[HN Gopher] Show HN: Sequel - Effortlessly understand your data
___________________________________________________________________
Show HN: Sequel - Effortlessly understand your data
Hey, I'm a solo developer building Sequel as a side project. I'm
excited to finally launch Sequel! Sequel is an AI tool that will
transform how your team interacts with data. Sequel helps you
translate natural language queries to SQL and helps you analyse the
result just like a data analyst, without even writing a single line
of SQL query. What you can do with Sequel? Chat with your
databse: Connect your database and ask questions to get results.
Generate Graphs and Charts: Ask sequel to give you charts and you
will get beautiful visualizations. Analyse your data: Sequel's AI
helps you understand and uncover hidden insights from your data.
Setup New Database: If you are setting up a new databse, just
describe your schema to Sequel and Sequel will create all the
tables and columns for you! Would love to see how your team uses
Sequel. Looking forward for your feedback and suggestions!
Author : haxzie
Score : 5 points
Date : 2024-09-24 13:52 UTC (9 hours ago)
(HTM) web link (sequel.sh)
(TXT) w3m dump (sequel.sh)
| bbor wrote:
| Looks awesome! Not the kind of thing I need, but I see the clear
| value prop. And the site is slick af.
|
| Some random critiques:
|
| 1. You have a tiny typo in the "Uncover insights with AI" example
| (quire).
|
| 2. The docs are broken on mobile screen sizes. Not a common use
| case for actually reading them, but I was curious!
|
| 3. The FAQ accordions aren't individually closable and seemingly
| use _MouseDown_ instead of _Click_ , which makes scrolling on
| mobile a little wonky.
|
| 4. Your (Midjourney?) images are killer. Some of the best usage
| I've seen yet for a professional site.
|
| And finally, two actual critiques/questions on real
| functionality:
|
| 1. I _love_ the most recent blog post and agree with it 100%, but
| was surprised to see this: "Instead of wrestling
| with complicated BI tools, users can instantly create charts,
| graphs, and live dashboards that help them make sense of complex
| data."
|
| That seems like a HUGE feature, with a ton of edge cases! I do
| now see that it's on the home page, too. You don't have to answer
| _at all_ if it's secret sauce, but I'd be curious to hear what
| library you're using for the visualizations -- presumably not
| direct d3?
|
| 2. Have you considered/explored paradigms that don't require
| users to send you their DB credentials, and instead have some
| sort of local agent as a middleman? I guess the difference is
| arbitrary at some point (a middleman could pull my data too!),
| and I'm so far from an expert, but it just feels weird sharing DB
| login info just to get help writing queries. More curious than
| anything.
|
| 3. I love the privacy doc, but I'd recommend sharing any LLM
| providers that would get access to this data, or if you're
| running it all yourself, bragging about that more explicitly! I
| personally don't care about OpenAI/Anthropic/Google seeing (more
| of) my data, but I know a lot of people do.
|
| 4. This is all fun and games for queries, but what sort of
| protections are in place for INSERT, UPDATE, or, god forbid, DROP
| statements? I'm the biggest LLM fan around, but I think it'll be
| a while before I'd let an LLM run such queries directly. Unless
| there's some sort of rollback-able transaction support, and/or
| deterministic safety checks?
|
| Thanks for sharing, no pressure to answer any of those lol. Very
| inspiring for my own work, especially as a solo dev! Keep cookin'
| ;)
| haxzie wrote:
| Wow, thank you for taking your time to go through the app!
|
| 2. Will fix the docs to work on mobile screens, this was a last
| moment run to setup docs, wanted to try it without any
| templates. Have to manually handle the responsiveness now :)
|
| 3. Makes sense, this shouldn't have caused issues on mobile.
| Thanks for letting me know.
|
| 4. Midjourney did a pretty good job with the images, got em
| done within no time! haha
|
| ---
|
| Visualisations are done using recharts. Currently I have
| implemented a very limited set of charts, hopefully will add
| more with much more customizability.
|
| I am planning on opensource the product soon, going ahead with
| an open core model, and a self hosted version along with it.
|
| The concern about LLM providers makes sense, will add them to
| the docs as well.
|
| Currently there is no hard wired instructions to not let the
| LLM insert/drop it. It's all as good as the user instructing it
| to do something and it does. There are soft instructions to not
| generate queries for that, but it's as good as telling a kid
| not to eat chocolates. I am planning to run a post analysis on
| the queries generated and marking them as potentially
| dangerous, so that the queries don't get executed
| automatically, but with a confirmation from the user. This
| should ideally be separate from the actuall LLM which generates
| the query.
|
| Again, thank you so much for the valuable feedback. This helped
| me a lot!
___________________________________________________________________
(page generated 2024-09-24 23:02 UTC)