Hello everyone, I’ve created an agent for sector monitoring using multiple newsletters (stored in a Dust folder) and the web. However, the results are not satisfactory, likely because my prompt is too complex and covers multiple steps: 1️⃣ Collecting and filtering relevant information from emails. 2️⃣ Collecting and filtering relevant information from the web. 3️⃣ Cross-referencing data to remove duplicates and keep only key insights. 4️⃣ Structuring the information into a formatted message for my team. 5️⃣ Validating the final output: sources, coherence, and formatting. Would it be more effective to create separate agents for each step? If so, how can I chain these agents so they work sequentially on a shared document? Thanks in advance for your advice!
hey Benoit, i think Remi is using Dust for a very similar usecase and nope, it doesn’t involve multiple agents. give him some time (he was traveling) and i’m sure he could share his steps 😉
Hello Benoit PAYEN 👋 Here is the assistant I built for the AI newsletter in 🔒[private feed]:
I sign up to 20+ AI newsletter
I push the content to Dust folder using Make
I have an assistant that selected the best articles (using "Include data" tool)
Here are some screenshots: Let me know if it helps or if you need more specific guidance!
A huge thank you for your help kevin straszburger and the very detailed feedback Remi . Indeed, it's exactly the same issue, but I think I'm still asking for a bit too much. I'll refine things a bit more and get back to you with a more concrete experience on the points that are blocking me.
Awesome Benoit PAYEN! AI still can't do everything, understanding how what exactly you can use it with is often the tricky part. Good luck! Happy to help 😊
thanks for the kind words Benoit 🙏 and please, come back here to share what you learned/ your experience. which is exactly what 🔒[private feed] is for 😉
