Closed cherry0411 closed 1 month ago
I guess you want to parse some json string in the start node? If so, you can use text type which supports long context. Then you can use a code node to parse these parameters to meet your requirements. Does this solves your problem?
No, it doesn't. Sorry for the confusion. I need function calling to extract parameters and values from text through LLM, which exactly Parameter Extractor node does. but currently parameters are fixed values we can only setup in the workflow editor, that's my problem, I need to dynamically setup parameters to be extract, on the runtime.
So this problem should be you need an agent during the workflow and it can call third-party services by it's own decision?
Emmm... I don't think agent in workflow can solve this problem' cos tools in agent still have fixed parameters.
Simplest way, I think, is supporting variables in Parameter Extractor's parameter.
Self Checks
1. Is this request related to a challenge you're experiencing? Tell me about your story.
Hello everyone, I would like to dynamically parse parameters from text, such as product model, brand, and size. Different products have different parameters, which may be passed in from the starting node or obtained through the HTTP node. Current parameter extractor can only support fixed parameters and cannot meet my needs. For example, I can pass in the following parameters at the start node: {"product_fields": [{"key": "field_1", "name": "brand", "description": "brand of the product", "type": "enum", "range": ["Nike", "Adidas", "Puma"]}, {"key": "field_2", "name": "size", "description": "size of the shoes of clothes", "type": "number"}}, and then use these parameters directly in the parameter extractor
2. Additional context or comments
No response
3. Can you help us with this feature?