BerriAI/litellm · error · ValueError

Palm was decommisioned on October 2024. Please use the `gemi

Error message

Palm was decommisioned on October 2024. Please use the `gemini/` route for Gemini Google AI Studio Models. Announcement: https://ai.google.dev/palm_docs/palm?hl=en

What it means

ValueError raised by the completion router when custom_llm_provider == 'palm': Google PaLM was decommissioned in October 2024, and litellm keeps the branch only to fail fast with a pointer to the replacement route. The PaLM API no longer exists server-side, so retrying against it can never work.

Source

Thrown at litellm/main.py:5698

        elif custom_llm_provider == "databricks":
            response = _complete_databricks(_dispatch_ctx)

        elif custom_llm_provider == "datarobot":
            response = _complete_datarobot(_dispatch_ctx)
        elif custom_llm_provider == "openrouter":
            response = _complete_openrouter(_dispatch_ctx)
        elif custom_llm_provider == "vercel_ai_gateway":
            response = _complete_vercel_ai_gateway(_dispatch_ctx)
        elif (
            custom_llm_provider == "together_ai"
            or ("togethercomputer" in model)
            or (model in litellm.together_ai_models)
        ):
            """
            Deprecated. We now do together ai calls via the openai client - https://docs.together.ai/docs/openai-api-compatibility
            """
        elif custom_llm_provider == "palm":
            raise ValueError(
                "Palm was decommisioned on October 2024. Please use the `gemini/` route for Gemini Google AI Studio Models. Announcement: https://ai.google.dev/palm_docs/palm?hl=en"
            )
        elif custom_llm_provider == "vertex_ai_beta" or custom_llm_provider == "gemini":
            response = _complete_vertex_ai_beta(_dispatch_ctx)

        elif custom_llm_provider == "vertex_ai":
            response = _complete_vertex_ai(_dispatch_ctx)
        elif custom_llm_provider == "predibase":
            response = _complete_predibase(_dispatch_ctx)
        elif custom_llm_provider == "text-completion-codestral":
            response = _complete_text_completion_codestral(_dispatch_ctx)
        elif custom_llm_provider == "text-completion-inception":
            response = _complete_text_completion_inception(_dispatch_ctx)
        elif custom_llm_provider in ("sagemaker_chat", "sagemaker_nova"):
            # boto3 reads keys from .env
            # sagemaker_chat: HF Messages API endpoints
            # sagemaker_nova: Nova models on SageMaker (OpenAI-compatible)
            response = _complete_sagemaker_chat(_dispatch_ctx)

View on GitHub (pinned to 77b7c6c40c)

Solutions

Switch the model string to the Gemini route: model='gemini/gemini-1.5-flash' (or a current Gemini model)
Set GEMINI_API_KEY (Google AI Studio key) for the gemini/ provider
Sweep configs/env/DB for 'palm' strings so nothing resurrects the old route
If you need Vertex, use 'vertex_ai/' prefixed Gemini models with GCP auth instead

Example fix

# before
resp = litellm.completion(model='palm/chat-bison', messages=m)  # ValueError: decommissioned

# after
import os
os.environ['GEMINI_API_KEY'] = 'AIza...'
resp = litellm.completion(model='gemini/gemini-1.5-flash', messages=m)

Defensive patterns

Strategy: validation

Validate before calling

if model.startswith('palm/'):
    model = 'gemini/' + model.split('/', 1)[1]  # migrate legacy PaLM names
    os.environ.setdefault('GEMINI_API_KEY', key)

Type guard

def is_deprecated_palm(model: str) -> bool:
    return model.startswith('palm/') or model == 'palm'

Try / catch

try:
    resp = litellm.completion(model=model, messages=m)
except ValueError as e:
    if 'decommisioned' in str(e):
        raise RuntimeError('PaLM is gone; use gemini/<model> with GEMINI_API_KEY') from e
    raise

Prevention

Grep configs, env files, and DB-stored model names for 'palm' and migrate them to gemini/ routes
Keep GEMINI_API_KEY configured wherever PaLM keys used to live
Add a startup lint that rejects any model prefix from a known-dead provider list
Track provider deprecations in litellm release notes when upgrading major versions

When it happens

Trigger: Legacy code or configs still using model='palm/...' / custom_llm_provider='palm' -- old tutorials, stale environment presets, or upgraded litellm in a project that never migrated off PaLM.

Common situations: Maintaining an old codebase after a litellm upgrade; docs/samples predating the shutdown; a model name pulled from an old database/config table that still stores 'palm/...'.

Related errors

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/ef616c578044571f. Report an issue: GitHub.