Has Claygent stopped PDF parsing functionality? I've had success getting it to analyse UK Company filings in the past, which requires opening a PDF and analysing it, but it's been working inconsistently often errors with a message suggesting it does not have this functionality. Am running on Argon.
PDF parsing functionality is still available and has actually been enhanced with our latest Claygent Neon model, which shows improved capabilities in parsing PDFs and extracting hard-to-find information.If you're experiencing inconsistent performance or errors, this could be due to several factors: 1. Processing delays or accuracy issues can be resolved by: • Making your search queries more specific • Breaking large data requests into smaller batches • Checking your usage limits and API quotas • Verifying your network connection stability • Double-checking your prompts and search parameters Source: Introducing Claygent Neon
Our PDF parsing functionality is still available and has actually been enhanced with Claygent Neon, which shows improved capabilities in parsing PDFs and extracting hard-to-find information.If you're experiencing inconsistent performance or errors, this could be due to several factors. Here's how to address these issues: 1. Make your search queries more specific and well-defined 2. Break large data requests into smaller batches 3. Check your usage limits and API quotas 4. Verify your network connection stability 5. Double-check your prompts and search parameters Source: Introducing Claygent Neon
Looks like you're still looking for help. We're escalating this over to our support team and someone should be in touch soon. In the meantime, feel free to include any other information that may be helpful in solving your issue!
Hi J, thanks for reaching out. Taking a look at this now!
it's returned some numbers that are all inaccurate where it has returned them. Didn't used to behave like this
Just to clarify, are the PDFs you're trying to deploy Claygent for publicly available?
yeah they are
Could you clarify which column you're building your prompt in so I can take a closer look?
Revenue Extraction
Could you help identify the actual revenue figure in the document linked above? This appears to be a complex reasoning task, and the document doesn’t seem to be optimized for searchability, which might explain the performance issues Claygent is encountering.
54,485,600
It seems Claygent might not be performing optimally in this case. I recommend fine-tuning your prompt to focus on a specific section of your PDF. Alternatively, you could use an API actor to extract information directly from the PDF. This one might be a good option: PDF Scraper by Apify.
So you know, I made a couple of tests, iterating over one that finds the page, which it did accurately and a second which found the Turnover, which it did right once out of 10. I'm moving on from this now for a cost POV and can take a look at PDF thing if required another time. The problem I'm having is it is highly confident in false answers, which it didn't seem to be last time. What I'd like to know is, if there has been any alterations to Claygents capabilities in the past couple of months? when I tried this kind of thing before, it worked much more accurately in one prompt, even paginating to the page and opening the file etc. : https://app.clay.com/workspaces/12893/tables/t_Um7NhF2Rawse/views/gv_8SYUjVt9rZj2
called Debtors amount in that table
No Claygent configuration should surpass or match your current scraping capabilities. To improve your results, consider using the 'Help Me' tool located at the bottom of the prompting window. Be sure to refine your prompt and provide as much context as possible for better precision.
.png){kind=logo}
