master.py

# from fastapi import FastAPI
# from fastapi.middleware.cors import CORSMiddleware
# from pydantic import BaseModel
# from sentence_transformers import SentenceTransformer, util

# app = FastAPI()

# app.add_middleware(
#     CORSMiddleware,
#     allow_origins=["http://localhost:5173"],
#     allow_credentials=True,
#     allow_methods=["*"],
#     allow_headers=["*"],
# )

# model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")

# class Profile(BaseModel):
#     name: str
#     budget: float
#     lifestyle: dict
#     interests: list

# class CompatibilityRequest(BaseModel):
#     user_profile: Profile
#     candidate_profiles: list[Profile]

# @app.post("/compute_compatibility")
# def compute_compatibility(data: CompatibilityRequest):
#     scores = []
#     user_text = f"Budget: {data.user_profile.budget}, Lifestyle: {data.user_profile.lifestyle}, Interests: {', '.join(data.user_profile.interests)}"
#     user_embedding = model.encode(user_text, convert_to_tensor=True)

#     for candidate in data.candidate_profiles:
#         candidate_text = f"Budget: {candidate.budget}, Lifestyle: {candidate.lifestyle}, Interests: {', '.join(candidate.interests)}"
#         candidate_embedding = model.encode(candidate_text, convert_to_tensor=True)

#         similarity_score = util.pytorch_cos_sim(user_embedding, candidate_embedding).item()
#         match_reasons = []

#         if similarity_score > 0.7:
#             match_reasons.append("Strong compatibility based on overall profile match")
#         elif similarity_score > 0.4:
#             match_reasons.append("Moderate compatibility with some common aspects")
#         else:
#             match_reasons.append("Low compatibility due to differing aspects")

#         scores.append({
#             "profile": candidate.name,
#             "compatibility": round(similarity_score * 100),
#             "matchReasons": match_reasons
#         })

#     return {"all_matches": scores}

# if __name__ == "__main__":
#     import uvicorn
#     uvicorn.run("master:app", host="127.0.0.1", port=8000, reload=True)

from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
import torch

app = FastAPI()

app.add_middleware(
    CORSMiddleware,
    allow_origins=["http://localhost:5173"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

model_name = "meta-llama/Meta-Llama-3-8B-Instruct"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
generator = pipeline("text-generation", model=model, tokenizer=tokenizer)

class Profile(BaseModel):
    name: str
    budget: float
    lifestyle: dict
    interests: list

class CompatibilityRequest(BaseModel):
    user_profile: Profile
    candidate_profiles: list[Profile]

@app.post("/compute_compatibility")
def compute_compatibility(data: CompatibilityRequest):
    scores = []
    user_text = f"Budget: {data.user_profile.budget}, Lifestyle: {data.user_profile.lifestyle}, Interests: {', '.join(data.user_profile.interests)}"

    for candidate in data.candidate_profiles:
        candidate_text = f"Budget: {candidate.budget}, Lifestyle: {candidate.lifestyle}, Interests: {', '.join(candidate.interests)}"
        
        prompt = f"Compare the following profiles and rate their compatibility from 0 to 100:\nUser: {user_text}\nCandidate: {candidate_text}\nCompatibility Score:"
        response = generator(prompt, max_length=50, do_sample=True)
        compatibility_score = int(''.join(filter(str.isdigit, response[0]["generated_text"])))

        scores.append({
            "profile": candidate.name,
            "compatibility": compatibility_score,
            "matchReasons": f"Generated by Llama-3 based on textual profile similarities"
        })

    return {"all_matches": scores}

if __name__ == "__main__":
    import uvicorn
    uvicorn.run("master:app", host="127.0.0.1", port=8000, reload=True)