(I get errors similar to these)

Plain Text

User: Hey
An error occurred: 'list' object has no attribute 'query_str'
(.venv) gavsgav@buntu-white:~/AthenaIndex/FlaskAPP$ /home/gavsgav/AthenaIndex/.venv/bin/python /home/gavsgav/AthenaIndex/StreamTest.py
User: Hey
An error occurred: name 'chat_engine' is not defined
(.venv) gavsgav@buntu-white:~/AthenaIndex/FlaskAPP$ /home/gavsgav/AthenaIndex/.venv/bin/python /home/gavsgav/AthenaIndex/StreamTest.py
Traceback (most recent call last):
  File "/home/gavsgav/AthenaIndex/StreamTest.py", line 28, in <module>
    chat_engine = index.as_chat_engine(chat_prompt=message, service_context=service_context, chat_mode="condense_plus_context", memory=memory, verbose=False)
NameError: name 'message' is not defined
(.venv) gavsgav@buntu-white:~/AthenaIndex/FlaskAPP$ /home/gavsgav/AthenaIndex/.venv/bin/python /home/gavsgav/AthenaIndex/StreamTest.py
User: Hey
An error occurred: 'list' object has no attribute 'query_str'
(.venv) gavsgav@buntu-white:~/AthenaIndex/FlaskAPP$ /home/gavsgav/AthenaIndex/.venv/bin/python /home/gavsgav/AthenaIndex/StreamTest.py
User: Hey
An error occurred: 'AgentChatResponse' object has no attribute 'format'
(.venv) gavsgav@buntu-white:~/AthenaIndex/FlaskAPP$ /home/gavsgav/AthenaIndex/.venv/bin/python /home/gavsgav/AthenaIndex/StreamTest.py
User: Hey.
An error occurred: 'AgentChatResponse' object has no attribute 'format'

14 comments

LLogan M

Can you share your code? Just seems like a few things not being used properly

LLirenco

Plain Text

import os
import sys
from llama_index import VectorStoreIndex, ServiceContext
from llama_index.prompts import PromptTemplate  
from llama_index.llms import Ollama, ChatMessage, MessageRole
from llama_index.chat_engine.condense_question import (
    CondenseQuestionChatEngine,
)
from llama_index import SimpleDirectoryReader
from pathlib import Path
from llama_index.embeddings import OptimumEmbedding, HuggingFaceEmbedding

OptimumEmbedding.create_and_save_optimum_model(
    "BAAI/bge-small-en-v1.5", "./bge_onnx"
)
embed_model = OptimumEmbedding(folder_name="./bge_onnx")

promt_path = Path("~/AthenaIndex/prompts/").expanduser()

documents = SimpleDirectoryReader(promt_path).load_data()

index = VectorStoreIndex.from_documents(documents)
service_context = ServiceContext.from_defaults(embed_model=embed_model, llm = Ollama(model="dolphin2.2-mistral:7b-q8_0", base_url="http://10.252.0.216:11434"))


file_path = Path("~/AthenaIndex/personas/Athena.txt").expanduser()
persona = open(file_path, "r").read()

while True:
    text_input = input("User: ")
    if text_input == "exit":
        break
    messages = [
        ChatMessage(
            role="system", content=persona, metadata={"persona": os.path.basename(file_path)}
        ),
        ChatMessage(role=MessageRole.USER, content=text_input),
    ]
    chat_engine = index.as_chat_engine(chat_mode="context", service_context=service_context)
    resp = chat_engine.stream_chat(messages)
    print(resp)

Not the best. 😄

LLirenco

Its all over the place atm because ive been trying to add/take away different things trying to get it to work

LLogan M

nah it seems pretty good!

So with this code, what is the error you get now? The only thing I noticed is that you are using stream_chat() but not streaming the result

I would do

Plain Text

resp = chat_engine.stream_chat(messages)
for token in resp.response_gen:
  print(token, end="")

LLirenco

That seems to have work. Thank you! I just need to figure out the memory now. (I do really appreciate you getting back to me. )

LLirenco

It was the query str error

LLirenco

Ok maybe not sorted ... i added

Plain Text

resp = chat_engine.stream_chat(messages[0].content)
    for token in resp.response_gen:
        print(token, end="")

I added [0].content For it to work but now the Agent is talking to itself 😄

LLirenco

With out the [0].content It gives me the error

LLogan M

stream chat expects a ChatMessage object -- I would do resp = chat_engine.stream_chat([messages[0]])

LLogan M

oh wait

LLogan M

this is a chat engine, I'm thinking of another API

LLogan M

Yea ok, for this, you'll want to do two things

LLogan M

resp = chat_engine.stream_chat("hello")

This needs to be a string. The chat engine will keep track of the chat history itself (which you can access with chat_engine.chat_history)

If you want to override the chat history, you can pass in a list of ChatMessage objects like

Plain Text

resp = chat_engine.stream_chat("Hello", chat_history=chat_history)

Hopefull that makes sense?

LLirenco

I think so. I'll have a play and call back if i have any issues if thats ok?

Add a reply

Find answers from the community

(I get errors similar to these)