Spaces:
Sleeping
Sleeping
sourabhzanwar
commited on
Commit
•
3cc9efa
1
Parent(s):
77a47ea
updateed app.py, added file to check if v1 is already changed
Browse files- .DS_Store +0 -0
- .gitignore +4 -0
- .streamlit/config.toml +6 -0
- README.md +105 -6
- app.py +249 -0
- change_log.txt +1 -0
- ml_logo.png +0 -0
- requirements.txt +7 -0
- utils/config.py +41 -0
- utils/haystack.py +120 -0
- utils/ui.py +16 -0
.DS_Store
ADDED
Binary file (6.15 kB). View file
|
|
.gitignore
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
.env
|
2 |
+
.vscode
|
3 |
+
.idea
|
4 |
+
*.pyc
|
.streamlit/config.toml
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
[theme]
|
2 |
+
primaryColor = "#E694FF"
|
3 |
+
backgroundColor = "#FFFFFF"
|
4 |
+
secondaryBackgroundColor = "#F0F0F0"
|
5 |
+
textColor = "#262730"
|
6 |
+
font = "sans-serif"
|
README.md
CHANGED
@@ -1,12 +1,111 @@
|
|
1 |
---
|
2 |
-
title:
|
3 |
-
emoji:
|
4 |
-
colorFrom:
|
5 |
-
colorTo:
|
6 |
sdk: streamlit
|
7 |
-
sdk_version: 1.
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
---
|
11 |
|
12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
title: Haystack Search Pipeline with Streamlit
|
3 |
+
emoji: 👑
|
4 |
+
colorFrom: indigo
|
5 |
+
colorTo: indigo
|
6 |
sdk: streamlit
|
7 |
+
sdk_version: 1.23.0
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
---
|
11 |
|
12 |
+
# Template Streamlit App for Haystack Search Pipelines
|
13 |
+
|
14 |
+
This template [Streamlit](https://docs.streamlit.io/) app set up for simple [Haystack search applications](https://docs.haystack.deepset.ai/docs/semantic_search). The template is ready to do QA with **Retrievel Augmented Generation**, or **Ectractive QA**
|
15 |
+
|
16 |
+
See the ['How to use this template'](#how-to-use-this-template) instructions below to create a simple UI for your own Haystack search pipelines.
|
17 |
+
|
18 |
+
Below you will also find instructions on how you could [push this to Hugging Face Spaces 🤗](#pushing-to-hugging-face-spaces-).
|
19 |
+
|
20 |
+
## Installation and Running
|
21 |
+
To run the bare application which does _nothing_:
|
22 |
+
1. Install requirements: `pip install -r requirements.txt`
|
23 |
+
2. Run the streamlit app: `streamlit run app.py`
|
24 |
+
|
25 |
+
This will start up the app on `localhost:8501` where you will find a simple search bar. Before you start editing, you'll notice that the app will only show you instructions on what to edit.
|
26 |
+
|
27 |
+
### Optional Configurations
|
28 |
+
|
29 |
+
You can set optional cofigurations to set the:
|
30 |
+
- `--task` you want to start the app with: `rag` or `extractive` (default: rag)
|
31 |
+
- `--store` you want to use: `inmemory`, `opensearch`, `weaviate` or `milvus` (default: inmemory)
|
32 |
+
- `--name` you want to have for the app. (default: 'My Search App')
|
33 |
+
|
34 |
+
E.g.:
|
35 |
+
|
36 |
+
```bash
|
37 |
+
streamlit run app.py -- --store opensearch --task extractive --name 'My Opensearch Documentation Search'
|
38 |
+
```
|
39 |
+
|
40 |
+
In a `.env` file, include all the config settings that you would like to use based on:
|
41 |
+
- The DocumentStore of your choice
|
42 |
+
- The Extractive/Generative model of your choice
|
43 |
+
|
44 |
+
While the `/utils/config.py` will create default values for some configurations, others have to be set in the `.env` such as the `OPENAI_KEY`
|
45 |
+
|
46 |
+
Example `.env`
|
47 |
+
|
48 |
+
```
|
49 |
+
OPENAI_KEY=YOUR_KEY
|
50 |
+
EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L12-v2
|
51 |
+
GENERATIVE_MODEL=text-davinci-003
|
52 |
+
```
|
53 |
+
|
54 |
+
|
55 |
+
## How to use this template
|
56 |
+
1. Create a new repository from this template or simply open it in a codespace to start playing around 💙
|
57 |
+
2. Make sure your `requirements.txt` file includes the Haystack and Streamlit versions you would like to use.
|
58 |
+
3. Change the code in `utils/haystack.py` if you would like a different pipeline.
|
59 |
+
4. Create a `.env`file with all of your configuration settings.
|
60 |
+
5. Make any UI edits you'd like to and [share with the Haystack community](https://haystack.deepeset.ai/community)
|
61 |
+
6. Run the app as show in [installation and running](#installation-and-running)
|
62 |
+
|
63 |
+
### Repo structure
|
64 |
+
- `./utils`: This is where we have 3 files:
|
65 |
+
- `config.py`: This file extracts all of the configuration settings from a `.env` file. For some config settings, it uses default values. An example of this is in [this demo project](https://github.com/TuanaCelik/should-i-follow/blob/main/utils/config.py).
|
66 |
+
- `haystack.py`: Here you will find some functions already set up for you to start creating your Haystack search pipeline. It includes 2 main functions called `start_haystack()` which is what we use to create a pipeline and cache it, and `query()` which is the function called by `app.py` once a user query is received.
|
67 |
+
- `ui.py`: Use this file for any UI and initial value setups.
|
68 |
+
- `app.py`: This is the main Streamlit application file that we will run. In its current state it has a simple search bar, a 'Run' button, and a response that you can highlight answers with.
|
69 |
+
|
70 |
+
### What to edit?
|
71 |
+
There are default pipelines both in `start_haystack_extractive()` and `start_haystack_rag()`
|
72 |
+
|
73 |
+
- Change the pipelines to use the embedding models, extractive or generative models as you need.
|
74 |
+
- If using the `rag` task, change the `default_prompt_template` to use one of our available ones on [PromptHub](https://prompthub.deepset.ai) or create your own `PromptTemplate`
|
75 |
+
|
76 |
+
|
77 |
+
## Pushing to Hugging Face Spaces 🤗
|
78 |
+
|
79 |
+
Below is an example GitHub action that will let you push your Streamlit app straight to the Hugging Face Hub as a Space.
|
80 |
+
|
81 |
+
A few things to pay attention to:
|
82 |
+
|
83 |
+
1. Create a New Space on Hugging Face with the Streamlit SDK.
|
84 |
+
2. Create a Hugging Face token on your HF account.
|
85 |
+
3. Create a secret on your GitHub repo called `HF_TOKEN` and put your Hugging Face token here.
|
86 |
+
4. If you're using DocumentStores or APIs that require some keys/tokens, make sure these are provided as a secret for your HF Space too!
|
87 |
+
5. This readme is set up to tell HF spaces that it's using streamlit and that the app is running on `app.py`, make any changes to the frontmatter of this readme to display the title, emoji etc you desire.
|
88 |
+
6. Create a file in `.github/workflows/hf_sync.yml`. Here's an example that you can change with your own information, and an [example workflow](https://github.com/TuanaCelik/should-i-follow/blob/main/.github/workflows/hf_sync.yml) working for the [Should I Follow demo](https://huggingface.co/spaces/deepset/should-i-follow)
|
89 |
+
|
90 |
+
```yaml
|
91 |
+
name: Sync to Hugging Face hub
|
92 |
+
on:
|
93 |
+
push:
|
94 |
+
branches: [main]
|
95 |
+
|
96 |
+
# to run this workflow manually from the Actions tab
|
97 |
+
workflow_dispatch:
|
98 |
+
|
99 |
+
jobs:
|
100 |
+
sync-to-hub:
|
101 |
+
runs-on: ubuntu-latest
|
102 |
+
steps:
|
103 |
+
- uses: actions/checkout@v2
|
104 |
+
with:
|
105 |
+
fetch-depth: 0
|
106 |
+
lfs: true
|
107 |
+
- name: Push to hub
|
108 |
+
env:
|
109 |
+
HF_TOKEN: ${{ secrets.HF_TOKEN }}
|
110 |
+
run: git push --force https://{YOUR_HF_USERNAME}:$HF_TOKEN@{YOUR_HF_SPACE_REPO} main
|
111 |
+
```
|
app.py
ADDED
@@ -0,0 +1,249 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import pydantic
|
2 |
+
module_file_path = pydantic.__file__
|
3 |
+
|
4 |
+
module_file_path = module_file_path.split('pydantic')[0] + 'haystack'
|
5 |
+
|
6 |
+
import os
|
7 |
+
import fileinput
|
8 |
+
|
9 |
+
|
10 |
+
def replace_string_in_files(folder_path, old_str, new_str):
|
11 |
+
for subdir, dirs, files in os.walk(folder_path):
|
12 |
+
for file in files:
|
13 |
+
file_path = os.path.join(subdir, file)
|
14 |
+
|
15 |
+
# Check if the file is a text file (you can modify this condition based on your needs)
|
16 |
+
if file.endswith(".txt") or file.endswith(".py"):
|
17 |
+
# Open the file in place for editing
|
18 |
+
with fileinput.FileInput(file_path, inplace=True) as f:
|
19 |
+
for line in f:
|
20 |
+
# Replace the old string with the new string
|
21 |
+
print(line.replace(old_str, new_str), end='')
|
22 |
+
|
23 |
+
with open('change_log.txt','r') as f:
|
24 |
+
status = f.readlines()
|
25 |
+
|
26 |
+
if status[-1] != 'changed':
|
27 |
+
replace_string_in_files(module_file_path, 'from pydantic', 'from pydantic.v1')
|
28 |
+
with open('change_log.txt','w'):
|
29 |
+
f.write('changed')
|
30 |
+
|
31 |
+
|
32 |
+
|
33 |
+
|
34 |
+
from operator import index
|
35 |
+
import streamlit as st
|
36 |
+
import logging
|
37 |
+
import os
|
38 |
+
|
39 |
+
from annotated_text import annotation
|
40 |
+
from json import JSONDecodeError
|
41 |
+
from markdown import markdown
|
42 |
+
from utils.config import parser
|
43 |
+
from utils.haystack import start_document_store, query, initialize_pipeline, start_preprocessor_node, start_retriever, start_reader
|
44 |
+
from utils.ui import reset_results, set_initial_state
|
45 |
+
import pandas as pd
|
46 |
+
import haystack
|
47 |
+
|
48 |
+
|
49 |
+
# Whether the file upload should be enabled or not
|
50 |
+
DISABLE_FILE_UPLOAD = bool(os.getenv("DISABLE_FILE_UPLOAD"))
|
51 |
+
# Define a function to handle file uploads
|
52 |
+
def upload_files():
|
53 |
+
uploaded_files = st.sidebar.file_uploader(
|
54 |
+
"upload", type=["pdf", "txt", "docx"], accept_multiple_files=True, label_visibility="hidden"
|
55 |
+
)
|
56 |
+
return uploaded_files
|
57 |
+
|
58 |
+
# Define a function to process a single file
|
59 |
+
|
60 |
+
def process_file(data_file, preprocesor, document_store):
|
61 |
+
# read file and add content
|
62 |
+
file_contents = data_file.read().decode("utf-8")
|
63 |
+
docs = [{
|
64 |
+
'content': str(file_contents),
|
65 |
+
'meta': {'name': str(data_file.name)}
|
66 |
+
}]
|
67 |
+
try:
|
68 |
+
names = [item.meta.get('name') for item in document_store.get_all_documents()]
|
69 |
+
#if args.store == 'inmemory':
|
70 |
+
# doc = converter.convert(file_path=files, meta=None)
|
71 |
+
if data_file.name in names:
|
72 |
+
print(f"{data_file.name} already processed")
|
73 |
+
else:
|
74 |
+
print(f'preprocessing uploaded doc {data_file.name}.......')
|
75 |
+
#print(data_file.read().decode("utf-8"))
|
76 |
+
preprocessed_docs = preprocesor.process(docs)
|
77 |
+
print('writing to document store.......')
|
78 |
+
document_store.write_documents(preprocessed_docs)
|
79 |
+
print('updating emebdding.......')
|
80 |
+
document_store.update_embeddings(retriever)
|
81 |
+
except Exception as e:
|
82 |
+
print(e)
|
83 |
+
|
84 |
+
try:
|
85 |
+
args = parser.parse_args()
|
86 |
+
preprocesor = start_preprocessor_node()
|
87 |
+
document_store = start_document_store(type=args.store)
|
88 |
+
retriever = start_retriever(document_store)
|
89 |
+
reader = start_reader()
|
90 |
+
st.set_page_config(
|
91 |
+
page_title="MLReplySearch",
|
92 |
+
layout="centered",
|
93 |
+
page_icon=":shark:",
|
94 |
+
menu_items={
|
95 |
+
'Get Help': 'https://www.extremelycoolapp.com/help',
|
96 |
+
'Report a bug': "https://www.extremelycoolapp.com/bug",
|
97 |
+
'About': "# This is a header. This is an *extremely* cool app!"
|
98 |
+
}
|
99 |
+
)
|
100 |
+
st.sidebar.image("ml_logo.png", use_column_width=True)
|
101 |
+
|
102 |
+
# Sidebar for Task Selection
|
103 |
+
st.sidebar.header('Options:')
|
104 |
+
|
105 |
+
# OpenAI Key Input
|
106 |
+
openai_key = st.sidebar.text_input("Enter OpenAI Key:", type="password")
|
107 |
+
|
108 |
+
if openai_key:
|
109 |
+
task_options = ['Extractive', 'Generative']
|
110 |
+
else:
|
111 |
+
task_options = ['Extractive']
|
112 |
+
|
113 |
+
task_selection = st.sidebar.radio('Select the task:', task_options)
|
114 |
+
|
115 |
+
# Check the task and initialize pipeline accordingly
|
116 |
+
if task_selection == 'Extractive':
|
117 |
+
pipeline_extractive = initialize_pipeline("extractive", document_store, retriever, reader)
|
118 |
+
elif task_selection == 'Generative' and openai_key: # Check for openai_key to ensure user has entered it
|
119 |
+
pipeline_rag = initialize_pipeline("rag", document_store, retriever, reader, openai_key=openai_key)
|
120 |
+
|
121 |
+
|
122 |
+
set_initial_state()
|
123 |
+
|
124 |
+
st.write('# ' + args.name)
|
125 |
+
|
126 |
+
|
127 |
+
# File upload block
|
128 |
+
if not DISABLE_FILE_UPLOAD:
|
129 |
+
st.sidebar.write("## File Upload:")
|
130 |
+
#data_files = st.sidebar.file_uploader(
|
131 |
+
# "upload", type=["pdf", "txt", "docx"], accept_multiple_files=True, label_visibility="hidden"
|
132 |
+
#)
|
133 |
+
data_files = upload_files()
|
134 |
+
if data_files is not None:
|
135 |
+
for data_file in data_files:
|
136 |
+
# Upload file
|
137 |
+
if data_file:
|
138 |
+
try:
|
139 |
+
#raw_json = upload_doc(data_file)
|
140 |
+
# Call the process_file function for each uploaded file
|
141 |
+
if args.store == 'inmemory':
|
142 |
+
processed_data = process_file(data_file, preprocesor, document_store)
|
143 |
+
st.sidebar.write(str(data_file.name) + " ✅ ")
|
144 |
+
except Exception as e:
|
145 |
+
st.sidebar.write(str(data_file.name) + " ❌ ")
|
146 |
+
st.sidebar.write("_This file could not be parsed, see the logs for more information._")
|
147 |
+
|
148 |
+
if "question" not in st.session_state:
|
149 |
+
st.session_state.question = ""
|
150 |
+
# Search bar
|
151 |
+
question = st.text_input("", value=st.session_state.question, max_chars=100, on_change=reset_results)
|
152 |
+
|
153 |
+
run_pressed = st.button("Run")
|
154 |
+
|
155 |
+
run_query = (
|
156 |
+
run_pressed or question != st.session_state.question #or task_selection != st.session_state.task
|
157 |
+
)
|
158 |
+
|
159 |
+
# Get results for query
|
160 |
+
if run_query and question:
|
161 |
+
if task_selection == 'Extractive':
|
162 |
+
reset_results()
|
163 |
+
st.session_state.question = question
|
164 |
+
with st.spinner("🔎 Running your pipeline"):
|
165 |
+
try:
|
166 |
+
st.session_state.results_extractive = query(pipeline_extractive, question)
|
167 |
+
st.session_state.task = task_selection
|
168 |
+
except JSONDecodeError as je:
|
169 |
+
st.error(
|
170 |
+
"👓 An error occurred reading the results. Is the document store working?"
|
171 |
+
)
|
172 |
+
except Exception as e:
|
173 |
+
logging.exception(e)
|
174 |
+
st.error("🐞 An error occurred during the request.")
|
175 |
+
|
176 |
+
elif task_selection == 'Generative':
|
177 |
+
reset_results()
|
178 |
+
st.session_state.question = question
|
179 |
+
with st.spinner("🔎 Running your pipeline"):
|
180 |
+
try:
|
181 |
+
st.session_state.results_generative = query(pipeline_rag, question)
|
182 |
+
st.session_state.task = task_selection
|
183 |
+
except JSONDecodeError as je:
|
184 |
+
st.error(
|
185 |
+
"👓 An error occurred reading the results. Is the document store working?"
|
186 |
+
)
|
187 |
+
except Exception as e:
|
188 |
+
if "API key is invalid" in str(e):
|
189 |
+
logging.exception(e)
|
190 |
+
st.error("🐞 incorrect API key provided. You can find your API key at https://platform.openai.com/account/api-keys.")
|
191 |
+
else:
|
192 |
+
logging.exception(e)
|
193 |
+
st.error("🐞 An error occurred during the request.")
|
194 |
+
# Display results
|
195 |
+
if (st.session_state.results_extractive or st.session_state.results_generative) and run_query:
|
196 |
+
|
197 |
+
# Handle Extractive Answers
|
198 |
+
if task_selection == 'Extractive':
|
199 |
+
results = st.session_state.results_extractive
|
200 |
+
|
201 |
+
st.subheader("Extracted Answers:")
|
202 |
+
|
203 |
+
if 'answers' in results:
|
204 |
+
answers = results['answers']
|
205 |
+
treshold = 0.2
|
206 |
+
higher_then_treshold = any(ans.score > treshold for ans in answers)
|
207 |
+
if not higher_then_treshold:
|
208 |
+
st.markdown(f"<span style='color:red'>Please note none of the answers achieved a score higher then {int(treshold) * 100}%. Which probably means that the desired answer is not in the searched documents.</span>", unsafe_allow_html=True)
|
209 |
+
for count, answer in enumerate(answers):
|
210 |
+
if answer.answer:
|
211 |
+
text, context = answer.answer, answer.context
|
212 |
+
start_idx = context.find(text)
|
213 |
+
end_idx = start_idx + len(text)
|
214 |
+
score = round(answer.score, 3)
|
215 |
+
st.markdown(f"**Answer {count + 1}:**")
|
216 |
+
st.markdown(
|
217 |
+
context[:start_idx] + str(annotation(body=text, label=f'SCORE {score}', background='#964448', color='#ffffff')) + context[end_idx:],
|
218 |
+
unsafe_allow_html=True,
|
219 |
+
)
|
220 |
+
else:
|
221 |
+
st.info(
|
222 |
+
"🤔 Haystack is unsure whether any of the documents contain an answer to your question. Try to reformulate it!"
|
223 |
+
)
|
224 |
+
|
225 |
+
# Handle Generative Answers
|
226 |
+
elif task_selection == 'Generative':
|
227 |
+
results = st.session_state.results_generative
|
228 |
+
st.subheader("Generated Answer:")
|
229 |
+
if 'results' in results:
|
230 |
+
st.markdown("**Answer:**")
|
231 |
+
st.write(results['results'][0])
|
232 |
+
|
233 |
+
# Handle Retrieved Documents
|
234 |
+
if 'documents' in results:
|
235 |
+
retrieved_documents = results['documents']
|
236 |
+
st.subheader("Retriever Results:")
|
237 |
+
|
238 |
+
data = []
|
239 |
+
for i, document in enumerate(retrieved_documents):
|
240 |
+
# Truncate the content
|
241 |
+
truncated_content = (document.content[:150] + '...') if len(document.content) > 150 else document.content
|
242 |
+
data.append([i + 1, document.meta['name'], truncated_content])
|
243 |
+
|
244 |
+
# Convert data to DataFrame and display using Streamlit
|
245 |
+
df = pd.DataFrame(data, columns=['Ranked Context', 'Document Name', 'Content'])
|
246 |
+
st.table(df)
|
247 |
+
|
248 |
+
except SystemExit as e:
|
249 |
+
os._exit(e.code)
|
change_log.txt
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
unchanged
|
ml_logo.png
ADDED
requirements.txt
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
safetensors==0.3.3.post1
|
2 |
+
farm-haystack[inference,weaviate,opensearch]==1.20.0
|
3 |
+
milvus-haystack
|
4 |
+
streamlit==1.23.0
|
5 |
+
markdown
|
6 |
+
st-annotated-text
|
7 |
+
datasets
|
utils/config.py
ADDED
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import argparse
|
2 |
+
import os
|
3 |
+
import os
|
4 |
+
from dotenv import load_dotenv
|
5 |
+
|
6 |
+
load_dotenv()
|
7 |
+
parser = argparse.ArgumentParser(description='This app lists animals')
|
8 |
+
|
9 |
+
document_store_choices = ('inmemory', 'weaviate', 'milvus', 'opensearch')
|
10 |
+
parser.add_argument('--store', choices=document_store_choices, default='inmemory', help='DocumentStore selection (default: %(default)s)')
|
11 |
+
parser.add_argument('--name', default="My Search App")
|
12 |
+
|
13 |
+
model_configs = {
|
14 |
+
'EMBEDDING_MODEL': os.getenv("EMBEDDING_MODEL", "sentence-transformers/all-MiniLM-L12-v2"),
|
15 |
+
'GENERATIVE_MODEL': os.getenv("GENERATIVE_MODEL", "gpt-4"),
|
16 |
+
'EXTRACTIVE_MODEL': os.getenv("EXTRACTIVE_MODEL", "deepset/roberta-base-squad2"),
|
17 |
+
'OPENAI_KEY': os.getenv("OPENAI_KEY"),
|
18 |
+
'COHERE_KEY': os.getenv("COHERE_KEY"),
|
19 |
+
}
|
20 |
+
|
21 |
+
document_store_configs = {
|
22 |
+
# Weaviate Config
|
23 |
+
'WEAVIATE_HOST': os.getenv("WEAVIATE_HOST", "http://localhost"),
|
24 |
+
'WEAVIATE_PORT': os.getenv("WEAVIATE_PORT", 8080),
|
25 |
+
'WEAVIATE_INDEX': os.getenv("WEAVIATE_INDEX", "Document"),
|
26 |
+
'WEAVIATE_EMBEDDING_DIM': os.getenv("WEAVIATE_EMBEDDING_DIM", 768),
|
27 |
+
|
28 |
+
# OpenSearch Config
|
29 |
+
'OPENSEARCH_SCHEME': os.getenv("OPENSEARCH_SCHEME", "https"),
|
30 |
+
'OPENSEARCH_USERNAME': os.getenv("OPENSEARCH_USERNAME", "admin"),
|
31 |
+
'OPENSEARCH_PASSWORD': os.getenv("OPENSEARCH_PASSWORD", "admin"),
|
32 |
+
'OPENSEARCH_HOST': os.getenv("OPENSEARCH_HOST", "localhost"),
|
33 |
+
'OPENSEARCH_PORT': os.getenv("OPENSEARCH_PORT", 9200),
|
34 |
+
'OPENSEARCH_INDEX': os.getenv("OPENSEARCH_INDEX", "document"),
|
35 |
+
'OPENSEARCH_EMBEDDING_DIM': os.getenv("OPENSEARCH_EMBEDDING_DIM", 768),
|
36 |
+
|
37 |
+
# Milvus Config
|
38 |
+
'MILVUS_URI': os.getenv("MILVUS_URI", "http://localhost:19530/default"),
|
39 |
+
'MILVUS_INDEX': os.getenv("MILVUS_INDEX", "document"),
|
40 |
+
'MILVUS_EMBEDDING_DIM': os.getenv("MILVUS_EMBEDDING_DIM", 768),
|
41 |
+
}
|
utils/haystack.py
ADDED
@@ -0,0 +1,120 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
|
3 |
+
from utils.config import document_store_configs, model_configs
|
4 |
+
from haystack import Pipeline
|
5 |
+
from haystack.schema import Answer
|
6 |
+
from haystack.document_stores import BaseDocumentStore
|
7 |
+
from haystack.document_stores import InMemoryDocumentStore, OpenSearchDocumentStore, WeaviateDocumentStore
|
8 |
+
from haystack.nodes import EmbeddingRetriever, FARMReader, PromptNode, PreProcessor
|
9 |
+
from milvus_haystack import MilvusDocumentStore
|
10 |
+
#Use this file to set up your Haystack pipeline and querying
|
11 |
+
|
12 |
+
@st.cache_resource(show_spinner=False)
|
13 |
+
def start_preprocessor_node():
|
14 |
+
print('initializing preprocessor node')
|
15 |
+
processor = PreProcessor(
|
16 |
+
clean_empty_lines= True,
|
17 |
+
clean_whitespace=True,
|
18 |
+
clean_header_footer=True,
|
19 |
+
#remove_substrings=None,
|
20 |
+
split_by="word",
|
21 |
+
split_length=100,
|
22 |
+
split_respect_sentence_boundary=True,
|
23 |
+
#split_overlap=0,
|
24 |
+
#max_chars_check= 10_000
|
25 |
+
)
|
26 |
+
return processor
|
27 |
+
#return docs
|
28 |
+
|
29 |
+
@st.cache_resource(show_spinner=False)
|
30 |
+
def start_document_store(type: str):
|
31 |
+
#This function starts the documents store of your choice based on your command line preference
|
32 |
+
print('initializing document store')
|
33 |
+
if type == 'inmemory':
|
34 |
+
document_store = InMemoryDocumentStore(use_bm25=True, embedding_dim=384)
|
35 |
+
'''
|
36 |
+
documents = [
|
37 |
+
{
|
38 |
+
'content': "Pi is a super dog",
|
39 |
+
'meta': {'name': "pi.txt"}
|
40 |
+
},
|
41 |
+
{
|
42 |
+
'content': "The revenue of siemens is 5 milion Euro",
|
43 |
+
'meta': {'name': "siemens.txt"}
|
44 |
+
},
|
45 |
+
]
|
46 |
+
document_store.write_documents(documents)
|
47 |
+
'''
|
48 |
+
elif type == 'opensearch':
|
49 |
+
document_store = OpenSearchDocumentStore(scheme = document_store_configs['OPENSEARCH_SCHEME'],
|
50 |
+
username = document_store_configs['OPENSEARCH_USERNAME'],
|
51 |
+
password = document_store_configs['OPENSEARCH_PASSWORD'],
|
52 |
+
host = document_store_configs['OPENSEARCH_HOST'],
|
53 |
+
port = document_store_configs['OPENSEARCH_PORT'],
|
54 |
+
index = document_store_configs['OPENSEARCH_INDEX'],
|
55 |
+
embedding_dim = document_store_configs['OPENSEARCH_EMBEDDING_DIM'])
|
56 |
+
elif type == 'weaviate':
|
57 |
+
document_store = WeaviateDocumentStore(host = document_store_configs['WEAVIATE_HOST'],
|
58 |
+
port = document_store_configs['WEAVIATE_PORT'],
|
59 |
+
index = document_store_configs['WEAVIATE_INDEX'],
|
60 |
+
embedding_dim = document_store_configs['WEAVIATE_EMBEDDING_DIM'])
|
61 |
+
elif type == 'milvus':
|
62 |
+
document_store = MilvusDocumentStore(uri = document_store_configs['MILVUS_URI'],
|
63 |
+
index = document_store_configs['MILVUS_INDEX'],
|
64 |
+
embedding_dim = document_store_configs['MILVUS_EMBEDDING_DIM'],
|
65 |
+
return_embedding=True)
|
66 |
+
return document_store
|
67 |
+
|
68 |
+
# cached to make index and models load only at start
|
69 |
+
@st.cache_resource(show_spinner=False)
|
70 |
+
def start_retriever(_document_store: BaseDocumentStore):
|
71 |
+
print('initializing retriever')
|
72 |
+
retriever = EmbeddingRetriever(document_store=_document_store,
|
73 |
+
embedding_model=model_configs['EMBEDDING_MODEL'],
|
74 |
+
top_k=5)
|
75 |
+
#
|
76 |
+
|
77 |
+
#_document_store.update_embeddings(retriever)
|
78 |
+
return retriever
|
79 |
+
|
80 |
+
|
81 |
+
@st.cache_resource(show_spinner=False)
|
82 |
+
def start_reader():
|
83 |
+
print('initializing reader')
|
84 |
+
reader = FARMReader(model_name_or_path=model_configs['EXTRACTIVE_MODEL'])
|
85 |
+
return reader
|
86 |
+
|
87 |
+
|
88 |
+
|
89 |
+
# cached to make index and models load only at start
|
90 |
+
@st.cache_resource(show_spinner=False)
|
91 |
+
def start_haystack_extractive(_document_store: BaseDocumentStore, _retriever: EmbeddingRetriever, _reader: FARMReader):
|
92 |
+
print('initializing pipeline')
|
93 |
+
pipe = Pipeline()
|
94 |
+
pipe.add_node(component=_retriever, name="Retriever", inputs=["Query"])
|
95 |
+
pipe.add_node(component= _reader, name="Reader", inputs=["Retriever"])
|
96 |
+
return pipe
|
97 |
+
|
98 |
+
@st.cache_resource(show_spinner=False)
|
99 |
+
def start_haystack_rag(_document_store: BaseDocumentStore, _retriever: EmbeddingRetriever, openai_key):
|
100 |
+
prompt_node = PromptNode(default_prompt_template="deepset/question-answering",
|
101 |
+
model_name_or_path=model_configs['GENERATIVE_MODEL'],
|
102 |
+
api_key=openai_key)
|
103 |
+
pipe = Pipeline()
|
104 |
+
|
105 |
+
pipe.add_node(component=_retriever, name="Retriever", inputs=["Query"])
|
106 |
+
pipe.add_node(component=prompt_node, name="PromptNode", inputs=["Retriever"])
|
107 |
+
|
108 |
+
return pipe
|
109 |
+
|
110 |
+
#@st.cache_data(show_spinner=True)
|
111 |
+
def query(_pipeline, question):
|
112 |
+
params = {}
|
113 |
+
results = _pipeline.run(question, params=params)
|
114 |
+
return results
|
115 |
+
|
116 |
+
def initialize_pipeline(task, document_store, retriever, reader, openai_key = ""):
|
117 |
+
if task == 'extractive':
|
118 |
+
return start_haystack_extractive(document_store, retriever, reader)
|
119 |
+
elif task == 'rag':
|
120 |
+
return start_haystack_rag(document_store, retriever, openai_key)
|
utils/ui.py
ADDED
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
|
3 |
+
def set_state_if_absent(key, value):
|
4 |
+
if key not in st.session_state:
|
5 |
+
st.session_state[key] = value
|
6 |
+
|
7 |
+
def set_initial_state():
|
8 |
+
set_state_if_absent("question", "Ask something here?")
|
9 |
+
set_state_if_absent("results_extractive", None)
|
10 |
+
set_state_if_absent("results_generative", None)
|
11 |
+
set_state_if_absent("task", None)
|
12 |
+
|
13 |
+
def reset_results(*args):
|
14 |
+
st.session_state.results_extractive = None
|
15 |
+
st.session_state.results_generative = None
|
16 |
+
st.session_state.task = None
|