Using local ChatGPT-like LLMs in OpenRefine for data wrangling

Support and Helpdesk
82

@Sunil_Natraj It is working as expected. Congratulations !!

image
image1847×1005 344 KB

1 Like
84

I studied it a bit. Here what I'm using : GitHub - harvard-lil/warc-gpt: WARC + AI - Experimental Retrieval Augmented Generation Pipeline for Web Archive Collections.

In fact I don't know how to include the retrieved results (document vectors) against a query (query vector). The following components are needed:

  • model: One of the models /api/models lists (required)
  • message: User prompt (required)
  • temperature: Defaults to 0.0
  • max_tokens: If provided, caps number of tokens that will be generated in response.
  • **search_results**: Array, output of /api/search.

In my request this 'search result' part is missing which is a product of another API call.

Can you guide me here?

Regards

85

Thanks for sharing details. WARC is a RAG implementation, there are 2 steps in it, first is to search for content based on user query & then generate an Inference using a LLM with the user prompt and search results. You can ry it out using CURL
1, Call the search endpoint - /api/search, pass the parameter message set the value as the user query e.g. "message": "what is Pragyan in the context of Chandrayaan-3?". Copy the response of the API.
2. Call the inference endpoint - api/complete, In the parameters message will have the user query same as what you passed in the search. For the param search_results pass set the value which is the response of the search API. Rest of the params are standard, you can ignore history param.

How do you plan to use the WARC RAG in OpenRefine flow?

86

@antonin_d @Martin The extension is ready for wider release. Shall i create a post for this and send it for review?

87

Thank you, @Sunil_Natraj. This is a great addition to the ecosystem! Yes, please go ahead and create a PR to add it to the extension page.

1 Like
88

Thanks everyone who help make this extension happen!

If you find bug or want to suggest new feature you can create issue directly in the new repository:

3 Likes
89

+1

Envoyé de mon iPhone

90

Quick preview of support for column update in the AI extraction extension. Feedback much appreciated.
Demo video

91

Is it available in version 0.1.1? I don't see update existing option rather it's showing New Column name. I downloaded the plugin from here - Releases · sunilnatraj/llm-extension · GitHub.

Best

Partha

92

Hi, The feature is still under development. Will notify when released.

93

Prompt history and reuse flow is nearing completion. Here is a quick demo of the same, do share any feedback.
Demo video

94

I am planning to make a release with the 2 additions this week. Do share any feedback on these

95

Great idea @Sunil_Natraj.

I haven't had the time to follow your work over the last few weeks and put the published extension to the test, but I can't wait to do so!

96

Thank you @archilecteur

97

@Sunil_Natraj Presently this plughin cannot store prompts (some prompts require lots of preview generation to get results in the desired format) and there is no trace of the prompt for later operations. The only option is to store prompts locally in a separate file.

Will it be possible in future to store a few prompts (like staring in History) in future release? It will help many of us really.

Thanks and regards

-partha

98

@psm this is something we are discussing with Sunil. You can join the conversation here:

99

@Martin Thanks, Joining asap.

Regards

1 Like
100

@Martin @psm Here is the video of the prompt history - reuse flow. Demo video

101

AI Extension V 0.1.2 is available for preview. Please try it out & share feedback.

AI Extension V0.1.2

102

Open refine couldnt start.
B.