Show HN: Checka – follow content beyond RSS – aggregate, filter, digest

[ace_zhao]

Hi HN, Our team has developed an information processing assistant called Checka. Compared to many other information services, Checka focuses on allowing users to describe their target information using natural language and use LLM workflow to filter the content. Here is the simplified flowchart of information processing: https://images.checka.ai/images/HN_pipeline.png

How Checka Works:

- Data Collection: It gathers information from selected data sources using scrapers, APIs, and RSS feeds, and structures the data.

- Filtering and Analysis: Through multiple LLM requests, it pre-filters and thoroughly analyzes all the information data objects to determine if they meet your specified "target information" criteria.

- Digest: For data that meets the criteria, it uses generative capabilities to digest the information based on your chosen format, making it easy to quickly access the insights you need.

As developers, we’ve all been there—building tools to tackle the endless stream of scattered information. But just simply aggregating content often ends up creating more noise. If you’ve ever faced hundreds of unread items in your RSS reader, you know the feeling. Therefore, we created Checka, hoping it would be useful to some of you. We would greatly appreciate any feedback as we know HN is the best crowd for it.

[YoungMan]

Looks promising. I've been dealing with information overload for years, and my RSS reader is full of unread articles. I noticed the site has a mobile landing page, but it sucks that the reading feature isn’t available on mobile. Using natural language to search for content sounds interesting, but I’d much prefer reading on mobile. I don't like reading news on a desktop.

[ace_zhao]

it's unfortunate we don't have a mobile version yet. we want to test the desktop solution first. If feedback is good, we'll quickly develop the mobile version.

[lelouchbing]

Fantastic and great work! Really like this idea of gathering information from various sources and generating abstract accordingly, which saves users tons of time.

One quick question: Can I customize the filtering based on my own preference? E.g. I'd like to see the finance news more than the others during weekdays.

[ace_zhao]

Thank you! The core feature that Checka aims to offer is personalization. You can select your preferred finance data sources and then describe the type of finance news you'd like to read in the target information, and its done!

[waideai]

This is a promising idea! I wonder how you check and ensure the quality of the filtering and digest. Can you share more about the underlying models?

[ace_zhao]

For filtering, we currently use a two-step process, similar to how humans read information. The first step involves an initial screening through the title and brief description, identifying data that clearly meets the criteria for a direct digest. For data that isn't clearly suitable, we perform a full data evaluation. To save costs, we employ a mentor-apprentice model, where state-of-the-art models (4o and Sonnet3.5) are used for the initial evaluations, and their outputs are recorded as few-shot examples. These examples are then used to guide more cost-effective models in subsequent processing.

[xinyuwu]

If I want to work with the result, what's the move, can I export it? And where can I find the raw data?

[ace_zhao]

Sorry, we currently don't offer any features for sharing or exporting results, and the raw data cannot be viewed within the product at this time.

[feiyue297]

Are you planning to release a local version thus can build local knowledge base?

[ace_zhao]

We currently don't have this plan; for a period of time, this will remain a pure SaaS service.