How to build automation workflow with function calling on local LLM?

[HomunMage]

This topic combine several questions.

First, is that there is no one general chat model. such gpt and ollama have different interface. This cannot create workflow template.

Second, such "Tool Agent" can use Read, Write File and function calling. But Tool Agent cannot output to prompt template.
Another side has same problem, "LLM Chain" cannot use tool or function calling.

Similar to first question, that the items "only" can use some spesific item. exhausted to find what can link.

Third question, this repo seems powerful, but just look like powerfull. It cannot do something more complex than any exist templates.
I see many questions such that cannot RAG multiple file, cannot combine agents.......

For my example, how to build automation workflow with function calling?

Such that X chat model with a template such Marketing search web with selenium, and next do analysis, write to TA.md.
Y chat model with a role such PM that read TA.md and design product and write to a file such Design.md.
And then another chat model Z with role programmer that write imp.py and run test.py.
and for different stage such model Z use gpt4o such PM model Y use ollama local.
(like metaGPT or CrewAI)

Even any one line of this workflow sub step cannot use flowise to acheieve.

[HomunMage]

Similar problem:
https://stackoverflow.com/questions/78537142/flowise-chain-output-of-tool-agent-to-an-llm-chain