Polyhistor

Introduction

Polyhistor is a AI bot that can answer questions from documents.

Installation

We use Poetry to manage dependencies.

poetry shell
poetry install

Create testdata directory in the root directory and add files.

Now we only support .md, .txt, .docx, .pdf files.

You can also import documents from github repositories.

Use pre-commit.

pre-commit install

Create a .env file in the root directory and add the following environment variables:

MODE="dev"
LANGCHAIN_TRACING_V2=true
LANGCHAIN_API_KEY=<LANGCHAIN_API_KEY>
MODEL_API=<the model api from Mattermost>
ACCESS_TOKEN=<Github access token>
MATTERMOST_TOKEN=<Mattermost token>

Note:

1. You can get the LANGCHAIN_API_KEY from LangSmith.
2. Get the MODEL_API from Mattermost.
3. (Optional) How to get Github access token: Github Docs
4. (Optional) How to get Mattermost token: Mattermost Docs

Usage

Qdrant

Default port:6333

If you want to change the port, please modify the docker-compose.yml and config.yaml.

cd vectorDB
docker-compose up -d

Create Qdrant collection

python vectorDB/create_collection.py

Import documents to vectorDB

python vectorDB/read_docs.py --source <source_type>

• <source_type>: local or github

If you choose github, you need to add ACCESS_TOKEN in .env

Imort Mattermost messages to vectorDB

python vectorDB/read_mattermost.py

You need to add MATTERMOST_TOKEN in .env It will import the messages from specific users.

Backend

poetry run uvicorn index:app --reload --host 0.0.0.0 --port 8080

Now you can access the API at http://localhost:8080/api/v1/ask.

Q & A

Open a new terminal and run the following command:

cd src
python main.py

Then you can ask questions to the model. If you want to exit, type bye.

Official

docker

.env 中的 MODE 設定成 "official"

docker-compose up -d --build