llama-hub/llama_hub/llama_packs/fuzzy

At a glance

The community member is puzzled about the need for an OpenAI API key for the FuzzyCitationEnginePack. Another community member suggests that the user can use any language model (LLM) and configure the index/query engine accordingly. However, the original poster mentions that they want to run without an LLM and are mostly interested in evaluating the citations, but they are encountering an OpenAI authentication error. The comments indicate that an LLM is required to evaluate the citations, as they are generated using fuzzy matching based on the LLM's output. Another community member suggests that if the user just wants the retrieved text, they can use a retriever without the need for an LLM, but they would still require an embedding mode.

Useful resources

MMotzWanted

Hello guys! I'm a bit puzzled over the need for an OpenAi API key for the FuzzyCitationEnginePack: https://github.com/run-llama/llama-hub/blob/main/llama_hub/llama_packs/fuzzy_citation/fuzzy_citation_example.ipynb

Can anyone clarify what I'm missing out on?

4 comments

LLogan M

You can use any llm, just configure your index/query engine to use whichever llm you want

MMotzWanted

@Logan M what if want to run without a LLM? Mostly interested in evaluating the citations, and not the LLM response. Howver, I get an openai.AuthenticationError imidiately:

Plain Text

openai.AuthenticationError: Error code: 401 - {'error': {'message': 'Incorrect API key provided: 3a03c59b********************1c2d. You can find your API key at https://platform.openai.com/account/api-keys.', 'type': 'invalid_request_error', 'param': None, 'code': 'invalid_api_key'}}

LLogan M

You need an LLM to evaluate citations? The citations are generated using fuzzy matching based on what the LLM wrote vs the retrieved text

LLogan M

if you just want the retrieved text, you don't need this -- just run a retriever

Plain Text

retriever = index.as_retriever(similarity_top_k=2)
nodes = retriever.retrieve("query")

You'll still need an embed mode for that though

Add a reply

Find answers from the community

llama-hub/llama_hub/llama_packs/fuzzy_ci...