> ## Documentation Index
> Fetch the complete documentation index at: https://novita.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# MLflow
> Rastreie chamadas da Novita AI no MLflow com instrumentação de SDK compatível com OpenAI para Python e JavaScript.
Este guia mostra como integrar a **Novita AI** ao MLflow Tracing. Ao usar o endpoint compatível com OpenAI da Novita AI (`https://api.novita.ai/openai`), você pode capturar prompts, respostas, latência, uso de tokens e metadados do modelo no MLflow.
# Pré-requisitos
Antes de começar, certifique-se de ter:
* Chave de API da Novita AI: crie uma em [Gerenciamento de Chaves](https://novita.ai/settings/key-management).
* Um servidor de rastreamento MLflow em execução. Você pode usar o padrão local `http://localhost:5000`.
* Runtime Python ou JavaScript.
# Etapas de integração
## Etapa 1: Instalar dependências
```bash Python icon="python" theme={"system"}
pip install 'mlflow[genai]' openai
```
```bash JavaScript / TypeScript icon="js" theme={"system"}
npm install @mlflow/openai openai
```
## Etapa 2: Iniciar o servidor MLflow
Se você tiver um ambiente Python local >= 3.10, poderá iniciar o MLflow com:
```bash theme={"system"}
mlflow server
```
O MLflow também fornece uma configuração Docker Compose:
```bash theme={"system"}
git clone --depth 1 --filter=blob:none --sparse https://github.com/mlflow/mlflow.git
cd mlflow
git sparse-checkout set docker-compose
cd docker-compose
cp .env.dev.example .env
docker compose up -d
```
Em seguida, abra `http://localhost:5000` para confirmar que a UI do MLflow está acessível.
## Etapa 3: Habilitar tracing e chamar a Novita AI
```python Python icon="python" theme={"system"}
import openai
import mlflow
# Enable auto-tracing for OpenAI-compatible calls
mlflow.openai.autolog()
# Optional: set tracking target and experiment
mlflow.set_tracking_uri("http://localhost:5000")
mlflow.set_experiment("Novita AI")
client = openai.OpenAI(
base_url="https://api.novita.ai/openai",
api_key="",
)
response = client.chat.completions.create(
model="deepseek/deepseek-r1",
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "What is the capital of France?"},
],
)
print(response.choices[0].message.content)
```
```ts JavaScript / TypeScript icon="js" theme={"system"}
import { OpenAI } from "openai";
import { tracedOpenAI } from "@mlflow/openai";
const client = tracedOpenAI(
new OpenAI({
baseURL: "https://api.novita.ai/openai",
apiKey: "",
})
);
const response = await client.chat.completions.create({
model: "deepseek/deepseek-r1",
messages: [
{ role: "system", content: "You are a helpful assistant." },
{ role: "user", content: "What is the capital of France?" },
],
temperature: 0.1,
max_tokens: 100,
});
console.log(response.choices[0].message?.content);
```
## Etapa 4: Visualizar traces na UI do MLflow
Abra sua UI do MLflow (por exemplo, `http://localhost:5000`) e acesse o experimento configurado para inspecionar os traces.
Você deverá ver:
* Conteúdo do prompt e da conclusão
* Latência e uso de tokens
* Metadados do modelo e da solicitação
* Erros/exceptions (se houver)
## Etapa 5: Referências avançadas de tracing
### Streaming e assíncrono
O MLflow oferece suporte a tracing para APIs de streaming e assíncronas da Novita AI. Consulte:
* [OpenAI Tracing](https://mlflow.org/docs/latest/genai/tracing/integrations/listing/openai/)
### Combinar com frameworks ou tracing manual
```python Python icon="python" theme={"system"}
import json
from openai import OpenAI
import mlflow
from mlflow.entities import SpanType
# Initialize the OpenAI client with Novita AI API endpoint
client = OpenAI(
base_url="https://api.novita.ai/openai",
api_key="",
)
# Create a parent span for the Novita AI call
@mlflow.trace(span_type=SpanType.CHAIN)
def answer_question(question: str):
messages = [{"role": "user", "content": question}]
response = client.chat.completions.create(
model="deepseek/deepseek-r1",
messages=messages,
)
# Attach session/user metadata to the trace
mlflow.update_current_trace(
metadata={
"mlflow.trace.session": "session-12345",
"mlflow.trace.user": "user-a",
}
)
return response.choices[0].message.content
answer = answer_question("What is the capital of France?")
```
```ts JavaScript / TypeScript icon="js" theme={"system"}
import * as mlflow from "@mlflow/core";
import { OpenAI } from "openai";
import { tracedOpenAI } from "@mlflow/openai";
mlflow.init({
trackingUri: "http://localhost:5000",
experimentId: "",
});
// Wrap the OpenAI client and point to Novita AI endpoint
const client = tracedOpenAI(
new OpenAI({
baseURL: "https://api.novita.ai/openai",
apiKey: "",
})
);
// Create a traced function that wraps the Novita AI call
const answerQuestion = mlflow.trace(
async (question: string) => {
const resp = await client.chat.completions.create({
model: "deepseek/deepseek-r1",
messages: [{ role: "user", content: question }],
});
return resp.choices[0].message?.content;
},
{ name: "answer-question" }
);
await answerQuestion("What is the capital of France?");
```
Para a referência upstream completa:
* Página de integração MLflow Novita AI: [Tracing Novita AI](https://mlflow.org/docs/latest/genai/tracing/integrations/listing/novitaai/)
* Documentação de tracing OpenAI do MLflow: [OpenAI Tracing](https://mlflow.org/docs/latest/genai/tracing/integrations/listing/openai/)
Para detalhes dos modelos Novita e uso do endpoint, consulte:
* Guia da API LLM da Novita: [LLM API](/docs/pt-BR/guides/llm-api)