Building AI and RAG Apps With MongoDB, Anyscale and PyMongo

1
{
2
  "fields": [
3
    {
4
      "type": "vector",
5
      "path": "embedding",
6
      "numDimensions": 1024,
7
      "similarity": "cosine"
8
    }
9
  ]
10
}

1
https://XXXXXXXXXXXX.anyscaleuserdata.com

1
https://YYYYYYYYYYYYY.anyscaleuserdata.com

1
pip install pymongo openai fastapi 

1
from typing import Any
2
from fastapi import FastAPI, HTTPException
3
from pydantic import BaseModel
4
from pymongo import MongoClient
5
import os
6
import openai
7
from dotenv import load_dotenv
8
from ray import serve
9
import logging
10
11
# Initialize FastAPI
12
fastapi = FastAPI()
13
14
# Initialize logger
15
logger = logging.getLogger("ray.serve")
16
17
# Data model for the request
18
class QueryModel(BaseModel):
19
   query: str
20
21
@serve.deployment
22
@serve.ingress(fastapi)
23
class MongoDBRag:
24
25
   def __init__(self):
26
       # MongoDB connection
27
       client = MongoClient(os.getenv("MONGODB_ATLAS_URI"))
28
       db = client["anyscale_db"]
29
       self.collection = db["stories"]
30
31
       # AI Client initialization
32
       # LLM endpoint for text generation
33
       self.openai_llm = openai.OpenAI(
34
           base_url="https://YYYYYYYYYYYYY.anyscaleuserdata.com/v1",
35
           api_key=os.getenv('LLM_API_KEY')
36
       )
37
38
       # Embedding endpoint for vector embeddings
39
       self.openai_embedding = openai.OpenAI(
40
           base_url="https://XXXXXXXXXXXX.anyscaleuserdata.com/v1",
41
           api_key=os.getenv('EMBEDDING_API_KEY')
42
       )
43
44
   @fastapi.post("/rag_endpoint/")
45
   async def rag_endpoint(self, query: QueryModel) -> Any:
46
       try:
47
           # Generate embedding for the input query
48
           embedding = self.openai_embedding.embeddings.create(
49
               model="thenlper/gte-large",
50
               input=query.query,
51
           )
52
53
           # Perform vector search in MongoDB to find relevant context
54
           context = list(self.collection.aggregate([
55
               {"$vectorSearch": {
56
                   "queryVector": embedding.data[0].embedding,
57
                   "index": "vector_index",
58
                   "numCandidates": 3,
59
                   "path": "embedding",
60
                   "limit": 1
61
               }},
62
               {
63
                   "$project": {
64
                       "_id": 0,
65
                       "embedding": 0
66
                   }
67
               }
68
           ]))
69
70
           # Construct prompt with context and query
71
           prompt = f""" Based on the following context please answer and elaborate the user query.
72
           context: {context}
73
74
           query: {query.query}
75
           Answer: """
76
77
           # Generate response using LLM
78
           chat_completion = self.openai_llm.chat.completions.create(
79
               model="meta-llama/Llama-2-70b-chat-hf",
80
               messages=[
81
                   {"role": "system", "content": "You are a helpful QA assistant"},
82
                   {"role": "user", "content": prompt}
83
               ],
84
               temperature=0.7
85
           )
86
87
           answer = chat_completion.choices[0].message.content
88
89
           return answer
90
91
       except Exception as e:
92
           raise HTTPException(status_code=500, detail=str(e))
93
94
# Bind the deployment
95
my_app = MongoDBRag.bind()

1
from pymongo import MongoClient
2
import os
3
import openai
4
from dotenv import load_dotenv
5
6
client = MongoClient(os.getenv("MONGODB_ATLAS_URI"))
7
db = client["anyscale_db"]
8
collection = db["stories"]
9
10
# Embedding endpoint for vector embeddings
11
openai_embedding = openai.OpenAI(
12
    base_url="https://XXXXXXXXXXXX.anyscaleuserdata.com/v1",
13
    api_key=os.getenv('EMBEDDING_API_KEY')
14
)
15
16
def generate_data() -> str:
17
    # Sample data to be inserted into MongoDB
18
    data = [
19
        "Karl lives in NYC and works as a chef",
20
        "Rio lives in Brasil and works as a barman",
21
        "Jessica lives in Paris and works as a programmer."
22
    ]
23
24
    for info in data:
25
        # Generate embedding for each piece of information
26
        embedding = openai_embedding.embeddings.create(
27
            model="thenlper/gte-large",
28
            input=info,
29
        )
30
31
        # Update or insert the data with its embedding into MongoDB
32
        collection.update_one(
33
            {'story': info},
34
            {'$set': {'embedding': embedding.data[0].embedding}},
35
            upsert=True
36
        )
37
38
    return "Data Ingested Successfully"
39
40
generate_data()

1
 data = [
2
               "Karl lives in NYC and works as a chef",
3
               "Rio lives in Brasil and works as a barman",
4
               "Jessica lives in Paris and works as a programmer."
5
           ]

1
python generate.py

1
serve run main:my_app --non-blocking

1
payload = {"query": "Main character list"}
2
3
4
# Perform the POST request
5
response = requests.post("http://localhost:8000/rag_endpoint", json=payload)
6
7
# Print the response in JSON format
8
print(response.json())

1
Based on the provided context, the main character list is:
2
3
4
1. Jessica
5
2. Rio
6
3. Karl

1
anyscale services deploy  main:my_app --name=my-rag-service

1
(anyscale +2.0s) Starting new service 'my-rag-service'.
2
(anyscale +4.5s) Using workspace runtime dependencies env vars: {'MONGODB_ATLAS_URI': '<YOUR_URI>', 'LLM_API_KEY': '<YOURKEY>',  'EMBEDDING_API_KEY': '<YOURKEY>'}.
3
(anyscale +4.5s) Uploading local dir '.' to cloud storage.
4
(anyscale +5.5s) Including workspace-managed pip dependencies.
5
(anyscale +6.1s) Service 'my-rag-service' deployed.
6
(anyscale +6.1s) View the service in the UI: 'https://console.anyscale.com/services/service2_wx8ar3cbuf84g9u87g3yvx2cq8'
7
(anyscale +6.1s) Query the service once it's running using the following curl command:
8
(anyscale +6.1s) curl -H 'Authorization: Bearer xxxxxxx' https://my-rag-service-yyyy.xxxxxxxxxxxx.s.anyscaleuserdata.com/

1
curl -X POST \
2
     -H "Authorization: Bearer XXXX" \
3
     -H "Content-Type: application/json" \
4
    https://my-rag-service-yyyy.cld-yga7qm8f6yamqev7.s.anyscaleuserdata.com/rag_endpoint/ \
5
     --data-raw '{"query": "What are the main character names?" }'
6
"The main character names mentioned in the stories are Jessica, Karl and Rio."