File size: 3,343 Bytes
09a9add 4b268d1 09a9add 4b268d1 09a9add 4b268d1 09a9add 4b268d1 09a9add 4b268d1 09a9add 4b268d1 09a9add | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 | from fastapi import FastAPI, UploadFile, File
import pandas as pd
import faiss
import pickle
import os
import json
from sentence_transformers import SentenceTransformer
from exa_py import Exa
from groq import Groq
app = FastAPI()
# =============================
# 🔑 KEYS
# =============================
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
EXA_API_KEY = os.getenv("EXA_API_KEY")
client = Groq(api_key=GROQ_API_KEY)
exa = Exa(api_key=EXA_API_KEY)
# =============================
# 🧠 LOAD MODELS
# =============================
embed_model = SentenceTransformer('all-mpnet-base-v2')
index = faiss.read_index("faiss_index.index")
with open("startup_texts.pkl", "rb") as f:
startup_texts = pickle.load(f)
# =============================
# 🔍 RETRIEVAL
# =============================
def retrieve_similar(problem, k=3):
vec = embed_model.encode([problem], convert_to_numpy=True)
distances, indices = index.search(vec, k)
return [
{"text": startup_texts[idx], "score": float(distances[0][i])}
for i, idx in enumerate(indices[0])
]
# =============================
# 🌐 WEB SEARCH
# =============================
def search_web(query):
try:
response = exa.search(query, num_results=5)
return [r.text or r.summary or "" for r in response.results]
except:
return []
# =============================
# 🤖 QWEN (Groq)
# =============================
def ask_qwen(prompt):
completion = client.chat.completions.create(
model="qwen/qwen3-32b",
messages=[
{"role": "system", "content": "You are a strict fact-checking analyst. Return ONLY valid JSON."},
{"role": "user", "content": prompt}
],
temperature=0.3,
max_tokens=512
)
return completion.choices[0].message.content
# =============================
# 🧠 PIPELINE (FIXED)
# =============================
def analyze_problem(problem):
retrieved = retrieve_similar(problem)
if retrieved and retrieved[0]["score"] < 2.0:
context = "\n\n".join([r["text"] for r in retrieved])
else:
web = search_web(problem)
context = "\n\n".join(web[:3])
prompt = f"""
Problem:
{problem}
Evidence:
{context}
Return ONLY valid JSON in this format:
{{
"status": "SOLVED or UNSOLVED",
"reason": "one short sentence",
"gaps": ["gap1", "gap2"],
"new_problem": "rewrite of the problem"
}}
"""
response = ask_qwen(prompt)
try:
parsed = json.loads(response)
except:
parsed = {
"status": "ERROR",
"reason": "invalid JSON from model",
"gaps": [],
"new_problem": ""
}
parsed["problem"] = problem
return parsed
# =============================
# 🚀 API
# =============================
@app.post("/analyze")
async def analyze(file: UploadFile = File(...)):
df = pd.read_csv(file.file)
results = []
for p in df.iloc[:, 0].tolist():
try:
results.append(analyze_problem(p))
except Exception as e:
results.append({
"problem": p,
"status": "ERROR",
"reason": str(e),
"gaps": [],
"new_problem": ""
})
return {
"success": True,
"results": results
} |