SDK configuration

This guide covers configuration for both Python and TypeScript SDKs, including basic setup, advanced options, and debugging capabilities.

The recommended approach to configuring the Python SDK is to use the opik configure command. This will prompt you to set up your API key and Opik instance URL (if applicable) to ensure proper routing and authentication. All details will be saved to a configuration file.

import opik

opik.configure(use_local=False)

opik configure

import opik

opik.configure(use_local=True)

opik configure --use_local

The configure methods will prompt you for the necessary information and save it to a configuration file (~/.opik.config). When using the command line version, you can use the -y or --yes flag to automatically approve any confirmation prompts:

opik configure --yes

The one-command path is uvx opik mcp configure, which needs no SDK; see Opik's MCP server. If you are already running opik configure, it offers the same step at the end: registering the MCP server with the AI clients it finds on your machine, so your assistant can read traces and log scores directly from the chat.

Pass --install-mcp to skip the prompt and register it:

opik configure --install-mcp

opik configure also offers the Opik skill pack, which teaches your assistant how to instrument code, run test suites, and use opik connect. --install-skills installs it without the prompt:

opik configure --install-mcp --install-skills

To set this up without re-running the whole configuration, use uvx opik mcp configure. See Opik's MCP server for the --ai-client flag, the skill pack, and per-client instructions.

For the TypeScript SDK, configuration is done through environment variables, constructor options, or configuration files.

Installation:

npm install opik

Basic Configuration:

You can configure the Opik client using environment variables in a .env file:

OPIK_API_KEY="your-api-key"
OPIK_URL_OVERRIDE="https://www.comet.com/opik/api"
OPIK_PROJECT_NAME="your-project-name"
OPIK_WORKSPACE="your-workspace-name"

Or pass configuration directly to the constructor:

import { Opik } from "opik";

const client = new Opik({
  apiKey: "<your-api-key>",
  apiUrl: "https://www.comet.com/opik/api",
  projectName: "<your-project-name>",
  workspaceName: "<your-workspace-name>",
});

Both SDKs support multiple configuration approaches with different precedence orders.

Python SDK: Constructor options → Environment variables → Configuration file → Defaults

TypeScript SDK: Constructor options → Environment variables → Configuration file (~/.opik.config) → Defaults

Both SDKs support environment variables for configuration. Here's a comparison of available options:

Both SDKs support .env files for managing environment variables. This is a good practice to avoid hardcoding secrets and to make your configuration more portable.

For Python projects, install python-dotenv:

pip install python-dotenv

For TypeScript projects, dotenv is automatically loaded by the SDK.

Create a .env file in your project's root directory:

# Opik Configuration
OPIK_API_KEY="YOUR_OPIK_API_KEY"
OPIK_URL_OVERRIDE="https://www.comet.com/opik/api"
OPIK_PROJECT_NAME="your-project-name"
OPIK_WORKSPACE="your-workspace-name"
OPIK_DEFAULT_LLM="openai/gpt-5-nano"

# LLM Provider API Keys (if needed)
OPENAI_API_KEY="YOUR_OPENAI_API_KEY"

# Logging Configuration (see Debug Mode and Logging section below)
OPIK_CONSOLE_LOGGING_LEVEL="WARNING"  # Python: Control console output (DEBUG, INFO, WARNING, ERROR, CRITICAL)
OPIK_FILE_LOGGING_LEVEL="DEBUG"       # Python: Enable file logging
OPIK_LOG_LEVEL="DEBUG"                # TypeScript: Control log level

Python usage with .env file:

from dotenv import load_dotenv

load_dotenv()  # Load before importing opik

import opik

# Your Opik code here

TypeScript usage with .env file:

The TypeScript SDK automatically loads .env files, so no additional setup is required:

import { Opik } from "opik";

// Configuration is automatically loaded from .env
const client = new Opik();

Both SDKs support configuration files for persistent settings.

The Python SDK uses the TOML format. The configure method creates this file automatically, but you can also create it manually:

[opik]
url_override = https://www.comet.com/opik/api
api_key = <API Key>
workspace = <Workspace name>
project_name = <Project Name>

[opik]
url_override = http://localhost:5173/api
workspace = default
project_name = <Project Name>

The TypeScript SDK also supports the same ~/.opik.config file format as the Python SDK. The configuration file uses INI format internally but shares the same structure:

[opik]
url_override = https://www.comet.com/opik/api
api_key = <API Key>
workspace = <Workspace name>
project_name = <Project Name>

[opik]
url_override = http://localhost:5173/api
workspace = default
project_name = <Project Name>

Both SDKs provide debug capabilities for troubleshooting integration issues.

The Python SDK provides two separate logging channels that can be configured independently:

• Console Logging: Controls log output to the console (stdout/stderr)
• File Logging: Controls log output to a file

Both channels support the following log levels: DEBUG, INFO (default), WARNING, ERROR, CRITICAL

To control the console log level, set the OPIK_CONSOLE_LOGGING_LEVEL environment variable before importing opik:

# Reduce console output to warnings and errors only
export OPIK_CONSOLE_LOGGING_LEVEL="WARNING"

Available log levels for console:

• DEBUG: Show all debug information
• INFO: Show informational messages (default)
• WARNING: Show only warnings and errors
• ERROR: Show only errors and critical messages
• CRITICAL: Show only critical errors

Using with .env file:

# Console Logging (reduce noise)
OPIK_CONSOLE_LOGGING_LEVEL="WARNING"

To enable debug mode with file logging, set these environment variables before importing opik:

export OPIK_FILE_LOGGING_LEVEL="DEBUG"
export OPIK_LOGGING_FILE=".tmp/opik-debug-$(date +%Y%m%d).log"

Using with .env file:

# File Logging (for debug)
OPIK_FILE_LOGGING_LEVEL="DEBUG"
OPIK_LOGGING_FILE=".tmp/opik-debug.log"

Example combining both console and file logging:

# Opik Logging Configuration

# Console: Show only warnings and errors
OPIK_CONSOLE_LOGGING_LEVEL="WARNING"

# File: Log everything for debugging
OPIK_FILE_LOGGING_LEVEL="DEBUG"
OPIK_LOGGING_FILE=".tmp/opik-debug.log"

Then in your Python script:

from dotenv import load_dotenv

load_dotenv()  # Load before importing opik

import opik

# Your Opik code here - console will be quiet, debug logs go to file

The TypeScript SDK uses structured logging with configurable levels:

Available log levels: SILLY, TRACE, DEBUG, INFO (default), WARN, ERROR, FATAL

Enable debug logging:

export OPIK_LOG_LEVEL="DEBUG"

Or in .env file:

OPIK_LOG_LEVEL="DEBUG"

Programmatic control:

import { setLoggerLevel, disableLogger } from "opik";

// Set log level
setLoggerLevel("DEBUG");

// Disable logging entirely
disableLogger();

The Opik Python SDK uses the httpx library to make HTTP requests. The default configuration applied to the HTTP client is suitable for most use cases, but you can customize it by registering a custom httpx client hook as in following example:

import opik.hooks

def custom_auth_client_hook(client: httpx.Client) -> None:
    client.auth = CustomAuth()

hook = opik.hooks.HttpxClientHook(
    client_init_arguments={"trust_env": False},
    client_modifier=custom_auth_client_hook,
)
opik.hooks.add_httpx_client_hook(hook)

# Use the Opik SDK as usual

Make sure to add the hook before using the Opik SDK.

The TypeScript SDK uses batching for optimal performance. You can configure batching behavior:

import { Opik } from "opik";

const client = new Opik({
  // Custom batching delay (default: 300ms)
  batchDelayMs: 1000,

  // Hold data until manual flush (default: false)
  holdUntilFlush: true,

  // Custom headers
  headers: {
    "Custom-Header": "value",
  },
});

// Manual flushing
await client.flush();

import { flushAll } from "opik";

// Flush all instantiated clients
await flushAll();

The Python SDK reports usage analytics that tell us which SDK features are being used, so we know what to invest in. Each API you call is reported once per run, by name only, alongside the same environment details the SDK already reports with error reports: the SDK and Python versions, the OS, whether you are in Jupyter, Colab or CI, the kind of installation (cloud, self-hosted or local), and the versions of the LLM libraries installed next to Opik.

Usage is attributed to your workspace name — the same identifier used for error reporting, so that an error and the usage around it describe the same user. Installations left on the default workspace are attributed to a one-way hash of the hostname and username instead.

Never sent: the contents of your traces, spans, prompts, datasets or evaluation results, your API key, or your project names. Reporting always happens on a background thread and never affects your application.

Events are sent to Comet's usage-reporting endpoint, the same pipeline the Opik UI and backend already report through.

To turn it off:

export OPIK_ANALYTICS_ENABLE=false

If you encounter the following error:

[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: self-signed certificate in certificate chain (_ssl.c:1006)

You can resolve it by either:

• Disable the TLS certificate check by setting the OPIK_CHECK_TLS_CERTIFICATE environment variable to false
• Add the Opik server's certificate to your trusted certificates by setting the REQUESTS_CA_BUNDLE environment variable

If you are experiencing problems with the Python SDK, such as receiving 400 or 500 errors from the backend, or being unable to connect at all, run the health check command:

opik healthcheck

This command will analyze your configuration and backend connectivity, providing useful insights into potential issues.

Reviewing the health check output can help pinpoint the source of the problem and suggest possible resolutions.

The TypeScript SDK validates configuration at startup. Common errors:

• "OPIK_URL_OVERRIDE is not set": Set the OPIK_URL_OVERRIDE environment variable
• "OPIK_API_KEY is not set": Required for Opik Cloud deployments
• "OPIK_WORKSPACE is not set": Optional, but can be set for Opik Cloud deployments

Enable debug logging to troubleshoot issues:

export OPIK_LOG_LEVEL="DEBUG"

If you are using the Opik Optimizer SDK, you can also enable optimizer-side debug logs:

export OPIK_OPTIMIZER_LOG_LEVEL="DEBUG"

Or programmatically:

import { setLoggerLevel } from "opik";
setLoggerLevel("DEBUG");

If data isn't appearing in Opik:

1. Check if data is batched: Call await client.flush() to force sending
2. Verify configuration: Ensure correct API URL and credentials
3. Check network connectivity: Verify firewall and proxy settings

1. Python: Ensure load_dotenv() is called before importing opik
2. TypeScript: The SDK automatically loads .env files
3. Verify file location: .env file should be in project root
4. Check file format: No spaces around = in .env files

1. File location: Default is ~/.opik.config
2. Custom location: Use OPIK_CONFIG_PATH environment variable
3. File format: Python uses TOML, TypeScript uses INI format
4. Permissions: Ensure file is readable by your application