photos are being drawn. Show every returned URL as a numbered markdown image and ask in one line which they want, or none. That is the only question in the publishing flow. 'Publish', 'fine', 'any' or 'you pick' means the first; 'none' means publish without a photo. Then call publish_recipe passing the choice as image_url. The links expire after about an hour.

get_groceries

Read the user's Paprika shopping list. Read-only.

Use it for 'what's on our shopping list', 'do we need anything', and before adding something when knowing what is already there would help.

Returns what is still to buy, grouped by aisle in by_aisle. **Read it out by aisle, not as one long list** - that is the order the items will be walked past in the shop. Items we could not place sit under 'Unsorted'.

add_groceries

Add items to the user's Paprika shopping list.

Send everything they named in ONE call - 'add oat milk, oats and bananas' is one sentence and should be one write.

**You do not need to check for duplicates first.** The service refuses anything already waiting to be bought and tells you which in already_there. Read the returned message back and say what was skipped, rather than claiming it was all added.

Each item is filed into an aisle by OUR GUESS, not by Paprika - the server files nothing itself. The aisles field says which. Mention them only when the user would care, and treat a wrong one as correctable in the app.

check_off_groceries

Tick items off the user's Paprika shopping list, as bought.

Matching ignores case and treats singular and plural alike, so 'bananas' finds 'banana'. Anything that was not on the list comes back in not_found - say so rather than implying everything was ticked off.

attach_photo

Puts a photograph the user attached to their message onto a recipe that is already in Paprika, replacing whatever picture it has now.

Use it when the user attaches a photo of something they have cooked and wants it on the recipe. Find the recipe with find_recipes first and pass the uid it returns.

You do NOT need to be able to see the picture. The plugin reads the attachment itself and sends it straight to the service, so this works whether or not you can look at images.

The photograph must be attached to the message that asks for this. If it was attached earlier in the conversation, ask the user to send it again with the request.

This OVERWRITES the recipe's current picture and there is no way back to the old one, so be sure of the recipe before calling it.

find_equipment

Look through the kitchen equipment list - knives, pots and pans, appliances, boards, utensils - by category, by words, or both. Read-only.

The owned equipment is usually already in your instructions under "Juergen's kitchen". Call this when it is not there, when he asks about the wish list (status wanted) or things he has got rid of (status gone), and before add_equipment, to check the thing is not already on the list.

Each item has notes (how he uses it, in his words), cautions (hard rules - never break one) and unconfirmed points (never state these as fact). Show photo_url as a markdown image when seeing it would help.

⛔ An empty result means nothing is WRITTEN DOWN, not that the kitchen lacks it. Never tell him he does not own something. Ask.

get_equipment

One piece of kitchen equipment in full, by id: every note, caution and unconfirmed point, and its photo link. Use it before correcting an item, so you remove and add against what is actually there, and when the kitchen list in your instructions is shown one line per item.

add_equipment

Add one piece of equipment to the kitchen list - a knife, a pan, the coffee machine, a whisk. It becomes part of what you know about the kitchen from the next new conversation.

**If he attached a photo, call this straight away with what you know**, and fill in the rest afterwards with update_equipment. The photo can only be read from the message it was attached to, so asking five questions first loses it.

You do NOT need to be able to see the photo; the plugin reads it and uploads it itself.

⛔ **Write only what he said or confirmed.** If you can see the photo and recognise a brand or model, say what you think it is and ask - do not put it in brand or model until he agrees. Anything he is unsure of goes in unconfirmed.

Write notes in his words, not yours. One thing per call; for several things, call it once for each and set use_attached_photo false unless the photo is of that one thing.

A name already on the list is refused - use update_equipment on that item instead.

update_equipment

Correct or add to a piece of equipment already on the kitchen list. Use it whenever he tells you something you had wrong, or something you did not know - 'the wok handle gets hot', 'that pan is 10 inches, not 8', 'never put this in the dishwasher', 'we got rid of the old nonstick'.

Three kinds of change, and they behave differently:
- **Adding** (add_notes, add_cautions, add_unconfirmed) keeps everything already there.
- **Replacing** (name, category, status, brand, model, size, material) swaps the value. When he settles something that was unconfirmed, also take it out with remove_unconfirmed.
- **Removing** (remove_notes, remove_cautions, remove_unconfirmed) takes out an entry he says is wrong. Give a few of its words; they must match exactly one entry, or nothing at all changes.

⛔ **Never rewrite his notes to read better.** Add what he said in his words. Read the item first with get_equipment if you are removing anything, so you match what is really there.

When something leaves the kitchen, set status to gone. Nothing is ever deleted. Every change keeps the version before it, so a wrong one can be undone.

To replace the photo, he attaches the new one to the message asking for it and you set use_attached_photo true. Say what changed, using the changed list that comes back.

Login to Polymash AI