GuidesCookbooksIntegration Mistral Sdk
This is a Jupyter notebook

Cookbook: Mistral SDK Integration (Python)

This cookbook provides step-by-step examples of integrating Langfuse with the Mistral AI SDK (v1) in Python. By following these examples, you’ll learn how to seamlessly log and trace interactions with Mistral’s language models, enhancing the transparency, debuggability, and performance monitoring of your AI-driven applications.


Note: Langfuse is also natively integrated with LangChain, LlamaIndex, LiteLLM, and other frameworks. If you use one of them, any use of Mistral models is instrumented right away.


Overview

In this notebook, we will explore various use cases where Langfuse can be integrated with Mistral AI SDK, including:

  • Basic LLM Calls: Learn how to wrap standard Mistral model interactions with Langfuse’s @observe decorator for comprehensive logging.
  • Chained Function Calls: See how to manage and observe complex workflows where multiple model interactions are linked together to produce a final result.
  • Async and Streaming Support: Discover how to use Langfuse with asynchronous and streaming responses from Mistral models, ensuring that real-time and concurrent interactions are fully traceable.
  • Function Calling: Understand how to implement and observe external tool integrations with Mistral, allowing the model to interact with custom functions and APIs.

For more detailed guidance on the Mistral SDK or the @observe decorator from Langfuse, please refer to the Mistral SDK repo and the Langfuse Documentation.

Setup

!pip install mistralai langfuse
import os
 
# get keys for your project from https://cloud.langfuse.com
os.environ["LANGFUSE_PUBLIC_KEY"] = "pk-lf-xxx"
os.environ["LANGFUSE_SECRET_KEY"] = "sk-lf-xxx"
os.environ["LANGFUSE_HOST"] = "https://cloud.langfuse.com" # 🇪🇺 EU region
# os.environ["LANGFUSE_HOST"] = "https://us.cloud.langfuse.com" # 🇺🇸 US region
 
# Your Mistral key
os.environ["MISTRAL_API_KEY"] = "xxx"
from mistralai import Mistral
 
# Initialize Mistral client
mistral_client = Mistral(api_key=os.environ["MISTRAL_API_KEY"])

Examples

Completions

We are integrating the Mistral AI SDK with Langfuse using the @observe decorator, which is crucial for logging and tracing interactions with large language models (LLMs). The @observe(as_type=“generation”) decorator specifically logs LLM interactions, capturing inputs, outputs, and model parameters. The resulting mistral_completion method can then be used across your project.

from langfuse.decorators import langfuse_context, observe
 
# Function to handle Mistral completion calls, wrapped with @observe to log the LLM interaction
@observe(as_type="generation")
def mistral_completion(**kwargs):
  # Clone kwargs to avoid modifying the original input
  kwargs_clone = kwargs.copy()
 
  # Extract relevant parameters from kwargs
  input = kwargs_clone.pop('messages', None)
  model = kwargs_clone.pop('model', None)
  min_tokens = kwargs_clone.pop('min_tokens', None)
  max_tokens = kwargs_clone.pop('max_tokens', None)
  temperature = kwargs_clone.pop('temperature', None)
  top_p = kwargs_clone.pop('top_p', None)
 
  # Filter and prepare model parameters for logging
  model_parameters = {
        "maxTokens": max_tokens,
        "minTokens": min_tokens,
        "temperature": temperature,
        "top_p": top_p
    }
  model_parameters = {k: v for k, v in model_parameters.items() if v is not None}
 
  # Log the input and model parameters before calling the LLM
  langfuse_context.update_current_observation(
      input=input,
      model=model,
      model_parameters=model_parameters,
      metadata=kwargs_clone,
 
  )
 
  # Call the Mistral model to generate a response
  res = mistral_client.chat.complete(**kwargs)
 
  # Log the usage details and output content after the LLM call
  langfuse_context.update_current_observation(
      usage={
          "input": res.usage.prompt_tokens,
          "output": res.usage.completion_tokens
      },
      output=res.choices[0].message.content
  )
 
  # Return the model's response object
  return res

Optionally, other functions (api handlers, retrieval functions, …) can be also decorated.

Simple Example

In the following example, we also added the decorator to the top-level function find_best_painter_from. This function calls the mistral_completion function, which is decorated with @observe(as_type=“generation”). This hierarchical setup hels to trace more complex applications which involve multiple LLM calls and other non-llm methods which are decorated with @observe.

You can use langfuse_context.update_current_observation or langfuse_context.update_current_trace to add additional details such as input, output, and model parameters to the trace.

@observe()
def find_best_painter_from(country="France"):
  response = mistral_completion(
      model="mistral-small-latest",
      max_tokens=1024,
      temperature=0.4,
      messages=[
        {
            "content": "Who is the best painter from {country}? Answer in one short sentence.".format(country=country),
            "role": "user",
        },
      ]
    )
  return response.choices[0].message.content
 
find_best_painter_from()
'Claude Monet, renowned for his role as a founder of French Impressionist painting, is often considered one of the best painters from France.'

Example trace in Langfuse: https://cloud.langfuse.com/project/cloramnkj0002jz088vzn1ja4/traces/836a9585-cfcc-47f7-881f-85ebdd9f601b

Chained Completions

This example demonstrates chaining multiple LLM calls using the @observe decorator. The first call identifies the best painter from a specified country, and the second call uses that painter’s name to find their most famous painting. Both interactions are logged by Langfuse as we use the wrapped mistral_completion method created above, ensuring full traceability across the chained requests.

@observe()
def find_best_painting_from(country="France"):
  response = mistral_completion(
      model="mistral-small-latest",
      max_tokens=1024,
      temperature=0.1,
      messages=[
        {
            "content": "Who is the best painter from {country}? Only provide the name.".format(country=country),
            "role": "user",
        },
      ]
    )
  painter_name = response.choices[0].message.content
  return mistral_completion(
      model="mistral-small-latest",
      max_tokens=1024,
      messages=[
        {
            "content": "What is the most famous painting of {painter_name}? Answer in one short sentence.".format(painter_name=painter_name),
            "role": "user",
        },
      ]
    )
 
find_best_painting_from("Germany")
ChatCompletionResponse(id='8bb8512749fd4ddf88720aec0021378c', object='chat.completion', model='mistral-small-latest', usage=UsageInfo(prompt_tokens=23, completion_tokens=23, total_tokens=46), created=1726597735, choices=[ChatCompletionChoice(index=0, message=AssistantMessage(content='Albrecht Dürer\'s most famous painting is "Self-Portrait at Twenty-Eight."', tool_calls=None, prefix=False, role='assistant'), finish_reason='stop')])

Example trace in Langfuse: https://cloud.langfuse.com/project/cloramnkj0002jz088vzn1ja4/traces/a3360c6f-24ad-455c-aae7-eb9d5c6f5dac

Streaming Completions

The following example demonstrates how to handle streaming responses from the Mistral model using the @observe(as_type=“generation”) decorator. The process is similar to the Completion example but includes handling streamed data in real-time.

Just like in the previous example, we wrap the streaming function with the @observe decorator to capture the input, model parameters, and usage details. Additionally, the function processes the streamed output incrementally, updating the Langfuse context as each chunk is received.

# Wrap streaming function with decorator
@observe(as_type="generation")
def stream_mistral_completion(**kwargs):
    kwargs_clone = kwargs.copy()
    input = kwargs_clone.pop('messages', None)
    model = kwargs_clone.pop('model', None)
    min_tokens = kwargs_clone.pop('min_tokens', None)
    max_tokens = kwargs_clone.pop('max_tokens', None)
    temperature = kwargs_clone.pop('temperature', None)
    top_p = kwargs_clone.pop('top_p', None)
 
    model_parameters = {
        "maxTokens": max_tokens,
        "minTokens": min_tokens,
        "temperature": temperature,
        "top_p": top_p
    }
    model_parameters = {k: v for k, v in model_parameters.items() if v is not None}
 
    langfuse_context.update_current_observation(
        input=input,
        model=model,
        model_parameters=model_parameters,
        metadata=kwargs_clone,
    )
 
    res = mistral_client.chat.stream(**kwargs)
    final_response = ""
    for chunk in res:
        content = chunk.data.choices[0].delta.content
        final_response += content
        yield content
 
        if chunk.data.choices[0].finish_reason == "stop":
            langfuse_context.update_current_observation(
                usage={
                    "input": chunk.data.usage.prompt_tokens,
                    "output": chunk.data.usage.completion_tokens
                },
                output=final_response
            )
            break
 
# Use stream_mistral_completion as you'd usually use the SDK
@observe()
def stream_find_best_five_painter_from(country="France"):
    response_chunks = stream_mistral_completion(
        model="mistral-small-latest",
        max_tokens=1024,
        messages=[
            {
                "content": "Who are the best five painter from {country}? Answer in one short sentence.".format(country=country),
                "role": "user",
            },
        ]
    )
    final_response = ""
    for chunk in response_chunks:
        final_response += chunk
        # You can also do something with each chunk here if needed
        print(chunk)
 
    return final_response
 
stream_find_best_five_painter_from("Spain")
The
 best
 five
 pain
ters
 from
 Spain
 are
 Diego
 Vel
áz
que
z
,
 Francisco
 G
oya
,
 P
ablo
 Pic
asso
,
 Salvador
 Dal
í
,
 and
 Joan
 Mir
ó
.






'The best five painters from Spain are Diego Velázquez, Francisco Goya, Pablo Picasso, Salvador Dalí, and Joan Miró.'

Example trace in Langfuse: https://cloud.langfuse.com/project/cloramnkj0002jz088vzn1ja4/traces/75a2a4fe-088d-4134-9797-ba9c21be01b2

Async Completion

This example showcases the use of the @observe decorator in an asynchronous context. It wraps an async function that interacts with the Mistral model, ensuring that both the request and the response are logged by Langfuse. The async function allows for non-blocking LLM calls, making it suitable for applications that require concurrency while maintaining full observability of the interactions.

# Wrap async function with decorator
@observe(as_type="generation")
async def async_mistral_completion(**kwargs):
  kwargs_clone = kwargs.copy()
  input = kwargs_clone.pop('messages', None)
  model = kwargs_clone.pop('model', None)
  min_tokens = kwargs_clone.pop('min_tokens', None)
  max_tokens = kwargs_clone.pop('max_tokens', None)
  temperature = kwargs_clone.pop('temperature', None)
  top_p = kwargs_clone.pop('top_p', None)
 
  model_parameters = {
        "maxTokens": max_tokens,
        "minTokens": min_tokens,
        "temperature": temperature,
        "top_p": top_p
    }
  model_parameters = {k: v for k, v in model_parameters.items() if v is not None}
 
  langfuse_context.update_current_observation(
      input=input,
      model=model,
      model_parameters=model_parameters,
      metadata=kwargs_clone,
 
  )
 
  res = await mistral_client.chat.complete_async(**kwargs)
 
  langfuse_context.update_current_observation(
      usage={
          "input": res.usage.prompt_tokens,
          "output": res.usage.completion_tokens
      },
      output=res.choices[0].message.content
  )
 
  return res
 
@observe()
async def async_find_best_musician_from(country="France"):
  response = await async_mistral_completion(
      model="mistral-small-latest",
      max_tokens=1024,
      messages=[
        {
            "content": "Who is the best musician from {country}? Answer in one short sentence.".format(country=country),
            "role": "user",
        },
      ]
    )
  return response
 
await async_find_best_musician_from("Spain")
ChatCompletionResponse(id='589fa6216c5346cc984586209c693a41', object='chat.completion', model='mistral-small-latest', usage=UsageInfo(prompt_tokens=17, completion_tokens=33, total_tokens=50), created=1726597737, choices=[ChatCompletionChoice(index=0, message=AssistantMessage(content="One of the most renowned musicians from Spain is Andrés Segovia, a classical guitarist who significantly impacted the instrument's modern repertoire.", tool_calls=None, prefix=False, role='assistant'), finish_reason='stop')])

Example trace in Langfuse: https://cloud.langfuse.com/project/cloramnkj0002jz088vzn1ja4/traces/1f7d91ce-45dd-41bf-8e6f-1875086ed32f

Async Streaming

This example demonstrates the use of the @observe decorator in an asynchronous streaming context. It wraps an async function that streams responses from the Mistral model, logging each chunk of data in real-time.

import asyncio
 
# Wrap async streaming function with decorator
@observe(as_type="generation")
async def async_stream_mistral_completion(**kwargs):
    kwargs_clone = kwargs.copy()
    input = kwargs_clone.pop('messages', None)
    model = kwargs_clone.pop('model', None)
    min_tokens = kwargs_clone.pop('min_tokens', None)
    max_tokens = kwargs_clone.pop('max_tokens', None)
    temperature = kwargs_clone.pop('temperature', None)
    top_p = kwargs_clone.pop('top_p', None)
 
    model_parameters = {
        "maxTokens": max_tokens,
        "minTokens": min_tokens,
        "temperature": temperature,
        "top_p": top_p
    }
    model_parameters = {k: v for k, v in model_parameters.items() if v is not None}
 
    langfuse_context.update_current_observation(
        input=input,
        model=model,
        model_parameters=model_parameters,
        metadata=kwargs_clone,
    )
 
    res = await mistral_client.chat.stream_async(**kwargs)
    final_response = ""
    async for chunk in res:
        content = chunk.data.choices[0].delta.content
        final_response += content
        yield content
 
        if chunk.data.choices[0].finish_reason == "stop":
            langfuse_context.update_current_observation(
                usage={
                    "input": chunk.data.usage.prompt_tokens,
                    "output": chunk.data.usage.completion_tokens
                },
                output=final_response
            )
            break
 
@observe()
async def async_stream_find_best_five_musician_from(country="France"):
    response_chunks = async_stream_mistral_completion(
        model="mistral-small-latest",
        max_tokens=1024,
        messages=[
            {
                "content": "Who are the best five musician from {country}? Answer in one short sentence.".format(country=country),
                "role": "user",
            },
        ]
    )
    final_response = ""
    async for chunk in response_chunks:
        final_response += chunk
        # You can also do something with each chunk here if needed
        print(chunk)
 
    return final_response
 
# Run the async function
await async_stream_find_best_five_musician_from("Spain")
The
 five
 most
 renown
ed
 musicians
 from
 Spain
 include
:
 Andr
és
 Seg
ov
ia
,
 Pac
o
 de
 Luc
ía
,
 En
rique
 I
gles
ias
,
 Ale
j
andro
 San
z
,
 and
 Ros
al
ía
.






'The five most renowned musicians from Spain include: Andrés Segovia, Paco de Lucía, Enrique Iglesias, Alejandro Sanz, and Rosalía.'

Example trace in Langfuse: https://cloud.langfuse.com/project/cloramnkj0002jz088vzn1ja4/traces/36608110-f6cf-4566-a080-7c18777e2dbf

Tool Calling

This snippet introduces Mistral’s function-calling capability, where you can define custom functions to retrieve specific data, like payment status and date, based on a transaction ID. These functions are then registered with the Mistral model, allowing it to call them when processing queries. For a deeper dive into function calling with Mistral, refer to the official Mistral documentation.

import pandas as pd
import json
import functools
 
 
# Sample payment transaction data
data = {
    'transaction_id': ['T1001', 'T1002', 'T1003', 'T1004', 'T1005'],
    'customer_id': ['C001', 'C002', 'C003', 'C002', 'C001'],
    'payment_amount': [125.50, 89.99, 120.00, 54.30, 210.20],
    'payment_date': ['2021-10-05', '2021-10-06', '2021-10-07', '2021-10-05', '2021-10-08'],
    'payment_status': ['Paid', 'Unpaid', 'Paid', 'Paid', 'Pending']
}
 
# Create a DataFrame from the data
df = pd.DataFrame(data)
 
# Function to retrieve payment status given a transaction ID
def retrieve_payment_status(df: data, transaction_id: str) -> str:
    if transaction_id in df.transaction_id.values:
        # Return the payment status as a JSON string
        return json.dumps({'status': df[df.transaction_id == transaction_id].payment_status.item()})
    return json.dumps({'error': 'transaction id not found.'})
 
# Function to retrieve payment date given a transaction ID
def retrieve_payment_date(df: data, transaction_id: str) -> str:
    if transaction_id in df.transaction_id.values:
        # Return the payment date as a JSON string
        return json.dumps({'date': df[df.transaction_id == transaction_id].payment_date.item()})
    return json.dumps({'error': 'transaction id not found.'})
 
# Define tools for the Mistral model with JSON schemas
tools = [
  {
      "type": "function",
      "function": {
          "name": "retrieve_payment_status",
          "description": "Get payment status of a transaction",
          "parameters": {
              "type": "object",
              "properties": {
                  "transaction_id": {
                      "type": "string",
                      "description": "The transaction id.",
                  }
              },
              "required": ["transaction_id"],
          },
      },
  },
  {
      "type": "function",
      "function": {
          "name": "retrieve_payment_date",
          "description": "Get payment date of a transaction",
          "parameters": {
              "type": "object",
              "properties": {
                  "transaction_id": {
                      "type": "string",
                      "description": "The transaction id.",
                  }
              },
              "required": ["transaction_id"],
          },
      },
  }
]
 
# Define tools for the Mistral model with JSON schemas
names_to_functions = {
  'retrieve_payment_status': functools.partial(retrieve_payment_status, df=df),
  'retrieve_payment_date': functools.partial(retrieve_payment_date, df=df)
}

The check_transaction_status function demonstrates the use of Mistral’s function-calling capabilities. The function’s result is then incorporated into the LLM’s response, which is logged and traced in Langfuse. This example illustrates how external function calls can be seamlessly integrated into a Langfuse by using the the wrapped mistral_completion function, ensuring that every step — from tool selection to final output - is captured for thorough observability.

@observe()
def tool_calling_check_transaction_status(id="T1001"):
 
  # Construct the initial user query message
  messages = [{"role": "user", "content": "What's the status of my transaction {id}?".format(id=id)}]
 
  # Use the Langfuse-decorated Mistral completion function to generate a tool-assisted response
  response = mistral_completion(
      model = "mistral-small-latest",
      messages = messages,
      max_tokens=512,
      temperature=0.1,
      tools = tools,
      tool_choice = "any",
  )
 
 
  messages.append(response.choices[0].message)
 
  # Extract the tool call details from the model's response
  tool_call = response.choices[0].message.tool_calls[0]
  function_name = tool_call.function.name
  function_params = json.loads(tool_call.function.arguments)
 
   # Execute the selected function with the extracted parameters
  function_result = names_to_functions[function_name](**function_params)
 
  messages.append({"role":"tool", "name":function_name, "content":function_result, "tool_call_id":tool_call.id})
 
  # Call the Langfuse-wrapped Mistral completion function again to generate a final response using the tool's result
  response = mistral_completion(
      model = "mistral-small-latest",
      max_tokens=1024,
      temperature=0.5,
      messages = messages
  )
 
  return response.choices[0].message.content
 
tool_calling_check_transaction_status("T1005")
'Your transaction T1005 is currently pending.'

Example trace in Langfuse: https://cloud.langfuse.com/project/cloramnkj0002jz088vzn1ja4/traces/e986408a-f96b-40dc-8278-5d0eb0286f82

Questions?

Join our Discord community in case you have any questions.

Was this page useful?

Questions? We're here to help

Subscribe to updates