adpmdrones commited on
Commit
7c5a5dc
·
1 Parent(s): 3a4af69

update readme

Browse files
Files changed (1) hide show
  1. README.md +9 -174
README.md CHANGED
@@ -1,174 +1,9 @@
1
-
2
- # pdfGPT
3
- ### Problem Description :
4
- 1. When you pass a large text to Open AI, it suffers from a 4K token limit. It cannot take an entire pdf file as an input
5
- 2. Open AI sometimes becomes overtly chatty and returns irrelevant response not directly related to your query. This is because Open AI uses poor embeddings.
6
- 3. ChatGPT cannot directly talk to external data. Some solutions use Langchain but it is token hungry if not implemented correctly.
7
- 4. There are a number of solutions like https://www.chatpdf.com, https://www.bespacific.com/chat-with-any-pdf, https://www.filechat.io they have poor content quality and are prone to hallucination problem. One good way to avoid hallucinations and improve truthfulness is to use improved embeddings. To solve this problem, I propose to improve embeddings with Universal Sentence Encoder family of algorithms (Read more here: https://tfhub.dev/google/collections/universal-sentence-encoder/1).
8
-
9
- ### Solution: What is PDF GPT ?
10
- 1. PDF GPT allows you to chat with an uploaded PDF file using GPT functionalities.
11
- 2. The application intelligently breaks the document into smaller chunks and employs a powerful Deep Averaging Network Encoder to generate embeddings.
12
- 3. A semantic search is first performed on your pdf content and the most relevant embeddings are passed to the Open AI.
13
- 4. A custom logic generates precise responses. The returned response can even cite the page number in square brackets([]) where the information is located, adding credibility to the responses and helping to locate pertinent information quickly. The Responses are much better than the naive responses by Open AI.
14
- 5. Andrej Karpathy mentioned in this post that KNN algorithm is most appropriate for similar problems: https://twitter.com/karpathy/status/1647025230546886658
15
- 6. Enables APIs on Production using **[langchain-serve](https://github.com/jina-ai/langchain-serve)**.
16
-
17
- ### Demo
18
- 1. **Demo URL**: https://bit.ly/41ZXBJM
19
- 2. **Original Source code** (for demo hosted in Hugging Face) : https://huggingface.co/spaces/bhaskartripathi/pdfChatter/blob/main/app.py
20
-
21
- **NOTE**: Please star this project if you like it!
22
-
23
- ### Docker
24
- Run `docker-compose -f docker-compose.yaml up` to use it with Docker compose.
25
-
26
-
27
- ## Use `pdfGPT` on Production using [langchain-serve](https://github.com/jina-ai/langchain-serve)
28
-
29
- #### Local playground
30
- 1. Run `lc-serve deploy local api` on one terminal to expose the app as API using langchain-serve.
31
- 2. Run `python app.py` on another terminal for a local gradio playground.
32
- 3. Open `http://localhost:7860` on your browser and interact with the app.
33
-
34
-
35
- #### Cloud deployment
36
-
37
- Make `pdfGPT` production ready by deploying it on [Jina Cloud](https://cloud.jina.ai/).
38
-
39
- `lc-serve deploy jcloud api`
40
-
41
- <details>
42
- <summary>Show command output</summary>
43
-
44
- ```text
45
- ╭──────────────┬──────────────────────────────────────────────────────────────────────────────────────╮
46
- │ App ID │ langchain-3ff4ab2c9d │
47
- ├──────────────┼──────────────────────────────────────────────────────────────────────────────────────┤
48
- │ Phase │ Serving │
49
- ├──────────────┼──────────────────────────────────────────────────────────────────────────────────────┤
50
- │ Endpoint │ https://langchain-3ff4ab2c9d.wolf.jina.ai │
51
- ├──────────────┼──────────────────────────────────────────────────────────────────────────────────────┤
52
- │ App logs │ dashboards.wolf.jina.ai │
53
- ├──────────────┼──────────────────────────────────────────────────────────────────────────────────────┤
54
- │ Swagger UI │ https://langchain-3ff4ab2c9d.wolf.jina.ai/docs │
55
- ├──────────────┼──────────────────────────────��───────────────────────────────────────────────────────┤
56
- │ OpenAPI JSON │ https://langchain-3ff4ab2c9d.wolf.jina.ai/openapi.json │
57
- ╰──────────────┴──────────────────────────────────────────────────────────────────────────────────────╯
58
- ```
59
-
60
- </details>
61
-
62
- #### Interact using cURL
63
-
64
- (Change the URL to your own endpoint)
65
-
66
- **PDF url**
67
- ```bash
68
- curl -X 'POST' \
69
- 'https://langchain-3ff4ab2c9d.wolf.jina.ai/ask_url' \
70
- -H 'accept: application/json' \
71
- -H 'Content-Type: application/json' \
72
- -d '{
73
- "url": "https://uiic.co.in/sites/default/files/uploads/downloadcenter/Arogya%20Sanjeevani%20Policy%20CIS_2.pdf",
74
- "question": "What'\''s the cap on room rent?",
75
- "envs": {
76
- "OPENAI_API_KEY": "'"${OPENAI_API_KEY}"'"
77
- }
78
- }'
79
-
80
- {"result":" Room rent is subject to a maximum of INR 5,000 per day as specified in the Arogya Sanjeevani Policy [Page no. 1].","error":"","stdout":""}
81
- ```
82
-
83
- **PDF file**
84
- ```bash
85
- QPARAMS=$(echo -n 'input_data='$(echo -n '{"question": "What'\''s the cap on room rent?", "envs": {"OPENAI_API_KEY": "'"${OPENAI_API_KEY}"'"}}' | jq -s -R -r @uri))
86
- curl -X 'POST' \
87
- 'https://langchain-3ff4ab2c9d.wolf.jina.ai/ask_file?'"${QPARAMS}" \
88
- -H 'accept: application/json' \
89
- -H 'Content-Type: multipart/form-data' \
90
- -F 'file=@Arogya_Sanjeevani_Policy_CIS_2.pdf;type=application/pdf'
91
-
92
- {"result":" Room rent is subject to a maximum of INR 5,000 per day as specified in the Arogya Sanjeevani Policy [Page no. 1].","error":"","stdout":""}
93
- ```
94
-
95
- ## Running on localhost
96
- ### Credits : [Adithya S](https://github.com/200901002)
97
- 1. Pull the image by entering the following command in your terminal or command prompt:
98
- ```bash
99
- docker pull registry.hf.space/bhaskartripathi-pdfchatter:latest
100
- ```
101
- 2. Download the Universal Sentence Encoder locally to your project's root folder. This is important because otherwise, 915 MB will be downloaded at runtime everytime you run it.
102
- 3. Download the encoder using this [link](https://tfhub.dev/google/universal-sentence-encoder/4?tf-hub-format=compressed).
103
- 4. Extract the downloaded file and place it in your project's root folder as shown below:
104
- ```text
105
- Root folder of your project
106
- └───Universal Sentence Encoder
107
- | ├───assets
108
- | └───variables
109
- | └───saved_model.pb
110
- |
111
- └───app.py
112
- ```
113
- 5. If you have downloaded it locally, replace the code on line 68 in the API file:
114
- ```python
115
- self.use = hub.load('https://tfhub.dev/google/universal-sentence-encoder/4')
116
- ```
117
- with:
118
- ```python
119
- self.use = hub.load('./Universal Sentence Encoder/')
120
- ```
121
- 6. Now, To run PDF-GPT, enter the following command:
122
-
123
- ```bash
124
- docker run -it -p 7860:7860 --platform=linux/amd64 registry.hf.space/bhaskartripathi-pdfchatter:latest python app.py
125
- ```
126
-
127
- ## UML
128
- ```mermaid
129
- sequenceDiagram
130
- participant User
131
- participant System
132
-
133
- User->>System: Enter API Key
134
- User->>System: Upload PDF/PDF URL
135
- User->>System: Ask Question
136
- User->>System: Submit Call to Action
137
-
138
- System->>System: Blank field Validations
139
- System->>System: Convert PDF to Text
140
- System->>System: Decompose Text to Chunks (150 word length)
141
- System->>System: Check if embeddings file exists
142
- System->>System: If file exists, load embeddings and set the fitted attribute to True
143
- System->>System: If file doesn't exist, generate embeddings, fit the recommender, save embeddings to file and set fitted attribute to True
144
- System->>System: Perform Semantic Search and return Top 5 Chunks with KNN
145
- System->>System: Load Open AI prompt
146
- System->>System: Embed Top 5 Chunks in Open AI Prompt
147
- System->>System: Generate Answer with Davinci
148
-
149
- System-->>User: Return Answer
150
- ```
151
-
152
- ### Flowchart
153
- ```mermaid
154
- flowchart TB
155
- A[Input] --> B[URL]
156
- A -- Upload File manually --> C[Parse PDF]
157
- B --> D[Parse PDF] -- Preprocess --> E[Dynamic Text Chunks]
158
- C -- Preprocess --> E[Dynamic Text Chunks with citation history]
159
- E --Fit-->F[Generate text embedding with Deep Averaging Network Encoder on each chunk]
160
- F -- Query --> G[Get Top Results]
161
- G -- K-Nearest Neighbour --> K[Get Nearest Neighbour - matching citation references]
162
- K -- Generate Prompt --> H[Generate Answer]
163
- H -- Output --> I[Output]
164
- ```
165
- ## Star History
166
-
167
- [![Star History Chart](https://api.star-history.com/svg?repos=bhaskatripathi/pdfGPT&type=Date)](https://star-history.com/#bhaskatripathi/pdfGPT&Date)
168
- I am looking for more contributors from the open source community who can take up backlog items voluntarily and maintain the application jointly with me.
169
-
170
- ## Also Try:
171
- This app creates schematic architecture diagrams, UML, flowcharts, Gantt charts and many more. You simple need to mention the usecase in natural language and it will create the desired diagram.
172
- https://github.com/bhaskatripathi/Text2Diagram
173
-
174
-
 
1
+ title: pdfADPM
2
+ emoji: 🏢
3
+ colorFrom: indigo
4
+ colorTo: green
5
+ sdk: gradio
6
+ sdk_version: 3.20.1
7
+ app_file: app.py
8
+ pinned: false
9
+ license: afl-3.0