Source Code Download
git clone /eosphoros-ai/
Miniconda environment installation
Miniconda Installation
mkdir -p ~/miniconda3
wget /miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/
bash ~/miniconda3/ -b -u -p ~/miniconda3
rm -rf ~/miniconda3/
Initialize Miniconda
~/miniconda3/bin/conda init bash
Enable the conda environment:
source ~/.bashrc
After that you have the conda base environment, you will see the (base) flag added in front of the end user name
Creating a Python Virtual Environment
python >= 3.10
conda create -n dbgpt_env python=3.10
conda activate dbgpt_env #beyondcondaThe environment becomes(dbgpt_env) tom@ubuntu:/home/apps$
# it will take some minutes
pip install -e ".[default]" -i /pypi/simple/
Copying Environment Variables
cp . .env
Model deployment
Download Embedding Model
cd DB-GPT
mkdir models and cd models
git lfs install
git clone /GanymedeNil/text2vec-large-chinese
Downloading the model is slow, use huggingface's mirror site
git clone /GanymedeNil/text2vec-large-chinese
and pytorch_model.bin can be downloaded manually and placed in the text2vec-large-english file directory.
Configuring the Agent
With limited resources, we use the proxy mode of installation here, which is to use the API interface provided by other big model vendors.
The glm-4 model of Wisdom Spectrum is used here
LLM_MODEL=zhipu_proxyllmi
PROXYLLM_BACKEND=glm-4
EMBEDDING_MODEL=text2vec
PROXY_SERVER_URL=/api/paas/v4/chat/completions
ZHIPU_PROXY_API_KEY=98141404f70e85263ef90ebd330e888c.DciXUwNK8b6R7aQm
Test data
Load the default test data into the SQLite database
bash ./scripts/examples/load_examples.sh
Operational services
python dbgpt/app/dbgpt_server.py
interviews
http://localhost:5670