mem0ai/mem0 · error · ValueError

Model '{model}' does not support function calling. Please us

Error message

Model '{model}' does not support function calling. Please use a model that supports function calling.

What it means

Mem0Proxy checks litellm.supports_function_calling(model) before routing the request, because the proxy injects memory-management tools into the completion call. If litellm's model registry says the requested model cannot do function/tool calling, it raises ValueError immediately rather than sending a request that would fail obscurely at the provider.

Source

Thrown at mem0/proxy/main.py:101

        parallel_tool_calls: Optional[bool] = None,
        deployment_id=None,
        extra_headers: Optional[dict] = None,
        # soon to be deprecated params by OpenAI
        functions: Optional[List] = None,
        function_call: Optional[str] = None,
        # set api_base, api_version, api_key
        base_url: Optional[str] = None,
        api_version: Optional[str] = None,
        api_key: Optional[str] = None,
        model_list: Optional[list] = None,  # pass in a list of api_base,keys, etc.
    ):
        if messages is None:
            messages = []
        if not any([user_id, agent_id, run_id]):
            raise ValueError("One of user_id, agent_id, run_id must be provided")

        if not litellm.supports_function_calling(model):
            raise ValueError(
                f"Model '{model}' does not support function calling. Please use a model that supports function calling."
            )

        prepared_messages = self._prepare_messages(messages)
        if prepared_messages[-1]["role"] == "user":
            self._async_add_to_memory(messages, user_id, agent_id, run_id, metadata, filters)
            relevant_memories = self._fetch_relevant_memories(messages, user_id, agent_id, run_id, filters, top_k)
            logger.debug(f"Retrieved {len(relevant_memories)} relevant memories")
            prepared_messages[-1]["content"] = self._format_query_with_memories(messages, relevant_memories)

        response = litellm.completion(
            model=model,
            messages=prepared_messages,
            temperature=temperature,
            top_p=top_p,
            n=n,
            timeout=timeout,
            stream=stream,

View on GitHub (pinned to 001c235229)

Solutions

1. Use a model known to support function calling (e.g. gpt-4o, gpt-4o-mini, claude-3.5-sonnet, gemini-1.5-pro)
2. Print litellm.supports_function_calling(model) in a quick REPL to see how litellm classifies your model string
3. Upgrade litellm so newer models are recognized, or pass the fully qualified provider-prefixed name (e.g. 'azure/<deployment>') so lookup succeeds
4. For local models, use a tool-calling-capable server (vLLM/Ollama with tools enabled) and the matching litellm provider prefix

Example fix

# before
resp = mem0_proxy.completion(messages=messages, model="some-base-model", user_id="alice")

# after
import litellm
assert litellm.supports_function_calling("gpt-4o-mini")
resp = mem0_proxy.completion(messages=messages, model="gpt-4o-mini", user_id="alice")

import litellm
if not litellm.supports_function_calling(model):
    raise ValueError(f"{model} cannot be used with the mem0 proxy; pick a tool-capable model")

When it happens

Trigger: Passing a chat-only or completion-only model name (e.g. some base/instruct models, older tiers, or unknown custom model strings) to mem0_proxy.completion(); typos in model names that make litellm fall back to a registry entry without function-calling support; using a custom 'openai/<deployed-name>' style string litellm cannot classify.

Common situations: Switching from gpt-4o to a cheaper/local model (e.g. some llama variants) that lacks tool support; on-prem deployments with custom model names litellm maps conservatively; litellm version changes that alter the supports_function_calling table.

Related errors

• LiteLLM failed: ${message}
• One of user_id, agent_id, run_id must be provided
• Invalid response format from ping endpoint
• NET_CONNECT

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/69125aaf4e68b2fa. Report an issue: GitHub.