[BFCL] Bugs in decode executable predictions

[XuHwang]

Describe the issue
BFCLv3 has modified the decode logic for executable predictions, while I found there are a lot of cases in REST category that could not be decoded by current method.

For example, here is one example:

result="[requests.get(url=\"https://covid-193.p.rapidapi.com/statistics\", headers={\"X-RapidAPI-Key\": \"xxxxxxx\", \"X-RapidAPI-Host\": \"covid-193.p.rapidapi.com\"}, timeout=5, params={\"country\": \"China\"})]"

It would be parsed by decoded_output_to_execution_list function.

Then the nested value {"X-RapidAPI-Key": "xxxxxxx", "X-RapidAPI-Host": \"covid-193.p.rapidapi.com"} would be parsed by parse_nested_value, which would raise error "str object has no attribute .items()"

In summary, this logic only consider the case that the nested value is a sub function while ignoring the truth that the argument type is just a dict.

ID datapoint

1. Datapoint / Model Handler permalink:
2. Issue:
3. Gorilla repo commit #:

What is the issue

Proposed Changes

{
'previous_datapoint':[],
'updated_datapoint':[]
}

Additional context
Add any other context about the problem here.

[HuanzhiMao]

Hi @XuHwang,

Thanks for pointing this out. This is indeed an oversight on our end. We will push a patch for this.