modelscope
[Bug]:react_agent 执行完tool后开始胡乱回答,并没有使用tool执行的结果
Describe the bug query:品质最高的装备
################################################################################ ##################### ITER 1, STEP 1: REASONING ###################### 2024-04-19 11:23:32.633 | DEBUG | agentscope.models.response:to_dict:74 - Text before parsing 我需要使用"get_best_card"函数来找到品质最高的装备。 2024-04-19 11:23:32.634 | DEBUG | agentscope.models.response:to_dict:81 - Text after parsing
PARSE FUNCTION:
@classmethod
def to_dict(cls, response: ModelResponse) -> ModelResponse:
"""Parse the response text to a dict, and feed it into the `json`
field."""
text = response.text
if text is not None:
logger.debug(f"Text before parsing\n{text}")
# extract from the first '{' to the last '}'
index_start = max(text.find("{"), 0)
index_end = min(text.rfind("}") + 1, len(text))
text = text[index_start:index_end]
logger.debug("Text after parsing", text)
response.text = text
response.json = json.loads(text)
return response
else:
raise ValueError(
f"The text field of the model response is None: {response}",
)
ERROR INFO:
JSONDecodeError: Expecting value: line 1 column 1 (char 0)
INPUT RESPONSE:
{
"text": "",
"embedding": null,
"image_urls": null,
"json": null,
"raw": {
"id": "chatcmpl-9FZ8WHTmdbTiMHy4zkIVmACNMqeLZ",
"choices": [
{
"finish_reason": "stop",
"index": 0,
"logprobs": null,
"message": {
"content": "我需要使用"get_best_card"函数来找到品质最高的装备。",
"role": "assistant",
"function_call": null,
"tool_calls": null
},
"content_filter_results": {
"hate": {
"filtered": false,
"severity": "safe"
},
"self_harm": {
"filtered": false,
"severity": "safe"
},
"sexual": {
"filtered": false,
"severity": "safe"
},
"violence": {
"filtered": false,
"severity": "safe"
}
}
}
],
"created": 1713497012,
"model": "gpt-35-turbo",
"object": "chat.completion",
"system_fingerprint": "fp_2f57f81c11",
"usage": {
"completion_tokens": 21,
"prompt_tokens": 569,
"total_tokens": 590
},
"prompt_filter_results": [
{
"prompt_index": 0,
"content_filter_results": {
"hate": {
"filtered": false,
"severity": "safe"
},
"self_harm": {
"filtered": false,
"severity": "safe"
},
"sexual": {
"filtered": false,
"severity": "safe"
},
"violence": {
"filtered": false,
"severity": "safe"
}
}
}
]
}
}
assistant:
system: Your response is not a JSON object, and cannot be parsed by json.loads in parse function:
Your Response:
[YOUR RESPONSE BEGIN]
[YOUR RESPONSE END]
Error Information:
JSONDecodeError: Expecting value: line 1 column 1 (char 0)
Analyze the reason, and re-correct your response in the correct format. ##################### ITER 2, STEP 1: REASONING ###################### 2024-04-19 11:23:35.452 | DEBUG | agentscope.models.response:to_dict:74 - Text before parsing { "thought": "I need to use the 'get_best_card' function to find the best equipment.", "speak": "我将使用'get_best_card'函数来寻找最好的装备。", "function": [{"name": "get_best_card", "arguments": {"type": "装备", "level": true}}] } 2024-04-19 11:23:35.452 | DEBUG | agentscope.models.response:to_dict:81 - Text after parsing assistant: { "thought": "I need to use the 'get_best_card' function to find the best equipment.", "speak": "\u6211\u5c06\u4f7f\u7528'get_best_card'\u51fd\u6570\u6765\u5bfb\u627e\u6700\u597d\u7684\u88c5\u5907\u3002", "function": [ { "name": "get_best_card", "arguments": { "type": "\u88c5\u5907", "level": true } } ] } ####################### ITER 2, STEP 2: ACTION #######################
Executing function get_best_card ... END system: 根据 Execution Results 回答user最初的问题:
- get_best_card: [EXECUTE STATUS]: SUCCESS [EXECUTE RESULT]: ['力量系头盔金3', '力量系衣服金3', '力量系鞋子金3', '力量系武器金3', '敏捷系头盔金3', '敏捷系衣服金3', '敏捷系鞋子金3', '敏捷系武器金3', '智力系头盔金3', '智力系衣服金3', '智力系鞋子金3', '智力系武器金3']
##################### ITER 3, STEP 1: REASONING ###################### 2024-04-19 11:23:37.219 | DEBUG | agentscope.models.response:to_dict:74 - Text before parsing {"thought": "The best equipment has been successfully retrieved.", "speak": "已成功找到品质最高的装备。", "function": []} 2024-04-19 11:23:37.219 | DEBUG | agentscope.models.response:to_dict:81 - Text after parsing assistant: { "thought": "The best equipment has been successfully retrieved.", "speak": "\u5df2\u6210\u529f\u627e\u5230\u54c1\u8d28\u6700\u9ad8\u7684\u88c5\u5907\u3002", "function": [] }
Hi！你好
- 看log应该已经是根据执行结果进行回答，但是没有回答更详细的内容，这个需要调整prompt进行调试，引导大模型把执行结果详细的表达出来。
- 关于打印出中文的转义码的问题，我这尝试复现了下，没有发现这个问题，可以尝试拉取最新的代码
If you have any further questions or concerns, please feel free to create a new issue. Closing this for now.