Replies: 10 comments 14 replies
-
interesting, using chat GPT for that would probably work. Many people who self host prefer to have everything run locally on their device so the ideal solution would have some kind of pre-trained model available directly in the application that is specialised for these kinds of tasks and can run on a lighter infrastructure but I am not sure if AI is advanced enough for that. Do you know any more about this or want to figure it out? Could you also post a few examples of image to json conversions that chat GPT can handle and which promts you used so I can get a feel of how good it is. maybe integrating with open AI as a optional component would be something useful for people. |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
I forced it to insert all ingredients and steps. Sadly Dall-e seems to be down, otherwise it would have added a tasty image.
|
Beta Was this translation helpful? Give feedback.
-
that looks really cool, I have no idea when I will be able to work on this but if anyone wants to pick it up I suggest to implement it in a way that a user can add an API token to OpenAI (or oauth or something) and then add a special page to the ImportView that allows images to be send to OpenAI. The response can either be directly or parsed into the format returned by recipe-scrapers or a special scraper could be build to then push the recipe into the normal parsing flow as any normally URL imported recipe is. |
Beta Was this translation helpful? Give feedback.
-
Open AIs ChatGPT integration would be helpful for my use case as well. I am not using images, but recipes published on Instagram (please don't judge me 😅) in instructered plain text. And I want these out of the walled garden in a more usable interface. I had success by manually using ChatGPT 3.5 to convert the plain text into JSON and then importing the generated JSON to tandoor. If this workflow could be automated (ideally, provide link or plain text) this would be great. |
Beta Was this translation helpful? Give feedback.
-
I'm doing something similar! I already have the recipes in a MySQL database. I have the code to extract the recipes, but I'm having a horrible time trying to post the data to Tandoor using the API. I can create a recipe, add the steps of directions, but then when I try to add the ingredients, through a find or create routine, I start getting complaints. When I look at the recipe with my browser, I see the recipe, there's steps but no ingredients. When I try to edit, there's no place to add ingredients! It seems like I have to add the ingredients to the data base and then link then back to the recipe. This is hard! LOL |
Beta Was this translation helpful? Give feedback.
-
Looking really forward to see here some progress. I actually wanted to checkout tandoor 2.0 implementation but docs in todays release notes were not accessible. FYI @vabene1111 |
Beta Was this translation helpful? Give feedback.
-
Has anyone managed to get the ai feature working with a self hosted ai? The docs seem to imply that it should be possible but it's not clear how that could be done |
Beta Was this translation helpful? Give feedback.
-
@vabene1111 would it be possible to do separation into steps (Schritte) automatically by AI import? I just wonder how hard it would be. I do have some basic python programming skills maybe I could attempt. This way it would be easy to follow stepwise introductions provided. ATM its imported as one. |
Beta Was this translation helpful? Give feedback.
-
Maybe helpful to share how I currently use AI and tandoor. I have a space (similar to a gpt) in perplexity with these settings and a recipeformat attached. I can include multiple images or just a link, and it converts it. I import this with the JSON/HTML option. Any improvements or ideas are welcome. I like the idea of adding a list of existing foods in my database to this space. Additionally, I use the kitshn app to import recipes from instagram. You are tasked with converting a recipe from a given source (which could be a picture, link, or video) into a standardized HTML format. Your goal is to accurately extract the recipe information and present it in the provided HTML template. For every question, first analyze the content of 'recipeformat.txt'. as the main source for the html structure. Custom promptNow, carefully analyze the recipe source provided: <recipe_source> Follow these steps to extract the recipe:
When filling out the template, make sure to:
Your final output should be the completed HTML template with all the recipe information filled in. Do not include any explanations or additional text outside of the HTML structure. Always use 'recipeformat.txt' as the main source for the html structure. Recipeformat.txt
Output
Another case I use AI for is converting a menu. I then give perplexity this prompt: create 1 recipe where each dish is a separate step with ingredients |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hi all! I just installed tandoor via docker/portainer. It works really nice.
I was looking into tandoor because a friend showed me crouton app, which can import recipes from a smartphone photo. I would like something similar and it probably requires AI to identify ingredients and amounts in recipe photo and text.
So far I was able to do it with OpenAI/ChatGPT. It can read photos. It can also convert a spoken explanation of a recipe into a nicely structured and written recipe. It can even generate a photo from the written recipe.
I got stuck with Tandoor in the import section. I asked ChatGPT to format the recipe as JSON, tried pasting that in source import, but that didn't work.
I'm reasonably fluid in Python, I never worked with Django, and I know very little of the Tandoor architecture and roadmap. What would be a good way to integrate an AI API into the source?
https://platform.openai.com/docs/api-reference/making-requests
Beta Was this translation helpful? Give feedback.
All reactions