Oct 2023, https://blog.sbensu.com/
For more like this one, see Friction logs
Setup: I want make Anki cards from chat sessions with ChatGPT and upload them to GDrive
https://github.com/FlowiseAI/Flowise
Takeaways
- Clearly for developers
- Need to install it from the command line with
npm
and call it with npx
- It felt like it took 10 minutes to install all these dependencies.
- Most of the integrations have names like “Embeddings” that don’t really explain what they are for
- They don’t have triggers! You can only trigger the scripts by doing an HTTP request to it
- UI
- The UI tells you the types of the arrows when you are connecting building blocks
Document
is one type
BaseLanguageModel
is a different type. This is different because “the configuration for a base LLM” is thought of as a data type. This is not something that would be easy to do with Zapier but Flowise allows this higher order configuration.
- You configure the steps inline and not on a sidebar
- Makes it easier to scan and see everything that the script is doing
- But it takes a lot of extra space
- When you are trying to connect elements together, it doesn’t tell you why some pieces don’t fit (they are different types) and you are left wondering.
- Market
- This seems to be a product for no-one: it clearly assumes that you are a developer but then has absolutely no escape-hatches to let you add code, if statements, or anything else that you might do with a Python script.
- And it doesn’t seem that much easier to use than Python
Chatflows
They call scripts or recipes “chatflows”. Even if they are not really about chat?
The editor matches the Zapier editor with only superficial differences:
- Zapier guides you to have all the steps go from top to bottom
- Zapier lets you edit the steps in a sidebar, Flowise puts all the options in the graph elements