Finding the most relevant chunks of content for a given prompt and generation is a tricky task. That’s why we use OpenAI’s embedding model default to build a powerful semantic index of your content, stored securely inside Promptitude.
To start enjoying the advantages that Content Storage brings to contextualize your prompts, it is not mandatory to perform the configuration, Promptitude previously set default parameters optimized to give you good results.
Even if you wish to learn and experiment by changing these generic settings, you are free to do so. Here, we explain how and where:
1️⃣ Sign up and log in to Promptitude: There's a free plan available, so you can try it out without any cost.
2️⃣ Go to the Content Storage area: This is where you can upload your own unique content.
3️⃣ Configure settings: Click Go to Settings and confirm the following two settings:
Maximum Chunk size: This determines how your content will be chopped up into smaller chunks. We recommend starting with a chunk size of 150 words.
Chunk overlap: Specifies how much overlap there should be between the chunks. This is often useful to make sure that the text is not split strangely.
4️⃣ Save your settings: Click Save to confirm your settings and complete the setup process.
Why is important the Content Storage Settings?
After uploading, your content gets sliced up into small chunks. Those small chunks help us find and add only the most relevant ones to your prompt. Why?
Today’s GPT prompts only accept a few thousand words, so we can’t just add all your content. We recommend starting with chunks of 150 words and an overlap of 20%. (current parameters by default)