hello again. I got this time api key error.

[klncgty]

when ı am creating knowledgebase, my codes here:

from giskard.rag import KnowledgeBase
import os
import giskard
from giskard.llm.client.mistral import MistralClient

api_key = os.environ["MISTRAL_API_KEY"]
model = "mistral-large-latest"


mc = MistralClient()
giskard.llm.set_default_client(mc)
knowledge_base=KnowledgeBase(df_base)

ı get this error:

OpenAIError: The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable

[mattbit]

Hi @klncgty, that's because by default we try to use OpenAI embeddings for the knowledge base. We currently support FastEmbed, OpenAI, and Bedrock. You can set the default embedding model similarly to how you set the LLM client:

from giskard.llm.embeddings import set_default_embedding

# If you use FastEmbed
from giskard.llm.embeddings.fastembed import FastEmbedEmbedding
from fastembed import TextEmbedding

text_embedding = TextEmbedding(model_name="BAAI/bge-small-en-v1.5")
embedding_model = FastEmbedEmbedding(text_embedding)

set_default_embedding(embedding_model)

We don't currently have a wrapper for Mistral embeddings, but should be easy to add (should be very similar to the OpenAI one).

[klncgty]

thank you ı am trying.

[klncgty]

hi mattbit, again your result doesnt work. here is my codes:

from giskard.rag import generate_testset, KnowledgeBase
from openai import OpenAI
from mistralai.client import MistralClient
from mistralai.models.chat_completion import ChatMessage
api_key = "…"
model = "mistral-large-latest"

knowledge_base = KnowledgeBase(df, llm_client=model)

from giskard.rag import testset_generation

testset = generate_testset(
    knowledge_base,
    num_questions=60,
    agent_description="A chatbot answering questions about cost of given materials with given measurements.",
)

And ı get error below :

---

2024-07-23 12:15:04,987 pid:20404 MainThread giskard.rag ERROR Encountered error in question generation: The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable. Skipping.
2024-07-23 12:15:04,999 pid:20404 MainThread giskard.rag ERROR The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable
Traceback (most recent call last):
File "c:\Users\cgtyk\AppData\Local\Programs\Python\Python311\Lib\site-packages\giskard\rag\question_generators\base.py", line 57, in generate_questions
yield self.generate_single_question(knowledge_base, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "c:\Users\cgtyk\AppData\Local\Programs\Python\Python311\Lib\site-packages\giskard\rag\question_generators\simple_questions.py", line 96, in generate_single_question
generated_qa = self._llm_complete(messages=messages)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "c:\Users\cgtyk\AppData\Local\Programs\Python\Python311\Lib\site-packages\giskard\rag\question_generators\base.py", line 42, in _llm_complete
out = self._llm_client.complete(
^^^^^^^^^^^^^^^^
File "c:\Users\cgtyk\AppData\Local\Programs\Python\Python311\Lib\site-packages\giskard\rag\question_generators\base.py", line 37, in llm_client
return get_default_client()
^^^^^^^^^^^^^^^^^^^^
File "c:\Users\cgtyk\AppData\Local\Programs\Python\Python311\Lib\site-packages\giskard\llm\client_init.py", line 69, in get_default_client
client = AzureOpenAI() if default_llm_api == "azure" else OpenAI()
^^^^^^^^
File "c:\Users\cgtyk\AppData\Local\Programs\Python\Python311\Lib\site-packages\openai_client.py", line 104, in init
raise OpenAIError(
openai.OpenAIError: The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable
2024-07-23 12:15:05,016 pid:20404 MainThread giskard.rag ERROR Encountered error in question generation: The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable. Skipping.
2024-07-23 12:15:05,029 pid:20404 MainThread giskard.rag ERROR The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable
Traceback (most recent call last):
File "c:\Users\cgtyk\AppData\Local\Programs\Python\Python311\Lib\site-packages\giskard\rag\question_generators\base.py", line 57, in generate_questions
...
^^^^^^^^
File "c:\Users\cgtyk\AppData\Local\Programs\Python\Python311\Lib\site-packages\openai_client.py", line 104, in init
raise OpenAIError(
openai.OpenAIError: The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable

[mattbit]

Yeah so your code has some issues. First, the LLM client you pass to the knowledge base should be an instance of giskard LLMClient class. Second, you should also set your embedding model.

I recommend to set both globally so that you don't have to pass the over every time:

# LLM client
from giskard.llm import set_default_client
from giskard.llm.mistral import MistralClient  # note that this is the Giskard wrapper

llm_client = MistralClient("mistral-large-latest")  # assuming you set your Mistral key in the env

set_default_client(llm_client)

# Embeddings
from giskard.llm.embeddings import set_default_embedding
from giskard.llm.embeddings.fastembed import FastEmbedEmbedding
from fastembed import TextEmbedding

text_embedding = TextEmbedding(model_name="BAAI/bge-small-en-v1.5")
embedding_model = FastEmbedEmbedding(text_embedding)

set_default_embedding(embedding_model)

# Now you can run the rest, for example

from giskard.rag import testset_generation, KnowledgeBase

knowledge_base = KnowledgeBase(df)

testset = generate_testset(
    knowledge_base,
    num_questions=60,
    agent_description="A chatbot answering questions about cost of given materials with given measurements.",
)