agent_tool_invocationTier 1 · 70% confidence

ai-agents-agent-tool-invocatio-structured-chat-zero-shot-react-description-agent--2138e2b9

agent: ai_agents

When does this happen?

IF STRUCTURED_CHAT_ZERO_SHOT_REACT_DESCRIPTION agent finishes the chain with a verbal commitment to use a tool but never actually invokes the tool, leaving the user without the expected tool output.

How others solved it

THEN Switch to the chat-conversational-react-description agent type, or modify any multi-input tool to accept a single JSON string instead of multiple parameters. For example, redefine the tool's _run method to accept a single string 'filters_string' and parse it with json.loads() inside the method.

class ComparablesTool(BaseTool):
    name = "get_comparables"
    description = "... Action input should always be an object inside a string with mentioned parameters"
    def _run(self, filters_string):
        filters = json.loads(filters_string)
        params = {
            'operation': filters.get('operation', '2'),
            'latitude': filters.get('latitude'),
            'longitude': filters.get('longitude'),
        }
        response = requests.get(url, params=params, headers=headers)
        return response.content

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics