FlowiseAI
[FEATURE] Agent output as JSON format (Sequential Agents)
I want to use the output of a Agent Node as JSON format
I am using Sequential Agents , I have an Agent Node with otput, for example, {'var1': 'value1', 'var2':'value2'}. I want to use this output value (node output) as JSON format for using this in input params of a Custom JS function!
I need something like Structured Output Perser that i can use in an Agent Node.
2 ways you can solve this:
1.) Specify in system message to ask LLM returns a JSON. and give few examples.
2.) Use LLM with JSON structured output:
I have tried both alternatives, but neither recognizes the output as JSON. The output is a JSON, but it cannot be passed as input to a Custom JS Function because it is not recognized as JSON, but rather as LLM output or Agent output.
yeah currently custom JS function is not possible to be connected to agentflows, will add that support
+1
Or please let us manually enter the desired JSON structure into the JSON Structured Output through a simple text box like in the Advanced Structured Output Parser node that maybe could then be parsed (no pun intended) by Flowise into the table-display that's currently in place. Adding the key-value pairs using the "add item" is very cumbersome when dealing with more then just few elements. Thank you! :)
I'd also like to have the tool agents and conversational chains being able to use the json output parser!
I would also like the tool agents and conversational chains to be compatible with the JSON output parser—it would be a game changer for me!
I’m thrilled about the potential of implementing JSON-formatted outputs for agent responses in Flowise! This enhancement is a game-changer, enabling seamless integration with custom JavaScript functions and empowering developers to craft more dynamic and efficient conversational agents.
By adopting this feature, Flowise will not only streamline workflows but also position itself at the forefront of AI development tools. Let’s make this happen and elevate Flowise to new heights! 🚀
Also interested in this feature. When it is easy to configure agents to output JSON it opens up a whole new world where this can be included in automation pipelines, used as API and so on.
I'd also like to have the tool agents being able to use the json output parser!
