deepseek r1 1.5 can not output json format vs gemma 2b and llama 1b can

[HuaizhengZhang]

[dosubot]

To resolve the TypeError: 'NoneType' object is not iterable in the process_plan function, you can modify the function to handle cases where plan_dict.get('tasks') returns None. Use a default empty list if tasks is not found:

def process_plan(plan_dict: dict):
    plan_str = ""
    tasks = plan_dict.get('tasks', [])
    for task in tasks:
        plan_str += f"[green][Task]:[/green] {task.get('task')}\n[green][Description]:[/green] {task.get('description')}\n\n"

    return plan_str

This change ensures that if plan_dict.get('tasks') returns None, an empty list is used instead, preventing the TypeError ^[1].

Regarding the JSON output issue with DeepSeek R1 1.5, ensure that the model's response adheres to the JSON format specified in the PlanAgent class. The class includes a JSON mode prompt with an example JSON output format. If the response is not in JSON format, the plan method attempts to parse it as JSON and uses the clean_json_string function if a JSONDecodeError occurs ^[1].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}