Hacker Newsnew | past | comments | ask | show | jobs | submit | yccheok's commentslogin

Hi,

I’m currently building a Q&A chatbot and facing challenges in addressing the following scenario:

When a user asks:

"What do you mean in your previous statement?"

How does your framework handle retrieving the correct small subset of "raw knowledge" and integrating it into the LLM for a relevant response?

Without relying on external frameworks, I’ve struggled with this issue - https://www.reddit.com/r/LocalLLaMA/comments/1gtzdid/d_optim...

I’d love to know how your framework solves this and whether it can streamline the process.

Thank you!


After a lot of experimentation, the only thing that worked in a chat style application is to pass maybe the last 4-5 messages (ideally the entire conversation history) and ask an LLM to summarize the question in the context of the conversation.

Without that it often failed when users asked something like ("Can you expand point 2? , Give a detailed example of the above").

Current implementation(I have 3 indexes) is to provide Query + Past messages and ask an LLM to break it down into Overall ask: BM25 optimized question: Keywords: Semantic optimized question:

Perform RAG + Rerank and pass the top N passages after this along with the Overall ask in the second LLM call.


If the user asks such a question, your agent should not invoke the RAG at all, but simply answer from the history. You need to focus on your orchestration step.

Search for ReAct agents, can build using either LangGraph or Bedrock Agents.


Have you tried allowing the LLM to decide the use of knowledge retrieval, through tool use or a direct query?


Spending time on social media platforms like Twitter, ProductHunt, and IndieHacker can often have negative effects on mental health, as seeing others' successes and high earnings can cause feelings of toxic jealousy and comparison.

To improve mental health, it may be beneficial for the author to reduce their time on these platforms and focus on practical tasks such as talking to customers, finding new customers, and developing their product. These actions can lead to a more fulfilling and balanced life.

If the author feels that quitting these social media platforms entirely is the best decision for their mental health, that is a valid choice. Alternatively, the author may find it more mentally healthy to transition from being a solo entrepreneur to an employee with a stable monthly paycheck. Ultimately, the goal is to find a path that leads to greater happiness and well-being.


It is still possible. I have been doing so for a few years, by focusing on single note taking app

https://play.google.com/store/apps/details?id=com.yocto.weno...

Here's my take :-

1) Just like in other businesses, it is a competitive market.

2) However, it is also full of limitless opportunities. Google Play store helps you to reach worldwide end consumers. Only a few channels can do that with a frictionless payment system. Google Play store is one of them.

3) Since it is a highly competitive market, you need to know your niche and know who your targetted customers are very well. Then, we provide a solid solution to fit customer needs.

4) Play store tax is not a concern. Once you publish the app, the only main concern is how to market the app and how to pitch the app so that consumers will choose your app over the others.

5) With some luck, one will hit overnight success. But, the chance is rare. Most of the time, we need to invest a lot of resources, and success is not guaranteed.


Mind elaborating a bit? How is your app doing? What's your model - a free and a paid version or just ads?


There are a few ways for us to make money.

1) We offer 1-time in-app purchases. The price tier ranged from $30.99 to $0.99 (We offer promo prices to users when they share the app with enough people)

2) We offer a monthly subscription model of $5 per month when the user requires better cloud storage to store their notes, image attachments, and voice recordings - https://www.wenote.me/cloud

3) Every 1-time in-app purchase comes with a 7-day free trial. After 7 days of the free trial, we provide an option for users to watch video ads in exchange for additional 2 days of the free trial.

4) We decide not to show banner ads/ full-screen ads. Even though it will make us earn less, it will provide a better user experience, gaining a better user retention rate and better app reputation.


How did you market your app to the point that it has 1m+ downloads?


First, we need to agree that "A great product will speak itself" is total BS.

But, in order to outpace millions of other apps in long run, we need to perform the following

1) ASO App store - So that our apps are searchable with the right keywords

2) Localization App store - So that our app can reach more users

3) A/B testing App store - So that we will find out what is the most effective graphics asset

4) Paid advertising - So that you will get real users

5) ... (other marketing activities)

As we can see, all the above activities require a substantial amount of money (We need to hire people with the right skill to complete the task or pay our ads network)

I think knowing how to spend the marketing money the right way, is the key to success in Google Play/ Apple App Store.

I am still learning the hard way, though.


Congratulation!

Building a marketplace to sell something (Apps, courses, clothes, food, ...) is a very challenging problem.

I am an App developer who sales app in Google Play. The only thing I am concerning is, how much users are Google Play (marketplace) is gonna to bring to me? Rest of the concern factors are secondary to none.

Even if Google Play choose to charge higher fee, or providing a more crappy publishing tool, I still will stay with Google Play. Because I know if I try to sell my Android app elsewhere, I will get 0 customer.

Good luck to your venture!


Thanks for your comment and feedback on the marketplace. Right now, the AlterClass marketplace is more a way for people to launch their courses quickly without worrying about setting up their own platform and hosting the content. I won't compare myself to any app marketplace (like Apple or Google) as it is not my end goal. I want to give the instructors full control of their courses and give them the freedom to access beyond AlterClass. I'm still working on this, so stay tuned ;)


Any chance to make the app available in App Store in Malaysia? Would love to try out on it. Thanks.


Thank you for your interest. Currently we're not operating outside the U.S. because our shipping operations are limited, however, if you're very interested we'd be happy to send you a Testflight link to try. We've had a few members pay for international shipping after they tried the app via the Testflight link.


Ya. I am particular to try out the app itself without hardware first, to see how effective it is. May I know how can I request for the Testflight link? Thanks.


I am indie developer who makes and sales apps on Android. I can tell you this is not true. Users are willing to pay for good apps.


No. Website doesn't provide experience as good as native app. Many ordinary users can tell you this.

- Privacy. A good intention mobile app, can be designed to work without Internet.

- Internet. User doesn't have access to Internet all-the-time, but you still need certain app to work.

- Fast. Web app can never be as fast as native app. It is not realistic to expect code which runs under an JavaScript interpreter layer, can be as fast as native (or almost native) code which is nearer to CPU layer.


I'm more interested to know what kind of product they are selling with that much of Ads marketing dollar.


Don't let Facebook get away with this!


Don't forget Eric as well. He is one of the trio, who turns Google into what it is today.


Eric's contribution to Google was paranoia and urgency from his bitter experiences losing twice to Microsoft. He knew Microsoft would come after them. So, they to widen the gap so far that when the "sleeping giant" woke up, it wouldn't be able to overtake and crush em.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: