mirror of
https://github.com/hwchase17/langchain
synced 2024-11-06 03:20:49 +00:00
41 lines
1.0 KiB
Markdown
41 lines
1.0 KiB
Markdown
|
# Extraction with OpenAI Function Calling
|
||
|
|
|
||
|
|
This template shows how to do extraction of structured data from unstructured data, using OpenAI [function calling](https://python.langchain.com/docs/modules/chains/how_to/openai_functions).
|
||
|
|
|
||
|
|
Specify the information you want to extract in `chain.py`
|
||
|
|
|
||
|
|
By default, it will extract the title and author of papers.
|
||
|
|
|
||
|
|
## LLM
|
||
|
|
|
||
|
|
This template will use `OpenAI` by default.
|
||
|
|
|
||
|
|
Be sure that `OPENAI_API_KEY` is set in your environment.
|
||
|
|
|
||
|
|
## Adding the template
|
||
|
|
|
||
|
|
Install the langchain package
|
||
|
|
```
|
||
|
|
pip install -e packages/extraction_openai_functions
|
||
|
|
```
|
||
|
|
|
||
|
|
Edit app/server.py to add that package to the routes
|
||
|
|
```
|
||
|
|
from fastapi import FastAPI
|
||
|
|
from langserve import add_routes
|
||
|
|
from extraction_openai_functions.chain import chain
|
||
|
|
|
||
|
|
app = FastAPI()
|
||
|
|
add_routes(app, chain)
|
||
|
|
```
|
||
|
|
|
||
|
|
Run the app
|
||
|
|
```
|
||
|
|
python app/server.py
|
||
|
|
```
|
||
|
|
|
||
|
|
You can use this template in the Playground:
|
||
|
|
|
||
|
|
http://127.0.0.1:8000/extraction-openai-functions/playground/
|
||
|
|
|
||
|
|
Also, see Jupyter notebook `openai_functions` for various other ways to connect to the template.
|