zilliztech · SimFG · Apr 21, 2023 · Apr 18, 2023 · Apr 21, 2023 · Apr 21, 2023
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-audio/.streamlit/config.toml b/docs/bootcamp/streamlit/gptcache-streamlit-audio/.streamlit/config.toml
@@ -0,0 +1,2 @@
+[theme]
+base="dark"
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-audio/README.md b/docs/bootcamp/streamlit/gptcache-streamlit-audio/README.md
@@ -0,0 +1,21 @@
+# GPTCache Demo for OpenAI Audio Transcription
+This project demonstrates how [GPTcache](https://github.com/zilliztech/GPTCache) can be used to save costs when using OpenAI’s audio transcription API. It provides a simple Streamlit app that allows users to input an audio file and see the corresponding transcribed text. The app uses a cache to store previously generated transcriptions and reuses them for the same audio file, thus avoiding making duplicate API calls.
+
+## Requirements
+* Python 3.6 or later 
+* Dependencies listed in requirements.txt 
+* OpenAI API key
+## Usage
+1. Clone the repository to your local machine
+Install the required packages: pip install -r requirements.txt
+2. Run the app: streamlit run audio.py
+3. Open the app in your browser at http://localhost:8501
+4. Enter your OpenAI API key and upload an audio file to transcribe, then click “generate” to wait for the transcribed text to appear.
+If a cache hit occurred, you should see a message like “cache” at the bottom of the transcribed text.
+
+<p align="center">
+  <img src="./example.png" alt="example"/>
+</p>
+
+
+
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-audio/audio.py b/docs/bootcamp/streamlit/gptcache-streamlit-audio/audio.py
@@ -0,0 +1,60 @@
+import streamlit as st
+import os
+import uuid
+
+from gptcache import cache
+from gptcache.manager import get_data_manager, CacheBase, VectorBase, ObjectBase
+from gptcache.adapter import openai
+from gptcache.processor.pre import get_file_name
+from gptcache.embedding import Data2VecAudio
+from gptcache.similarity_evaluation.distance import SearchDistanceEvaluation
+
+
+@st.cache_resource
+def initialize_configuration():
+    data2vec = Data2VecAudio()
+    data_manager = get_data_manager(CacheBase('sqlite', sql_url='sqlite:///./local/gptcache20.db'),
+                                    VectorBase('faiss', dimension=data2vec.dimension, index_path='./local/faiss20.index'),
+                                    ObjectBase('local', path='./local'))
+    cache.init(
+        pre_embedding_func=get_file_name,
+        embedding_func=data2vec.to_embeddings,
+        data_manager=data_manager,
+        similarity_evaluation=SearchDistanceEvaluation(),
+        )
+    return data_manager
+
+data_manager = initialize_configuration()
+
+def api_call(audio_bytes, open_ai_key):
+    os.environ['OPENAI_API_KEY'] =  open_ai_key
+    os.environ['CURL_CA_BUNDLE'] = ''
+    os.environ['KMP_DUPLICATE_LIB_OK']='TRUE'
+    cache.set_openai_key()
+    transcript = openai.Audio.transcribe('whisper-1', audio_bytes, api_key=open_ai_key)
+    is_cached = transcript.get('gptcache', False)
+    return transcript['text'], is_cached
+
+def main():
+
+    st.title('GPTCache for Audio Demo')
+    open_ai_key = st.text_input('OpenAI key')
+    audio_file = st.file_uploader('Choose an audio file (.mp3, .wav, or .ogg)', type=['mp3', 'wav', 'ogg'])
+
+    if st.button('generate', key='button'):
+        file_extension = os.path.splitext(audio_file.name)[1]  # Get the extension of the uploaded file
+        random_filename = str(uuid.uuid4()) + file_extension
+        with open(random_filename, 'wb') as f:
+            f.write(audio_file.getbuffer())
+        audio_file_handler = open(random_filename, 'rb')
+        text, is_cached = api_call(audio_file_handler, open_ai_key)
+        st.write(text)
+        os.remove(random_filename)
+
+        if is_cached:
+            st.markdown('<div style="display: flex; align-items: center; justify-content: center; \
+                    background-color: green; padding: 10px; color: white; font-weight: bold; border-radius: \
+                    5px; margin: 10px auto; max-width: 100px;">cache</div>', unsafe_allow_html=True)
+
+if __name__ == '__main__':
+    main()
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-audio/example.png b/docs/bootcamp/streamlit/gptcache-streamlit-audio/example.png
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-audio/local/.cache b/docs/bootcamp/streamlit/gptcache-streamlit-audio/local/.cache
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-audio/requirements.txt b/docs/bootcamp/streamlit/gptcache-streamlit-audio/requirements.txt
@@ -0,0 +1,8 @@
+gptcache
+pillow
+streamlit
+torch
+faiss-cpu
+torchaudio
+transformers
+sqlalchemy
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-image/README.md b/docs/bootcamp/streamlit/gptcache-streamlit-image/README.md
@@ -0,0 +1,22 @@
+# GPTCache Demo for OpenAI Image Generation
+This project demonstrates how [GPTcache](https://github.com/zilliztech/GPTCache) can be used to save costs when using OpenAI’s DALL-E API. It provides a simple Streamlit app that allows users to input a prompt and see the corresponding DALL-E output image. The app uses a cache to store previously generated images and reuses them for the same prompt, thus avoiding making duplicate API calls. There is an online [demo](https://gptcache-openai-image.streamlit.app/) hosted for preview.
+
+## Requirements
+* Python 3.6 or later 
+* Dependencies listed in requirements.txt 
+* OpenAI API key
+## Usage
+1. Clone the repository to your local machine
+Install the required packages: pip install -r requirements.txt
+2. Run the app: streamlit run imagen.py
+3. Open the app in your browser at http://localhost:8501
+4. Enter your OpenAI key and prompt then click “generate” to
+wait for the DALL-E output image to appear.
+If a cache hit occurred, you should see a message like “cache” at the bottom of the image.
+
+<p align="center">
+  <img src="./example.png" alt="example"/>
+</p>
+
+
+
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-image/example.png b/docs/bootcamp/streamlit/gptcache-streamlit-image/example.png
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-image/imagen.py b/docs/bootcamp/streamlit/gptcache-streamlit-image/imagen.py
@@ -0,0 +1,81 @@
+import streamlit as st
+from PIL import Image
+import os
+import io
+import base64
+from io import BytesIO
+import requests
+
+from gptcache import cache
+from gptcache.manager import get_data_manager, CacheBase, VectorBase, ObjectBase
+from gptcache.adapter import openai
+from gptcache.processor.pre import get_prompt
+from gptcache.embedding import Onnx
+from gptcache.similarity_evaluation import ExactMatchEvaluation
+
+st.title('GPTCache for Image Demo')
+
+@st.cache_resource
+def initialize_configuration():
+    onnx = Onnx()
+    data_manager = get_data_manager(CacheBase('sqlite', sql_url='sqlite:///./local/gptcache10.db'),
+                                    VectorBase('faiss', dimension=onnx.dimension, index_path='./local/faiss10.index'),
+                                    ObjectBase('local', path='./local'))
+    cache.init(
+        pre_embedding_func=get_prompt,
+        embedding_func=onnx.to_embeddings,
+        data_manager=data_manager,
+        similarity_evaluation=ExactMatchEvaluation(),
+        )
+    return data_manager
+
+data_manager = initialize_configuration()
+
+def api_call(text_input, open_ai_key):
+    os.environ['CURL_CA_BUNDLE'] = ''
+    response = openai.Image.create(
+      prompt=text_input,
+      n=1,
+      size='256x256',
+      api_key=open_ai_key
+    )
+    image_url = response['data'][0]['url']
+
+    is_cached = response.get('gptcache', False)
+    if is_cached is False:
+        response = requests.get(image_url)
+        img = Image.open(BytesIO(response.content))
+    else:
+        img = Image.open(image_url)
+    return img, is_cached
+
+def main():
+    open_ai_key = st.text_input('OpenAI key:')
+    text_input = st.text_input('prompt:')
+
+    if st.button('generate', key='button'):
+        try:
+            image, is_cached = api_call(text_input, open_ai_key)
+            width, height = image.size
+            desired_width = 500
+            desired_height = int(height * desired_width / width)
+            resized_image = image.resize((desired_width, desired_height))
+            img_bytes = io.BytesIO()
+            resized_image.save(img_bytes, format='PNG')
+            img_str = base64.b64encode(img_bytes.getvalue()).decode()
+
+            st.markdown(
+                f'<div style="display: flex; justify-content: center;"><img src="data:image/png;base64,{img_str}" \
+                        alt="Uploaded Image" width="{desired_width}"></div>',
+                unsafe_allow_html=True
+            )
+
+            if is_cached:
+                st.markdown('<div style="display: flex; align-items: center; justify-content: center; background-color: \
+                        green; padding: 10px; color: white; font-weight: bold; border-radius: 5px; margin: 10px auto; \
+                        max-width: 100px;">cache</div>', unsafe_allow_html=True)
+        except Exception as e:
+            st.error('invalid OpenAI API key or inappropriate prompt rejected by OpenAI.')
+
+if __name__ == '__main__':
+    main()
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-image/local/.cache b/docs/bootcamp/streamlit/gptcache-streamlit-image/local/.cache
diff --git a/docs/bootcamp/streamlit/gptcache-streamlit-image/requirements.txt b/docs/bootcamp/streamlit/gptcache-streamlit-image/requirements.txt
@@ -0,0 +1,7 @@
+gptcache
+pillow
+streamlit
+onnxruntime
+faiss-cpu
+transformers
+sqlalchemy