Skip to content

Steps.md

Latest commit

 

History

History
41 lines (39 loc) · 2.37 KB

Steps.md

File metadata and controls

41 lines (39 loc) · 2.37 KB

One step at a time

  1. Take example in langgraph course , try to build the same functionality of langgraph but with your code base - #Done

    context_history(agentic memories) - should only contain the messages in the list of dict that are feed to ai models (dict with role and content). This will be per agent - #Done graph_state - comman place to acess data between boxes - #Done meta_history - all the data results(output + metadata) we get from the each box , also should contain the index of messsage length with key as unique identifier - #Done openai support - write inference fucntions - #Done Add conditional edges -- added two types of conditional edges, one that only runs next node if condtion is met , second is branching edge which will take to node/nodes based on the condition - #Done claude support - extend the inference functions from openai to claude - #DoneHopefully 1.1. multi node input - #Done and fixed most of bugs of update1 by default all the edges are required and if user specifies the edge as not required like optional , then it will not trigger the next node , only required edge will trigger the next node Bydefault we need make conditional edges to next node as not required edges , as they only get riggered in some cases - if i donot add to required list , then its fine i guess

  2. Graph Communincation funtionality - #Done

  3. Interept and Human in loop - #Done Human as a tool can be done , but this means llm is decided to call the human support Human approval before function call - yet to be done

  4. GUI first version Create Basic UI setup - #Done Add the agent config json system to be able to create a graph from graph - ##Done

  5. Replicate real world examples graph_websearch_agent - Basic Done While building i got a feeling that goal i want to achieve is create a sharable agentic flow space that anyone improve on if it is on sharable (public/private) space SuperMemory Basic Rag -- llama index maybe Mem0 Copali version TexttoSQL TelcoGPT Claude and Ollama support Code Change Reviewer Stock market helper

  6. GUI second version code for deploy graph as api endpoint option function calls to send and recieve from frontend Functional UI

  7. multi workflow versioning

  8. Add streaming

  10. RAG examples

  11. parellel exectution , Asyc funcionality