mistralai/cookbook

cookbook's Issues

Basic RAG Guide typo/mismatch

While working through the basic RAG guide I discovered a mismatch between the code on Colab and the code on the website.

Website:

Colab:

Flactuating and Corrupted Response

Sometimes proper response is generated with answer and sometimes following kind of error response is generated,Thoroughly inspected the code,flactuating response is what iam facing.

llama_index.core.output_parsers.base.OutputParserException: Got invalid JSON object. Error: Expecting ',' delimiter: line 26 column 10 (char 885) while parsing a flow sequence
in "", line 2, column 14:
"items": [
^
expected ',' or ']', but got ''in "", line 26, column 10:
} ^. Got JSON string: { "items": [
{ "sub_question": "What is the definition of folliculitis?",
"tool_name": "pdf1" }, { "subpor": "Which patients are more likely to develop gram-negative folliculitis?",
"tool_name": "pdf1" }, { "sub_question": "What are the common symptoms of folliculitis?",
"tool_name": "pdf1" }, {
"sub_question": "What are the common causes of folliculitis?",
"tool_name": "pdf1"
},
{
"sub_question": "What are the recommended treatments for folliculitis?",
"tool_name": "pdf1"
},
{
"sub_question": "What are the potential complications of folliculitis?",
"tool_name": "pdf1"
}

No CAMEL-AI package in PIP

Trying the mistralai cookbook for knowledge graph, but can't install it in pip. CAn hou hihlight the procedure to download/install the dependencies please?

https://github.com/mistralai/cookbook/blob/main/third_party/CAMEL_AI/camel_graph_rag.ipynb

Kind regards

Add license

Hi @sophiamyang et al,

that's a fantastic repository! I was just wondering under which conditions one could reuse the code ansd materials provided here. Would you mind adding a license file? If you're new to licensing and/or wonder which license to use, you can read more in this blob post: https://focalplane.biologists.com/2023/05/06/if-you-license-it-itll-be-harder-to-steal-it-why-we-should-license-our-work/

Thanks!

Best,
Robert

[Question] Restricting prior internal knowledge for RAG

Hello Everyone,

I am building RAG application with model Mistral-7B-Instruct-v0.2. It works well for questions related to the knowledge content. But I want to restrict the LLM from answering to any out of scope questions. This is the current prompt I am using:

You are an assistant to users who have queries about {topic} .
Only USE the following pieces of contexts under **Contexts** to answer. Do not use any external knowledge or information. If the answer cannot be determined from the context, respond with "I don't have enough information to answer that"

**Contexts**:

{contexts}

What is the effective way to stop LLM from answering questions if it is out of scope?.
Can we prompt like above and restrict or is there a better way?

Thanks in Advance

HTTPStatusError during Querying step in Ollama/Mistral/LlamaIndex Cookbook

I am trying to run Ollama <> Mistral <> LlamaIndex Cookbook's notebook.

The Querying part of the notebook raises a HTTPStatusError:

In [3]: from llama_index.core.llms import ChatMessage
   ...:
   ...: messages = [
   ...:     ChatMessage(role="system", content="You are a helpful assistant."),
   ...:     ChatMessage(role="user", content="What is the capital city of France?"),
   ...: ]
   ...: response = llm.chat(messages)
---------------------------------------------------------------------------
HTTPStatusError                           Traceback (most recent call last)
Cell In[3], line 7
      1 from llama_index.core.llms import ChatMessage
      3 messages = [
      4     ChatMessage(role="system", content="You are a helpful assistant."),
      5     ChatMessage(role="user", content="What is the capital city of France?"),
      6 ]
----> 7 response = llm.chat(messages)

File ~/.py-venv/lib/python3.11/site-packages/llama_index/core/instrumentation/dispatcher.py:230, in Dispatcher.span.<locals>.wrapper(func, instance, args, kwargs)
    226 self.span_enter(
    227     id_=id_, bound_args=bound_args, instance=instance, parent_id=parent_id
    228 )
    229 try:
--> 230     result = func(*args, **kwargs)
    231 except BaseException as e:
    232     self.event(SpanDropEvent(span_id=id_, err_str=str(e)))

File ~/.py-venv/lib/python3.11/site-packages/llama_index/core/llms/callbacks.py:172, in llm_chat_callback.<locals>.wrap.<locals>.wrapped_llm_chat(_self, messages, **kwargs)
    163 event_id = callback_manager.on_event_start(
    164     CBEventType.LLM,
    165     payload={
   (...)
    169     },
    170 )
    171 try:
--> 172     f_return_val = f(_self, messages, **kwargs)
    173 except BaseException as e:
    174     callback_manager.on_event_end(
    175         CBEventType.LLM,
    176         payload={EventPayload.EXCEPTION: e},
    177         event_id=event_id,
    178     )

File ~/.py-venv/lib/python3.11/site-packages/llama_index/llms/ollama/base.py:135, in Ollama.chat(self, messages, **kwargs)
    130 with httpx.Client(timeout=Timeout(self.request_timeout)) as client:
    131     response = client.post(
    132         url=f"{self.base_url}/api/chat",
    133         json=payload,
    134     )
--> 135     response.raise_for_status()
    136     raw = response.json()
    137     message = raw["message"]

File ~/.py-venv/lib/python3.11/site-packages/httpx/_models.py:761, in Response.raise_for_status(self)
    759 error_type = error_types.get(status_class, "Invalid status code")
    760 message = message.format(self, error_type=error_type)
--> 761 raise HTTPStatusError(message, request=request, response=self)

HTTPStatusError: Client error '404 Not Found' for url 'http://localhost:11434/api/chat'
For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/404

I have installed all the needed packages to run the notebook, from Ollama and pip.
In addition, the Ollama server is running and I can query it with the example curl code from the documentation:

 $ curl -X POST http://localhost:11434/api/generate -d '{
  "model": "mistral",
  "prompt":"Here is a story about llamas eating grass"
 }'
{"model":"mistral","created_at":"2024-06-18T08:03:01.76453978Z","response":" Title","done":false}
{"model":"mistral","created_at":"2024-06-18T08:03:02.057336093Z","response":":","done":false}
{"model":"mistral","created_at":"2024-06-18T08:03:02.346361586Z","response":" The","done":false}
{"model":"mistral","created_at":"2024-06-18T08:03:02.55772899Z","response":" L","done":false}
{"model":"mistral","created_at":"2024-06-18T08:03:02.670552331Z","response":"l","done":false}
{"model":"mistral","created_at":"2024-06-18T08:03:02.784986299Z","response":"ama","done":false}
{"model":"mistral","created_at":"2024-06-18T08:03:02.891749852Z","response":" G","done":false}
{"model":"mistral","created_at":"2024-06-18T08:03:02.995237782Z","response":"raz","done":false}
{"model":"mistral","created_at":"2024-06-18T08:03:03.102841835Z","response":"ers","done":false}
{"model":"mistral","created_at":"2024-06-18T08:03:03.229874023Z","response":" of","done":false}
{"model":"mistral","created_at":"2024-06-18T08:03:03.334951599Z","response":" the","done":false}
^C

Looks like everything is working well except the Querying done in the notebook. In fact, the server seems to work without failure using requests:

In [13]: payload = {"model":"mistral", "prompt": "is the sky blue ?", "stream": False}
In [14]: r = requests.post('http://localhost:11434/api/generate', json=payload)
In [15]: r
Out[15]: <Response [200]>

is it possible lower the codestral to 7B or small model?

is it possible lower the codestral to 7B or small model?
my laptop not able to running codestral model with 22B size.

Still confused about dataset format used for fine-tuning model using APIs

I've been diving into fine-tuning for this virtual hackathon but I am confused about the format used to fine-tune the model using its apis.Below is the format used in the 'capabilities' section under 'fine-tuning' in the docs.
{
"messages": [
{
"role": "user",
"content": "User interaction n°1 contained in document n°2"
},
{
"role": "assistant",
"content": "Bot interaction n°1 contained in document n°2"
},
{
"role": "user",
"content": "User interaction n°2 contained in document n°1"
},
{
"role": "assistant",
"content": "Bot interaction n°2 contained in document n°1"
}
]
}

But under the 'Guide' section 'fine-tuning' a different format is used and later converted into binary and fed to the api.here is the format:
{
"prompt":"Are there any specific neighborhoods in the city that are particularly well-served by public transportation?",
"prompt_id":"ea9a0c38af7a8f9602c40a2e666283e8bb5101adec34421ec4a7df90bb081666",
"messages":[
{
"content":"Are there any specific neighborhoods in the city that are particularly well-served by public transportation?",
"role":"user"
}] , enlighten me!!