BerriAI/litellm · error · Exception

[91mLangfuse not installed, try running 'pip install langfu

Error message

[91mLangfuse not installed, try running 'pip install langfuse' to fix this error: {e}\n{traceback.format_exc()}[0m

What it means

LiteLLM's Langfuse callback could not import the 'langfuse' Python package. The import happens lazily inside LangfuseLogger.__init__, so the error only surfaces when the callback is first constructed (e.g. litellm.success_callback = ['langfuse'] or proxy startup with langfuse enabled), not at litellm install time. The message also embeds the underlying import exception and traceback, which can reveal a broken install rather than a missing package.

Source

Thrown at litellm/integrations/langfuse/langfuse.py:134

    return public_key, secret_key, resolved_host


class LangFuseLogger:
    # Class variables or attributes
    def __init__(
        self,
        langfuse_public_key=None,
        langfuse_secret=None,
        langfuse_host=None,
        flush_interval=1,
        allow_env_credentials: bool = True,
    ):
        try:
            import langfuse
            from langfuse import Langfuse
        except Exception as e:
            raise Exception(
                f"\033[91mLangfuse not installed, try running 'pip install langfuse' to fix this error: {e}\n{traceback.format_exc()}\033[0m"
            )
        self.public_key, self.secret_key, self.langfuse_host = resolve_langfuse_credentials(
            langfuse_public_key=langfuse_public_key,
            langfuse_secret=langfuse_secret,
            langfuse_host=langfuse_host,
            allow_env_credentials=allow_env_credentials,
        )
        if not (self.langfuse_host.startswith("http://") or self.langfuse_host.startswith("https://")):
            # add http:// if unset, assume communicating over private network - e.g. render
            self.langfuse_host = "http://" + self.langfuse_host
        self.langfuse_release = os.getenv("LANGFUSE_RELEASE")
        self.langfuse_debug = os.getenv("LANGFUSE_DEBUG")
        self.langfuse_flush_interval = LangFuseLogger._get_langfuse_flush_interval(flush_interval)

        if should_use_langfuse_mock():
            self.langfuse_client = create_mock_langfuse_client()
            self.is_mock_mode = True

View on GitHub (pinned to 6c2dcb801b)

Solutions

pip install langfuse (or pip install 'litellm[langfuse]' if using extras)
Run python -c 'import langfuse' and read the embedded traceback in the error message — if the package IS installed, fix the underlying import failure (commonly a pydantic version conflict) instead of reinstalling
Pin compatible versions, e.g. pip install -U langfuse pydantic, then retry
If you do not need Langfuse, remove it from success_callback / proxy litellm_settings

Example fix

# before
import litellm
litellm.success_callback = ["langfuse"]  # Exception: Langfuse not installed...

# after
# shell: pip install langfuse
import litellm
litellm.success_callback = ["langfuse"]

Defensive patterns

Strategy: validation

Validate before calling

import importlib.util, sys

def langfuse_available() -> bool:
    return importlib.util.find_spec("langfuse") is not None

if not langfuse_available():
    print("Install langfuse or disable the langfuse callback")
    sys.exit(1)

Try / catch

try:
    import litellm
    litellm.success_callback = ["langfuse"]
except Exception as e:
    if "Langfuse not installed" in str(e):
        litellm.success_callback = []  # degrade gracefully: run without observability
    else:
        raise

Prevention

Install optional observability extras in the same commit that enables the callback
Add a 'python -c import langfuse' preflight check to CI for services using langfuse callbacks
Pin langfuse and pydantic versions to avoid import-time breakage

When it happens

Trigger: Setting litellm.success_callback = ['langfuse'] (or defining langfuse callback params in the proxy config) without 'pip install langfuse'; installing a langfuse version whose transitive deps fail to import (e.g. incompatible pydantic); a corrupted venv where 'import langfuse' raises any Exception.

Common situations: Using litellm's optional observability integrations — langfuse is an optional extra, so a plain 'pip install litellm' lacks it. Also seen after upgrading pydantic/langfuse to incompatible versions, where the import itself throws and is misreported as 'not installed'.

Related errors

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/99b5a23dff043a46. Report an issue: GitHub.

[91mLangfuse not installed, try running 'pip install langfu