BerriAI/litellm · error · OCIError

Tool call `function.arguments` must be a JSON string

Error message

Tool call `function.arguments` must be a JSON string

What it means

OCI GENERIC carries a tool call's arguments as a JSON string (matching OpenAI's convention), so function.arguments must be a str. A dict, list, None-as-absent-then-dict, or number raises OCIError 400. Note the translator defaults missing arguments to "{}" — only a present-but-non-string value fails.

Solutions

1. Serialize dict arguments back to a JSON string: json.dumps(args) if isinstance(args, dict) else args
2. Send arguments exactly as received from the previous model response (always a string there)
3. Add a normalization pass over history before OCI calls (see validationCode)
4. Beware mixed histories: normalize per-item, not globally

Example fix

# before
{"function": {"name": "get_weather", "arguments": {"city": "SF"}}}

# after
{"function": {"name": "get_weather", "arguments": "{\"city\":\"SF\"}"}}

import json

def normalize_tool_arguments(tool_calls: list) -> list:
    out = []
    for tc in tool_calls:
        fn = tc.get("function", {})
        args = fn.get("arguments", "{}")
        if not isinstance(args, str):
            args = json.dumps(args)
            tc = {**tc, "function": {**fn, "arguments": args}}
        out.append(tc)
    return out

def tool_arguments_is_json_string(tc) -> bool:
    args = tc.get("function", {}).get("arguments", "{}")
    return isinstance(args, str)

When it happens

Trigger: Replaying assistant tool_calls where arguments was left as a parsed dict (common after json.loads of the whole history or from frameworks that auto-parse), e.g. {"function":{"name":"f","arguments":{"city":"SF"}}} to an oci/generic.* model.

Common situations: Auto-parsing arguments for display/editing and forgetting to re-serialize; frameworks (some agent libs) exposing arguments as dicts; hand-constructed tool_calls from code; LLM-generated JSON where arguments is naturally an object.

Related errors

• Each tool call must be a dictionary
• Message `tool_calls` must be a list
• OCI only supports function tool calls
• Tool call `function` must be a dictionary
• Tool call `function.name` must be a string

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/8554522c47658361. Report an issue: GitHub.

Appendix: source

Thrown at litellm/llms/oci/chat/generic.py:137

            raise OCIError(status_code=400, message="Each tool call must be a dictionary")
        if tool_call.get("type") != "function":
            raise OCIError(status_code=400, message="OCI only supports function tool calls")

        tool_call_id = tool_call.get("id")
        if not isinstance(tool_call_id, str):
            raise OCIError(status_code=400, message="Tool call `id` must be a string")

        tool_function = tool_call.get("function")
        if not isinstance(tool_function, dict):
            raise OCIError(status_code=400, message="Tool call `function` must be a dictionary")

        function_name = tool_function.get("name")
        if not isinstance(function_name, str):
            raise OCIError(status_code=400, message="Tool call `function.name` must be a string")

        arguments = tool_call["function"].get("arguments", "{}")
        if not isinstance(arguments, str):
            raise OCIError(
                status_code=400,
                message="Tool call `function.arguments` must be a JSON string",
            )

        tool_calls_formatted.append(
            OCIToolCall(
                id=tool_call_id,
                type="FUNCTION",
                name=function_name,
                arguments=arguments,
            )
        )

    return OCIMessage(
        role=open_ai_to_generic_oci_role_map[role],
        content=None,
        toolCalls=tool_calls_formatted,
        toolCallId=None,

View on GitHub (pinned to 77b7c6c40c)