My friend Sami and I built Vizly, a project that lets you query and visualize your data in seconds<p>It works for any data so you can upload and starting querying/visualizing your dataset right away
I'd say this project is rather useful in that it made queries far much more intuitive, something SQL aimed for in 1970s but eventually fails: for simple ones like `SELECT type_1 FROM roles;` it is indeed intuitive, but for something a little bit complicated, that intuitiveness quickly fall apart:<p>```sql
SELECT DISTINCT type_1 FROM roles AS r1
WHERE NOT EXISTS(
SELECT * FROM roles AS r2
WHERE r2.type_1 = r1.type_1
AND r2.legendary = 'True'
);
```<p>This queries all "`TYPE_1`s that all characters in this type are not legendary", and it is near impossible for a newcomer who had never learnt about SQL before to figure out the query, yet typing the sentence in quotes in `Vizly` gives exactly what is needed. Sure it took 1 minute to execute (which is probably why some thought it is down), but formulating the above SQL statement took much longer, and I am not sure that statement even works (I haven't touched SQL for quite some time).
Nice idea and interesting space. I tried 'show counts by type_1 ordered by occurrences' and it showed a bar chart, as expected. But 'show statistical distributions' or 'show something surprising about the data' gave the response 'Vizly does not yet support this chart type.' Note that some of the standard tools like Tableau or Observable show basic charts over columns out-of-box after loading a dataset, so the default tabular view could be augmented. I'm aware of at least one mature-looking tool in this space [1] and an interesting thesis [2]. Best wishes with building it out!<p>[1] <a href="https://datachat.ai/" rel="nofollow">https://datachat.ai/</a>
[2] Iris Agent - <a href="https://youtu.be/3VZZbKoXDVM" rel="nofollow">https://youtu.be/3VZZbKoXDVM</a>
Nice idea. Not sure it’s fully ready yet, and I have a few suggestions.<p>One, you need more content to explain the concept to the user. A table and an input box is not sufficient. Sure, I could mess around and figure it out, but for all I know I’d be missing out on all the cool features.<p>Two, provide some examples, and some alternative datasets (e.g. tabular, nested, relational).<p>I’m not sure how to use it, so I gave the prompt:<p>> plot speed vs defence<p>on the Pokémon dataset. It provided me with an empty scatter plot.<p>Three, you need an FAQ. Some example questions: Can I run this locally? Can I integrate this with private data sources such as Prometheus? How does the project work? Who worked on it? Are there terms and conditions? What kind of data sources does it currently work on? What are the ongoing plans? Etc.
Super cool demo. I uploaded a messy fraud dataset and asked it some queries like "when have you seen the most fraud", and it was able to convert this to a COUNT aggregated on days.<p>At least for investigative workflows (where it's pretty common for data scientist to often have to rebuild custom dashboards), it'd be really cool to have the system handle prompts such as "give me a regex on email address that seem to incur the most amount of fraud" -- or unveil some insight on the data that is usually just a mix of operator/data scientist domain specific knowledge + pattern matching.