Spaces:
Running
Running
talexm
commited on
Commit
•
e4a2031
1
Parent(s):
c3c1187
update
Browse files- rag_sec/README.md +44 -17
rag_sec/README.md
CHANGED
@@ -1,24 +1,51 @@
|
|
|
|
1 |
|
2 |
-
|
3 |
|
4 |
-
|
5 |
-
This project demonstrates a RAG system enhanced with Chagu features for:
|
6 |
-
- Data Poisoning Detection
|
7 |
-
- Model Drift Handling
|
8 |
-
- Query Injection Attack Prevention
|
9 |
-
- Adversarial Embedding Detection
|
10 |
|
11 |
-
##
|
12 |
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
```
|
17 |
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
|
23 |
## Requirements
|
24 |
-
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Document Search and Response Generation System
|
2 |
|
3 |
+
This project implements a **Document Search and Response Generation System** combining semantic search, malicious query detection, and generative response capabilities. It is designed for efficient and context-aware information retrieval and response generation.
|
4 |
|
5 |
+
---
|
|
|
|
|
|
|
|
|
|
|
6 |
|
7 |
+
## Features
|
8 |
|
9 |
+
1. **Semantic Search**:
|
10 |
+
- Uses SentenceTransformer embeddings for document similarity.
|
11 |
+
- Retrieves top-k relevant documents for a given query.
|
|
|
12 |
|
13 |
+
2. **Malicious Query Detection**:
|
14 |
+
- Identifies and blocks malicious or harmful queries using sentiment analysis.
|
15 |
+
|
16 |
+
3. **Query Transformation**:
|
17 |
+
- Rephrases or enhances ambiguous queries for better processing.
|
18 |
+
|
19 |
+
4. **Generative Response**:
|
20 |
+
- Generates a context-aware response using Hugging Face models like `distilgpt2`.
|
21 |
+
|
22 |
+
5. **Expandable Architecture**:
|
23 |
+
- Modular components for easy enhancement and integration.
|
24 |
+
- Compatible with lightweight and resource-efficient models.
|
25 |
+
|
26 |
+
---
|
27 |
+
|
28 |
+
## Architecture
|
29 |
+
|
30 |
+
1. **Bad Query Detector**:
|
31 |
+
- Detects malicious or inappropriate queries using sentiment analysis (`distilbert-base-uncased-finetuned-sst-2-english`).
|
32 |
+
|
33 |
+
2. **Query Transformer**:
|
34 |
+
- Rephrases or improves queries for better retrieval results.
|
35 |
+
|
36 |
+
3. **Document Retriever**:
|
37 |
+
- Encodes documents into dense vectors using `all-MiniLM-L6-v2` embeddings.
|
38 |
+
- Finds similar documents using cosine similarity.
|
39 |
+
|
40 |
+
4. **Semantic Response Generator**:
|
41 |
+
- Generates context-aware responses using models like `distilgpt2` or `EleutherAI/gpt-neo-1.3B`.
|
42 |
+
|
43 |
+
---
|
44 |
|
45 |
## Requirements
|
46 |
+
|
47 |
+
### Python Libraries
|
48 |
+
Install the necessary libraries using `pip`:
|
49 |
+
```bash
|
50 |
+
pip install transformers sentence-transformers scikit-learn numpy flask
|
51 |
+
```
|