FoundationAgents/MetaGPT · error · ValueError

task_id is not found in the insight_pool

Error message

task_id is not found in the insight_pool

What it means

Raised by InstructionGenerator.load_insight_pool when any entry in the loaded insight pool JSON lacks a 'task_id' key. Every insight must carry a task_id because the pool is filtered by task (int(item['task_id']) == int(task_id)) and later grouped per task.

Source

Thrown at metagpt/ext/sela/insights/instruction_generator.py:109

        rsp_list.append(rsp)
        for item in rsp_list:
            item_dict = json.loads(item)
            data = {
                "Insights": item_dict,
            }
            new_data.append(data)
        return new_data

    @staticmethod
    def load_insight_pool(file_path, use_fixed_insights, task_id=None):
        data = InstructionGenerator.load_json_data(file_path)
        if use_fixed_insights:
            current_directory = os.path.dirname(__file__)
            fixed_insights = InstructionGenerator.load_json_data(f"{current_directory}/fixed_insights.json")
            data.extend(fixed_insights)
        for item in data:
            if "task_id" not in item:
                raise ValueError("task_id is not found in the insight_pool")

        if task_id:
            data = [item for item in data if int(item["task_id"]) == int(task_id)]
        return data

    async def generate_new_instructions(self, task_id, original_instruction, max_num, ext_info=None):
        data = self.insight_pool
        new_instructions = []
        if len(data) == 0:
            mcts_logger.log("MCTS", f"No insights available for task {task_id}")
            # return [original_instruction]  # Return the original instruction if no insights are available
        for i in range(max_num):
            if len(data) == 0:
                insights = "No insights available"
            else:
                item = data[i]
                insights = item["Analysis"]
            new_instruction = await InstructionGenerator.generate_new_instruction(

View on GitHub (pinned to 11cdf466d0)

Solutions

1. Add an integer 'task_id' field to every entry in the insight pool JSON
2. Validate the JSON offline: assert all('task_id' in item for item in data)
3. Keep the schema of your added insights identical to existing entries

Example fix

# before
{"Analysis": "..."}

# after
{"task_id": 0, "Analysis": "..."}

data = json.load(open(pool_path))
missing = [i for i, item in enumerate(data) if "task_id" not in item]
assert not missing, f"entries missing task_id: {missing}"

def insight_pool_valid(data: list[dict]) -> bool:
    return all(isinstance(item, dict) and "task_id" in item for item in data)

When it happens

Trigger: Loading an analysis_pool/insight_pool JSON where at least one item was added without a task_id field — typically hand-written or externally generated insights.

Common situations: Editing the pool JSON to add domain knowledge and omitting task_id; merging insights from another pool format; use_fixed_insights pulling a malformed fixed_insights.json.

Related errors

• Dataset {dataset_name} not found in config file. Available d
• Dataset {task_name} not found in config file. Available data
• Number of classes {num_classes} not supported
• Target column not provided
• Invalid exp_mode: {args.exp_mode}

AI-assisted analysis of FoundationAgents/MetaGPT@11cdf466d0 (2026-08-14). Data as JSON: /api/errors/5dfccb058b64c7fa. Report an issue: GitHub.