I don't use llms much but i saw see how you wrote the prompt and was wondering if you think that any of these tips could help improve the results.
https://blog.bytebytego.com/p/a-crash-course-on-using-ai-to-save
Today I built my first LLM-based bot! Based on nostr:nprofile1qy2hwumn8ghj7enjv4h8xtn4w3ux7tn0dejj7qguwaehxw309aex2mrp0yhxy6t5vdhkjmnsv9exktnrdakj7qg4waehxw309aex2mrp0yh82arcduhx7mn99uq3xamnwvaz7tmhda6zuat50phjummwv5hsqg8zenmu7gzq8ulj5jj4kv50ph3muwz43f747vmr9ld2alrjdswgavgyuvdz's news.utxo.one relay, it scrapes the relay every six hours and (attempts) to summarize the top 5 stories. Here's post number one:
I attempted to use huggingface to no avail, then tried to use mistral via ollama, which brought my computer to a grinding halt, so I decided to go with the Mistral API.
Source code: https://github.com/coracle-social/condenser
Discussion
No replies yet.