I am trying to run the app locally. However, I don't have access to the files on Azure. The error message is:
File ".../climate-question-answering/app.py", line 53, in <module> account_key = os.environ["BLOB_ACCOUNT_KEY"] ~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^ File "<frozen os>", line 679, in __getitem__ KeyError: 'BLOB_ACCOUNT_KEY'
Can you provide information on how to obtain the credentials (account key, name, url) for the files on the server? Or alternatively, access to them and a way to store and use them locally.
Thanks a lot for the support!
Hello all ! Thanks for the interest !
I am working on a reproducible local setup to allow for contributions.
Meanwhile there are three components you can change
- You can comment everything about logging (on Azure Blob Storage), this part is logging the question and answer provided for debugging purposes
- You need to change the AzureChatOpenAI to your LLM config (OpenAI or else) using Langchain and provide your own API key. I'm also working on using local open source LLM instead of OpenAI to ease the process
- The hardest part is the access to the Pinecone vector store for IPCC chunk embeddings, you can find the dataset on hugging face https://huggingface.co/datasets/Ekimetrics/ipcc-ar6. It's not well documented, but you can use it to recreate the embeddings and store in a local Chroma or Qdrant VectorStore.
I am planning to document and make it easier to work on it locally in the next weeks. Tell me in the meantime if you have any questions with the above precisions