Skip to content

Latest commit

 

History

History
164 lines (119 loc) · 7.71 KB

README.md

File metadata and controls

164 lines (119 loc) · 7.71 KB
Raw

🇨🇳中文 | 🌐English

SmartSearch: Build your own conversational search engine with LLMs

HF Models Contributions welcome License Apache 2.0 python_version GitHub issues Wechat Group

Features

  • Built-in support for LLM, you can build API with local model
  • Support OpenAI LLMs, such as gpt-4
  • Built-in support for search engine, support bing/google/DDGS
  • Customizable pretty UI interface
  • Shareable, cached search results
  • Support for follow-up questions, continuous Q&A
  • Supports query analysis, rewrites queries based on context for precise search

Install requirements

pip install -r requirements.txt

Run

1. Build the front-end web

There are two ways to build the front end:

  1. Download the pre-packaged front-end UI from https://github.com/shibing624/SmartSearch/releases/download/0.1.0/ui.zip and unzip it directly into the project root directory for use.
  2. Use npm to build the front end yourself (requires nodejs version 18 or above)
cd web && npm install && npm run build

Output: The project root directory produces a ui folder, which contains static files for the frontend.

2. Run server with Lepton API

Note

We recommend using the built-in llm and kv functions with Lepton. Running the following commands to set up them automatically.

lep login
python search.py

ok, now your search app running on http://0.0.0.0:8081

  • Here is the demo for our online Colab execution service: demo.ipynb, it's Colab: Open In Colab

Using Search Engine API

you can using Bing Search API or Google Search API.

Using Bing Search API

To use the Bing Web Search API, please visit this link to obtain your Bing subscription key.

export BING_SEARCH_V7_SUBSCRIPTION_KEY=YOUR_BING_SUBSCRIPTION_KEY
BACKEND=BING python search.py

Using Google Search Api

You have three options for Google Search: you can use the SearchApi Google Search API from SearchApi, Serper Google Search API from Serper, or opt for the Programmable Search Engine provided by Google.

For Google Search using SearchApi:

export SEARCHAPI_API_KEY=YOUR_SEARCHAPI_API_KEY
BACKEND=SEARCHAPI python search.py

For Google Search using Serper:

export SERPER_SEARCH_API_KEY=YOUR_SERPER_API_KEY
BACKEND=SERPER python search.py

For Google Search using Programmable Search Engine:

export GOOGLE_SEARCH_API_KEY=YOUR_GOOGLE_SEARCH_API_KEY
export GOOGLE_SEARCH_CX=YOUR_GOOGLE_SEARCH_ENGINE_ID
BACKEND=GOOGLE python search.py

Using OpenAI LLM

If you are pursuing better LLM results, you can use OpenAI's LLM model gpt-4.

export SERPER_SEARCH_API_KEY=YOUR_SERPER_API_KEY
export OPENAI_API_KEY=YOUR_OPENAI_API_KEY
export OPENAI_BASE_URL=https://xxx/v1
BACKEND=SERPER LLM_TYPE=OPENAI LLM_MODEL=gpt-4 python search.py

Deployment Configurations

Here are the configurations you can set for your deployment, see search.py:

  • resource_shape: most of heavy lifting will be done by the LLM server and the search engine API, so you can choose a small resource shape. cpu.small is usually good enough.

Then, set the following environmental variables.

  • BACKEND: the search backend to use. If you don't have bing or google set up, simply use LEPTON to try the demo. Otherwise, do BING, GOOGLE, SERPER, SEARCHAPI, or use DDGS.
  • LLM_TYPE: the LLM type to use. If you are using Lepton, set it to lepton. Otherwise, set it to openai.
  • LLM_MODEL: the LLM model to run. We recommend using mixtral-8x7b, but if you want to experiment other models, you can try the ones hosted on LeptonAI, for example, llama2-70b, llama2-13b, llama2-7b. Note that small models won't work that well.
  • KV_NAME: the Lepton KV to use to store the search results. You can use the default search-with-lepton.
  • RELATED_QUESTIONS: whether to generate related questions. If you set this to true, the search engine will generate related questions for you. Otherwise, it will not.
  • REWRITE_QUESTION: whether to rewrite the question. If you set this to true, the LLM will rewrite the question and send it to the search engine. Otherwise, it will not.
  • GOOGLE_SEARCH_CX: if you are using google, specify the search cx. Otherwise, leave it empty.
  • LEPTON_ENABLE_AUTH_BY_COOKIE: this is to allow web UI access to the deployment. Set it to true.
  • OPENAI_BASE_URL: if you are using OpenAI, you can specify the base url. It is usually https://api.openai.com/v1.
  • ENABLE_HISTORY: whether to enable history. If you set this to true, the LLM will store the search history. Otherwise, it will not.

In addition, you will need to set the following secrets:

  • LEPTON_WORKSPACE_TOKEN: this is required to call Lepton's LLM and KV apis. You can find your workspace token at Settings.
  • BING_SEARCH_V7_SUBSCRIPTION_KEY: if you are using Bing, you need to specify the subscription key. Otherwise it is not needed.
  • GOOGLE_SEARCH_API_KEY: if you are using Google, you need to specify the search api key. Note that you should also specify the cx in the env. If you are not using Google, it is not needed.
  • SEARCHAPI_API_KEY: if you are using SearchApi, a 3rd party Google Search API, you need to specify the api key.
  • OPENAI_API_KEY: if you are using OpenAI, you need to specify the api key.

Todo

  1. Support multi-round retrieval, mainly displaying multi-round retrieval results on the page.
  2. Support third-party LLM's API, such as qwen, baichuan, etc.
  3. Mini program support, currently only supports web end.

Contact

  • Issue(suggestion): GitHub issues
  • Email: xuming: xuming624@qq.com
  • Wechat: Add my Wechat ID: xuming624, note: name-company-NLP to join the NLP discussion group.

License

The license agreement is The Apache License 2.0, which can be used for commercial purposes for free. Please include SmartSearch's link and license agreement in the product description.

Contribute

The project code is still rough, if everyone has improvements to the code, welcome to submit back to this project.

Reference