langgraph/tutorials/storm/storm/

[giscus[bot]]

langgraph/tutorials/storm/storm/

Build language agents as graphs

https://langchain-ai.github.io/langgraph/tutorials/storm/storm/

[Asdafers]

Hi,
You wouldn't happen to have an example of reworking this to be able to work with Ollama?

[jondoescoding]

Replace every instance of OpenAI with Ollama instead

[hinthornw]

I think the prompts at least would need updating, and the function calling integrations aren't built out yet, so would need your own parser - less turnkey than we'd like!

[SweetSourPeter]

Love it!!!

[xeniode]

Did anyone reading this get this to work?

[arisylafeta]

I did, some slight modifications to make it work with gpt-4o-mini

[JosephArcila]

I'm trying to run the code, I tried with the duckduck implementation and with the tavily implementation but always get an empty final_state["references"]
I appreciate if someone could help me

[elliotvaucher]

Thanks for this amazing work. The overall architecture of the software, the way it was designed and implemented is fantastic. Great work. I still find it quite funny though that the result is... okay (not many references, not so many paragraphs, etc.).

[elliotvaucher]

Note : if I understand correctly and the result of the process is the article written at the end of the tutorial, of course.

[SharoMonk]

If and when you encountered errors related to pattern mismatch in class Editor, try to use "^.{1,64}$" to pass all characters.

[SharoMonk]

You can test it with this:

from pydantic import ValidationError

try:
Editor(
name="Mia González",
affiliation="UNILAG",
role="Editor_1",
description="Wiki_editor",
)
except ValidationError as exc:
print(repr(exc.errors()[0]))

[NaMoCv]

Why does return have state? And it should be easy to make mistakes.
为什么后面return要有state？而且这么应该很容易造成错误吧。。

[itsuresh248]

Tried executing this with ChatOllama using the Qwen 2.5 7B model to make it run locally on my Mac. Got the following error:

tool_call = queries["raw"].tool_calls[0] line produces the error.

I set the method to "json-schema" in the with_structured_output method of the LLM. The result is that there are no tool_calls in the "queries" object (screenshot below).

Please do let me know on how to fix this issue to enable local execution of the code with ollama.

[itsuresh248]

Claude 3.7 Sonnet helped me fix this.

FIX: