(If you are using Google Chrome browser, it offers a “Pretty print” option which reformat the data, adding spaces and newlines, to make it more readable) Within that JSON object you will find a list of messages, each containing the following fields which you are after:
Has anyone tipps how to lead better conversations with the AI? Right now I seem to be stuck in a circle with every second question being ‘oh that is great. What is your favorite..?’
Hi Steve, I have been inactive for a bit and just logged back in to pick up where I left off on my language studies. I absolutely love the new conversational feature!!! You do such fantastic work here! Thanks for adding this feature.
This is an awesome feature!!! Any dates when it’ll be possible to use this feature with languages such as Arabic and Chinese? I would really appreciate it
I think it’s worth pointing out to anyone who hasn’t tried it that it is Ami, the chatbot, who initiates the conversation. I didn’t use it for a long time because I wasn’t sure how to start. I now see that I needn’t have worried.
I’ve used it four times now, and this seems a really useful tool, especially for those of us who don’t get much chance to actually practise the language. And I like the fact that this gives you time to think (and maybe have a quick check in Google Translate) before you answer. And I really like the follow-up flashcards.
I can certainly see this becoming a daily habit. Thanks so much, Steve.
I’m really liking this feature, and it’s helping me with a long-neglected part of my language learning. I’m fairly good at reading books and watching movies with good comprehension but am very bad at language production.
My feedback has to do with the flashcards that are generated from this feature. For flashcards generated from what I’m reading, passive recall and blitz mode work best. It’s testing my reading comprehension.
For flashcards generated from these chatbot conversations, active recall works best. Many of my cards are fundamental things like using the right form for “the” and getting the end of adjectives right.
However, when I am doing flashcards, these all get mixed into one pool. A flashcard that is insultingly simple with passive recall (“Which word means ‘the’?”), it is incredibly valuable in active mode, especially because the context sentence came from me.
Is there any way to separate these two sets of words? The drilling of them is very different. Or a simple solution is to make “Chat Conversations” its own source. Then I can select “Mastery” mode with that source, and for all others, I can use Blitz.
Edit: I now see that with new vocab it now asks whether to treat it as an active or passive card. I think this would get me most of the way there. However, if I select “active”, it still seems to show me the blitz, multiple-choice version half of the time. It would be nice if the card stayed in the form that I selected.
Still an option to see “Chat Conversations” as its own source would be nice. As well as an option to automatically make cards generated from chat conversations as “active”.
I tried a conversation today for the first time today, so here’s my feedback
What I like
It’s awesome that I can use German or English and can directly add the unknown words to my word list without friction. That’s something that some designated AI chat apps can’t do.
I love that I can easily review the conversation and still add words after completing it.
Room for improvement
Like someone else already mentioned: I found it difficult to keep Ami on (my desired) topic. I tried to talk about my experience of reading a (well known) Spanish text and the conversation went more into a “what other Spanish authors do you want to read” direction. So while I wanted to dig deeper it seemed that Ami wanted to branch out. I would have preferred the conversation to stay more focused.
Is this something that could get better with using the more powerful AI model?
I’d also prefer if the context of saved words was just the sentence and not the whole message.
Wishes for the future
Like mentioned by others before I would love to have conversations about specific texts in my library. This would be a neat integration of Readlang’s core feature and the conversation feature.
I also second the request to have a “Chat Conversations” word source (ideally with a link to the conversation as a source for the context). Maybe even “Chat Conversations” and “Mistakes” so that one could differentiate between words that should be studied passively or actively?
Right now the chatbot is always using the larger gpt-4o model so it’s not quite that easy. The difference between the plans is that with Premium you only get 2 chats per day and with Premium Plus gives you unlimited chats per day.
I agree that chats about a specific text and a way to filter by chat conversations are both good ideas!
