Trace OpenAI-compatible providers

Many LLM providers accept requests in the same format as the OpenAI API. To trace calls from these providers to LangSmith, construct an OpenAI client pointed at the provider’s base URL, then wrap it with wrap_openai / wrapOpenAI. Use wrap_openai / wrapOpenAI for direct API calls. Use @traceable when you need to trace agent logic around the call or set metadata per invocation.

	`wrap_openai` / `wrapOpenAI`	`@traceable` / `traceable`
Token tracking	Automatic	Requires `run_type="llm"`
Run type	LLM (set automatically)	Chain by default
Traces	The API call	The function wrapping it
Metadata	Client-level only (Python); client-level or per-call (TypeScript)	Per-call via `langsmith_extra`

To trace OpenAI directly, refer to Trace with OpenAI.

In an agent-based workspace, create an agent once before sending traces. Use agent addressing to select its identifier and environment. If an integration requires project_name, use the selected environment’s tracer_session_name as a compatibility value.

Setup

pip install langsmith openai

npm install langsmith openai

export LANGSMITH_API_KEY=<your-api-key>
export LANGSMITH_AGENT_ID="<your-agent-id>"
export LANGSMITH_AGENT_ENVIRONMENT="local"
export LANGSMITH_TRACING=true

Trace API calls

import os

import openai
from langsmith import wrappers

client = wrappers.wrap_openai(
    openai.OpenAI(
        base_url="https://<provider-base-url>/v1",
        api_key=os.environ["PROVIDER_API_KEY"],
    )
)

completion = client.chat.completions.create(
    model="<provider-model-name>",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(completion.choices[0].message.content)

import OpenAI from "openai";
import { wrapOpenAI } from "langsmith/wrappers/openai";

const client = wrapOpenAI(
  new OpenAI({
    baseURL: "https://<provider-base-url>/v1",
    apiKey: process.env.PROVIDER_API_KEY!,
  })
);

const completion = await client.chat.completions.create({
  model: "<provider-model-name>",
  messages: [{ role: "user", content: "Hello!" }],
});
console.log(completion.choices[0].message.content);

Add metadata

Python
TypeScript

Pass tracing_extra when wrapping the client. The metadata applies to all calls made with that client.

import os

import openai

from langsmith import wrappers

client = wrappers.wrap_openai(
    openai.OpenAI(
        base_url="https://<provider-base-url>/v1",
        api_key=os.environ["PROVIDER_API_KEY"],
    ),
    tracing_extra={"metadata": {"environment": "production"}},
)

Pass options as the second argument to wrapOpenAI for client-level metadata, or pass langsmithExtra per call.

import OpenAI from "openai";
import { wrapOpenAI } from "langsmith/wrappers/openai";

const client = wrapOpenAI(
  new OpenAI({
    baseURL: "https://<provider-base-url>/v1",
    apiKey: process.env.PROVIDER_API_KEY!,
  }),
  { metadata: { environment: "production" } }
);

// Per-call metadata
const completion = await client.chat.completions.create(
  {
    model: "<provider-model-name>",
    messages: [{ role: "user", content: "Hello!" }],
  },
  { langsmithExtra: { metadata: { request_id: "abc123" } } }
);

Some providers have dedicated setup guides that use @traceable or a native callback. These approaches trace at the function level rather than wrapping the client directly, or integrate with the provider’s own SDK and routing layer.

DeepSeek: OpenAI-compatible API; guide uses @traceable with custom provider metadata
LiteLLM: proxy that exposes an OpenAI-compatible endpoint; guide covers @traceable and LiteLLM’s built-in LangSmith callback

Connect these docs to your agent of choice via MCP for real-time answers.

Edit this page on GitHub or file an issue.

Tracing setup

Configuration & troubleshooting

Trace OpenAI-compatible providers

Setup

Trace API calls

Add metadata

​Setup

​Trace API calls

​Add metadata

​Related guides

Setup

Trace API calls

Add metadata

Related guides