Skip to content

Commit

Permalink
update readme
Browse files Browse the repository at this point in the history
  • Loading branch information
@diptanu
diptanu committed Aug 16, 2024
1 parent 5acb82a commit 30229e7
Show file tree
Hide file tree
Showing 3 changed files with 54 additions and 123 deletions.
101 changes: 28 additions & 73 deletions examples/pdf/indexing_and_rag/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,6 @@ Before we begin, ensure you have the following:
```
- `pip` (Python package manager)
- An OpenAI API key
- Basic familiarity with Python and command-line interfaces

## Setup

Expand Down Expand Up @@ -96,10 +95,8 @@ Create a file [`upload_and_retrieve.py`](upload_and_retrieve.py):
```python
from indexify import IndexifyClient
from indexify.data_loaders import LocalDirectoryLoader
import os, requests
from indexify.data_loaders import UrlLoader
from openai import OpenAI
import tempfile
client = IndexifyClient()
Expand Down Expand Up @@ -146,43 +143,34 @@ def answer_question(question):
)
return chat_completion.choices[0].message.content
# Example usage
if __name__ == "__main__":
# Uncomment the lines if you want to upload more than 1 pdf
pdf_urls = [
"https://arxiv.org/pdf/2304.08485.pdf",
"https://arxiv.org/pdf/0910.2029.pdf",
"https://arxiv.org/pdf/2402.01968.pdf",
"https://arxiv.org/pdf/2401.13138.pdf",
"https://arxiv.org/pdf/2402.03578.pdf",
"https://arxiv.org/pdf/2309.07864.pdf",
"https://arxiv.org/pdf/2401.03568.pdf",
"https://arxiv.org/pdf/2312.10256.pdf",
"https://arxiv.org/pdf/2312.01058.pdf",
"https://arxiv.org/pdf/2402.01680.pdf",
"https://arxiv.org/pdf/2403.07017.pdf"
"http://arxiv.org/pdf/2304.08485"
#"https://arxiv.org/pdf/2304.08485.pdf",
# "https://arxiv.org/pdf/0910.2029.pdf",
# "https://arxiv.org/pdf/2402.01968.pdf",
# "https://arxiv.org/pdf/2401.13138.pdf",
# "https://arxiv.org/pdf/2402.03578.pdf",
# "https://arxiv.org/pdf/2309.07864.pdf",
# "https://arxiv.org/pdf/2401.03568.pdf",
# "https://arxiv.org/pdf/2312.10256.pdf",
# "https://arxiv.org/pdf/2312.01058.pdf",
# "https://arxiv.org/pdf/2402.01680.pdf",
# "https://arxiv.org/pdf/2403.07017.pdf"
]
os.makedirs("pdfs", exist_ok=True)
data_loader = UrlLoader(pdf_urls)
content_ids = client.ingest_from_loader(data_loader, "rag_pipeline")
for url in pdf_urls:
filename = url.split("/")[-1]
response = requests.get(url)
if response.status_code == 200:
with open(os.path.join("pdfs", filename), "wb") as file:
file.write(response.content)
print(f"Downloaded {filename}")
else:
print(f"Failed to download {filename}")
director_loader = LocalDirectoryLoader("pdfs", file_extensions=["pdf"])
content_ids = client.ingest_from_loader(director_loader, "rag_pipeline")
print(f"Processed {len(content_ids)} documents")
print(f"Uploaded {len(content_ids)} documents")
client.wait_for_extraction(content_ids)
question = "What is the performance of LLaVa across across multiple image domains / subjects?"
answer = answer_question(question)
print(f"Question: {question}")
print(f"Answer: {answer}")
```
Setup the OPENAI API KEY in your terminal before running the script.
Expand Down Expand Up @@ -239,18 +227,7 @@ extraction_policies:
Create a new Python file called [`setup_graph_mm.py`](setup_graph_mm.py) and add the following code:
```python
import os
from indexify import IndexifyClient, ExtractionGraph
client = IndexifyClient()
script_dir = os.path.dirname(os.path.abspath(__file__))
yaml_file_path = os.path.join(script_dir, "graph_mm.yaml")
extraction_graph = ExtractionGraph.from_yaml_file(yaml_file_path)
client.create_extraction_graph(extraction_graph)
```
It's pretty similar to the one we created above, we will simply replace the graph file name to `graph_mm.yaml`

Run this script to set up the multi-modal pipeline:
```bash
Expand All @@ -263,11 +240,10 @@ Create a file [`upload_and_retrieve_mm.py`](upload_and_retrieve_mm.py), which is

```python
from indexify import IndexifyClient
from indexify.data_loaders import LocalDirectoryLoader
from indexify.data_loaders import UrlLoader
import requests
from openai import OpenAI
import base64
import tempfile
client = IndexifyClient()
client_openai = OpenAI()
Expand Down Expand Up @@ -327,38 +303,26 @@ def answer_question(question):
)
return chat_completion.choices[0].message.content
# Example usage
if __name__ == "__main__":
# Uncomment the lines if you want to upload more than 1 pdf
pdf_urls = [
"https://proceedings.neurips.cc/paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf",
"https://arxiv.org/pdf/1810.04805.pdf",
"https://arxiv.org/pdf/2304.08485"
# "https://arxiv.org/pdf/1810.04805.pdf",
# "https://arxiv.org/pdf/2304.08485"
]
os.makedirs("pdfs", exist_ok=True)
for url in pdf_urls:
filename = url.split("/")[-1]
response = requests.get(url)
if response.status_code == 200:
with open(os.path.join("pdfs", filename), "wb") as file:
file.write(response.content)
print(f"Downloaded {filename}")
else:
print(f"Failed to download {filename}")
data_loader = UrlLoader(pdf_urls)
content_ids = client.ingest_from_loader(data_loader, "rag_pipeline_mm")
director_loader = LocalDirectoryLoader("pdfs", file_extensions=["pdf"])
content_ids = client.ingest_from_loader(director_loader, "rag_pipeline_mm")
print(f"Processed {len(content_ids)} documents")
print(f"Uploaded {len(content_ids)} documents")
client.wait_for_extraction(content_ids)
# Ask questions
questions = [
"What does the architecture diagram show?",
"Explain the attention mechanism in transformers.",
"What are the key contributions of BERT?",
]
for question in questions:
answer = answer_question(question)
print(f"\nQuestion: {question}")
Expand Down Expand Up @@ -401,12 +365,3 @@ You can customize both RAG systems in several ways:
4. Experiment with different OpenAI models or adjust the prompt structure in the `create_prompt` function.
5. For the multi-modal RAG, you can adjust the number of images retrieved or how they are incorporated into the prompt.
## Conclusion
These RAG systems demonstrate the power of combining Indexify with large language models:
1. **Scalability**: Indexify can process and index large numbers of PDFs efficiently, including both text and images.
2. **Flexibility**: You can easily swap out components or adjust parameters to suit your specific needs.
3. **Integration**: The systems seamlessly integrate PDF processing, embedding generation, and text generation.
4. **Multi-Modal Capabilities**: The second system shows how to incorporate both text and image data for more comprehensive question answering.
47 changes: 18 additions & 29 deletions examples/pdf/indexing_and_rag/upload_and_retrieve.py
View file
Original file line number Diff line number Diff line change
@@ -1,6 +1,5 @@
from indexify import IndexifyClient
from indexify.data_loaders import LocalDirectoryLoader
import os, requests
from indexify.data_loaders import UrlLoader
from openai import OpenAI

client = IndexifyClient()
Expand Down Expand Up @@ -48,40 +47,30 @@ def answer_question(question):
)
return chat_completion.choices[0].message.content

# Example usage
if __name__ == "__main__":
# Uncomment the lines if you want to upload more than 1 pdf
pdf_urls = [
"http://arxiv.org/pdf/2304.08485"
"https://arxiv.org/pdf/2304.08485.pdf",
"https://arxiv.org/pdf/0910.2029.pdf",
"https://arxiv.org/pdf/2402.01968.pdf",
"https://arxiv.org/pdf/2401.13138.pdf",
"https://arxiv.org/pdf/2402.03578.pdf",
"https://arxiv.org/pdf/2309.07864.pdf",
"https://arxiv.org/pdf/2401.03568.pdf",
"https://arxiv.org/pdf/2312.10256.pdf",
"https://arxiv.org/pdf/2312.01058.pdf",
"https://arxiv.org/pdf/2402.01680.pdf",
"https://arxiv.org/pdf/2403.07017.pdf"
# "https://arxiv.org/pdf/0910.2029.pdf",
# "https://arxiv.org/pdf/2402.01968.pdf",
# "https://arxiv.org/pdf/2401.13138.pdf",
# "https://arxiv.org/pdf/2402.03578.pdf",
# "https://arxiv.org/pdf/2309.07864.pdf",
# "https://arxiv.org/pdf/2401.03568.pdf",
# "https://arxiv.org/pdf/2312.10256.pdf",
# "https://arxiv.org/pdf/2312.01058.pdf",
# "https://arxiv.org/pdf/2402.01680.pdf",
# "https://arxiv.org/pdf/2403.07017.pdf"
]

os.makedirs("pdfs", exist_ok=True)
data_loader = UrlLoader(pdf_urls)
content_ids = client.ingest_from_loader(data_loader, "rag_pipeline")

for url in pdf_urls:
filename = url.split("/")[-1]
response = requests.get(url)
if response.status_code == 200:
with open(os.path.join("pdfs", filename), "wb") as file:
file.write(response.content)
print(f"Downloaded {filename}")
else:
print(f"Failed to download {filename}")

director_loader = LocalDirectoryLoader("pdfs", file_extensions=["pdf"])
content_ids = client.ingest_from_loader(director_loader, "rag_pipeline")

print(f"Processed {len(content_ids)} documents")
print(f"Uploaded {len(content_ids)} documents")
client.wait_for_extraction(content_ids)

question = "What is the performance of LLaVa across across multiple image domains / subjects?"
answer = answer_question(question)
print(f"Question: {question}")
print(f"Answer: {answer}")
print(f"Answer: {answer}")
29 changes: 8 additions & 21 deletions examples/pdf/indexing_and_rag/upload_and_retrieve_mm.py
View file
Original file line number Diff line number Diff line change
@@ -1,9 +1,8 @@
from indexify import IndexifyClient
from indexify.data_loaders import LocalDirectoryLoader
from indexify.data_loaders import UrlLoader
import requests
from openai import OpenAI
import base64
import os

client = IndexifyClient()
client_openai = OpenAI()
Expand Down Expand Up @@ -63,38 +62,26 @@ def answer_question(question):
)
return chat_completion.choices[0].message.content

# Example usage
if __name__ == "__main__":
# Uncomment the lines if you want to upload more than 1 pdf
pdf_urls = [
"https://proceedings.neurips.cc/paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf",
"https://arxiv.org/pdf/1810.04805.pdf",
"https://arxiv.org/pdf/2304.08485"
# "https://arxiv.org/pdf/1810.04805.pdf",
# "https://arxiv.org/pdf/2304.08485"
]

os.makedirs("pdfs", exist_ok=True)
data_loader = UrlLoader(pdf_urls)
content_ids = client.ingest_from_loader(data_loader, "rag_pipeline_mm")

for url in pdf_urls:
filename = url.split("/")[-1]
response = requests.get(url)
if response.status_code == 200:
with open(os.path.join("pdfs", filename), "wb") as file:
file.write(response.content)
print(f"Downloaded {filename}")
else:
print(f"Failed to download {filename}")

director_loader = LocalDirectoryLoader("pdfs", file_extensions=["pdf"])
content_ids = client.ingest_from_loader(director_loader, "rag_pipeline_mm")

print(f"Processed {len(content_ids)} documents")
print(f"Uploaded {len(content_ids)} documents")
client.wait_for_extraction(content_ids)

# Ask questions
questions = [
"What does the architecture diagram show?",
"Explain the attention mechanism in transformers.",
"What are the key contributions of BERT?",
]

for question in questions:
answer = answer_question(question)
print(f"\nQuestion: {question}")
Expand Down

0 comments on commit 30229e7

Please sign in to comment.