Mistral — CAELUM docs

pip install "caelum-ai[mistral]"

from caelum import Mistral  # instead of: from mistralai.client import Mistral

caelum.Mistral accepts exactly the same arguments as mistralai.client.Mistral, plus the optional CAELUM options.

Method	CAELUM features
`chat.complete`, `chat.complete_async`	exact cache, retries, metrics
everything else	forwarded untouched (`chat.stream`, `chat.parse`, `embeddings`, `agents`...)

Chat

import os
from caelum import Mistral

client = Mistral(api_key=os.environ["MISTRAL_API_KEY"], cache=True)
response = client.chat.complete(
    model="mistral-small-latest",
    messages=[{"role": "user", "content": "What is an inference cache?"}],
    max_tokens=256,
)
print(response.choices[0].message.content)
client.metrics.print_summary()

Asynchronous

chat.complete_async shares the same cache and metrics.

import asyncio
import os
from caelum import Mistral

async def main() -> None:
    async with Mistral(api_key=os.environ["MISTRAL_API_KEY"], cache=True) as client:
        response = await client.chat.complete_async(
            model="mistral-small-latest", messages=[{"role": "user", "content": "Hi"}]
        )
        print(response.choices[0].message.content)

asyncio.run(main())

Retries

client = Mistral(api_key="...", retry=True)  # 429, 5xx and network errors, with backoff

Do not combine retry=True with the SDK's own retry_config: attempts would multiply.

Costs

Prices of the current Small, Medium and Large generations are built in; -latest aliases and dated versions match them. Add other models with caelum.pricing.register_price("model-name", input_per_1m, output_per_1m).

Testing status

caelum.Mistral is tested against a simulated Mistral server. Mistral models were also tested for real through an OpenAI-compatible API (OpenRouter), but not yet through the paid Mistral API itself. Please report anything that behaves differently.

Not supported yet

The semantic cache, routing and cross-provider fallback do not work with Mistral yet.