We don't publish
your competitive advantage.

IFAgent fails to execute tool with error: 'the Action Input is not a valid key, value dictionary.'

THENDefine tool's args_schema using a Pydantic BaseModel with explicit Field descriptions for each parameter. In the agent's task description, instruct the LLM to output a valid JSON object matching that schema without extra string escaping. Avoid passing JSON strings directly in tool config; let the LLM generate the raw dictionary.

IFWhen using Llama 3 8B or similar models with CrewAI, custom tools that expect a JSON string argument fail with 'Error: the Action Input is not a valid key, value dictionary' because the model outputs the JSON object directly without proper string quoting.

THENEnsure the tool's `args_schema` Pydantic model has a single string field for the JSON argument, and inside `_run`, parse it with `json.loads`. In the task configuration, pass `tool_input` as a JSON string (e.g., `json.dumps(params)`). Also, instruct the agent to output the tool input as a JSON string enclosed in double quotes, like `'{"key": "value"}'`.

IFLLM outputs JSON string tool input but it is not parsed as a valid dictionary, causing 'Error: the Action Input is not a valid key, value dictionary.'

THENEnsure tool input parsing handles JSON strings robustly, especially for boolean and None values. One workaround is to instruct the LLM to output JSON strings with string representations of booleans and None (e.g., 'True' → '"True"') or to implement a pre-processing step that converts JSON strings into dictionaries before passing to the tool. Alternatively, adjust the tool's args_schema to accept a JSON string directly and parse it internally.