Topology graph Key Error (After Jina 3.21.0)
ZhihaoAIRobotic opened this issue · 5 comments
Bug Description
There is a KeyError: 'type' in the File "/home/lzh/.conda/envs/DIMA/lib/python3.8/site-packages/jina/serve/runtimes/gateway/graph/topology_graph.py", line 90.
Solution
I have printed the schema_1_properties[property_1], the result is as follows:
There is no key named 'type' for Env Info
Here is the code for Env Info:
class EnvInfo(BaseDoc):
env_memory: ShortTermMemory = ShortTermMemory() # Interaction information of all agent
history: str = ''class ShortTermMemory(DocList[Info]):
def add(self, info: Info):
if info in self:
return
self.append(info)
def add_batch(self, infos: DocList[Info]):
for info in infos:
self.add(info)
def remember(self, k=0) -> DocList[Info]:
"""Return the most recent k memories, return all when k=0"""
return self[-k:]
def remember_news(self, observed: DocList[Info], k=0) -> DocList[Info]:
"""remember the most recent k memories from observed Messages, return all when k=0"""
already_observed = self.remember(k)
news = DocList[Info]()
for i in observed:
if i.id in already_observed.id:
continue
news.append(i)
return news
def remember_by_action(self, action: str) -> DocList[Info]:
"""Return all messages triggered by a specified Action"""
storage_index = InMemoryExactNNIndex[Info]()
storage_index.index(self)
query = {'cause_by': {'$eq': action}}
content = storage_index.filter(query)
return content
def remember_by_actions(self, actions: Iterable[str]) -> DocList[Info]:
"""Return all messages triggered by specified Actions"""
contents = DocList[Info]()
for action in actions:
storage_index = InMemoryExactNNIndex[Info]()
storage_index.index(self)
query = {'cause_by': {'$eq': action}}
contents = contents + storage_index.filter(query) # become a list after + operation
return DocList[Info](contents)class Info(BaseDoc):
content: List = []
instruction: str = ''
agent_id: str = '' # the profile of the agent
role: str = 'user' # system / user / assistant
cause_by: str = ''
@property
def Info_str(self):
# prefix = '-'.join([self.role, str(self.cause_by)])
return f"{self.role}: {self.content}"
def Info_str_repr(self):
return self.Info_str()
def to_dict(self) -> dict:
return {
"role": self.role,
"content": self.content
}Environment
Python 3.8
Jina>3.21.0
How are these Docs chained together? Can you share a Topology of Executors using them as Input and Outulput? you do not need to share what they actually do, just the endpoints and type annotations and the Flow structure.
What is the version of docarray and Pydantic in your system?
Ok, wait for good news, I will use Jina<3.21.0 temporarily.
@ZhihaoAIRobotic the patch has been releassd. Thanks for opening the issue