Help for code node and llm node

[kavyakalivarapu-lab]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:)
• Please do not modify this template :) and fill in all the required fields.

1. Is this request related to a challenge you're experiencing? Tell me about your story.

i am trying to build a simple workflow. on uploading a file sin start node, doc extractor node extracts the text and that text is summarized by llm node. upto this its working fine. My doubt is how to access the llm nodes output in code node to write it in python code. how to access the llm nodes output variable in code node python code? i am running dify on localhost

2. Additional context or comments

No response

[dosubot[bot]]

Hi @kavyakalivarapu-lab! I'm Dosu and I’m helping the dify team.

To access the output of an LLM node in a code node, you need to configure the code node's input variables to reference the LLM node's output. In Dify, each node's outputs are stored in a shared variable pool, and you map them to your code node's function arguments using the variable selector format: [llm_node_id, "text"] (assuming you want the summary text).

Here's how you can do it:

1. In the workflow editor, add an input variable to your code node. Set its value selector to [llm_node_id, "text"], where llm_node_id is the ID of your LLM node (you can find this in the workflow editor).
2. In your code node's Python function, declare a parameter with the same name as your input variable. For example:

def main(summary_text: str) -> dict:
    # Now you can use summary_text, which contains the LLM node's output
    print("LLM summary:", summary_text)
    return {"result": summary_text}

If you want to access other fields (like usage info), use [llm_node_id, "usage"] and add a corresponding parameter.

Common pitfalls: Make sure the variable name and selector match exactly, and that the LLM node output is available (the LLM node must run before the code node). If you see "Output result is missing," double-check your variable mapping and output schema in the code node config. For more details and troubleshooting, see this discussion and this PR.

Let me know if you need a more specific example based on your node IDs!

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[kavyakalivarapu-lab]

hello @dosu . thank you for your response. i have written the code that you gave, still the problem is not solved. i am attaching the screenshots of my llm node and code node. please verify and guide me.

[dosubot[bot]]

Based on your screenshots and the error, here’s what to check:

1. In your code node, the output dictionary key must match the output schema exactly. If your output schema expects "result", but your code returns {"llm_summary": summary_text}, you’ll get the "Output result is missing" error. Change your return statement to return {"result": summary_text} and make sure the output variable name in the workflow editor is also "result" source.

2. Double-check the input mapping for your code node in the workflow editor. The input variable (e.g., summary_text) should be mapped to [llm_node_id, "text"], where llm_node_id is the actual ID of your LLM node. The argument name in your Python function must match the input variable name exactly source.

3. If you recently duplicated or edited your workflow, variable mappings can break. Remap the input variable in the UI to ensure it’s correctly connected to the LLM node’s output source.

4. If the LLM node output is blank, check if the stream parameter is set to true in your LLM node config. Setting stream to false can resolve missing output issues source.

Let me know if you need help finding the node IDs or adjusting the schema!

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[kavyakalivarapu-lab]

wow, that worked for me. thank you for the help @dosu

[dosubot[bot]]

You're welcome! Glad I could help.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}