peace2024's picture
saving changes
992bd88
import random
import logging
import json
import os
from typing import List
from fastapi import APIRouter, Depends, HTTPException
from pydantic import BaseModel
from sqlalchemy import select, delete
from sqlalchemy.ext.asyncio import AsyncSession
from app.auth import get_current_user
from app.database import get_db
from app.models import Feed, User
from langchain_groq import ChatGroq
from langchain_openai import OpenAIEmbeddings
from langchain_community.vectorstores import SupabaseVectorStore
from supabase.client import create_client
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_core.prompts import ChatPromptTemplate
try:
import wikipediaapi
except Exception:
wikipediaapi = None
# Logger early init (used during module init below)
logger = logging.getLogger("app.feeds")
if not logger.handlers:
handler = logging.StreamHandler()
formatter = logging.Formatter("[%(asctime)s] %(levelname)s - %(message)s")
handler.setFormatter(formatter)
logger.addHandler(handler)
logger.setLevel(logging.INFO)
# Supabase Initialization (independent of other modules)
SUPABASE_URL = os.getenv("SUPABASE_URL")
SUPABASE_KEY = os.getenv("SUPABASE_KEY")
supabase_client = None
if SUPABASE_URL and SUPABASE_KEY:
try:
supabase_client = create_client(SUPABASE_URL, SUPABASE_KEY)
logger.info("✅ Supabase client (feeds) initialized.")
except Exception as e:
logger.warning(f"Supabase init failed in feeds: {e}")
router = APIRouter()
class FeedOut(BaseModel):
id: int
title: str
content: str
media_url: str | None = None
tags: list[str] | None = None
category: str | None = None
source: str | None = None
relevance_score: int
likes: int
comments_count: int
shares: int
class Config:
from_attributes = True
def _fallback_generate(user: User) -> List[dict]:
topics = [
"career-growth", "interview-prep", "project-ideas", "resume-tips",
"networking", "internships", "leetcode", "system-design", "ai-ml",
"cloud", "devops", "frontend", "backend"
]
base_tags = ["tips", "guide", "beginner", "advanced", "2025", "best-practices"]
user_topic = (user.preparing_for or "career-growth").lower().replace(" ", "-")
chosen_topics = list({user_topic} | set(random.sample(topics, k=4)))
items: List[dict] = []
for _ in range(25):
topic = random.choice(chosen_topics)
title = f"{topic.title()} insights for {user.name or 'you'}"
content = (
f"Actionable {topic.replace('-', ' ')} advice tailored for {user.name or 'your profile'}. "
f"Focus on consistent practice, portfolio building, and networking."
)
tags = list({topic, *(random.sample(base_tags, k=3))})
items.append({
"title": title[:200],
"content": content,
"media_url": None,
"tags": ",".join(tags),
"category": topic,
"source": "dubsway-ai",
"relevance_score": random.randint(70, 100),
"likes": random.randint(0, 50),
"comments_count": random.randint(0, 20),
"shares": random.randint(0, 10),
})
return items
def _agentic_generate(user: User) -> List[dict]:
"""Generate context-aware feeds using vector DB, public info, and LLM.
Falls back to heuristic generation if LLM not available.
"""
# 1) Gather user context
user_focus = (user.preparing_for or "career growth").strip()
# 2) Retrieve user-related docs from Supabase vector store via LangChain
doc_snippets = ""
try:
if supabase_client is not None:
embeddings = OpenAIEmbeddings()
vector_store = SupabaseVectorStore(
client=supabase_client,
embedding=embeddings,
table_name="documents",
query_name="match_documents",
)
retriever = vector_store.as_retriever(search_kwargs={"k": 8, "filter": {"user_id": user.id}})
# Use a simple seed question from the user's focus
seed_q = f"Key takeaways for {user_focus}"
retrieved = retriever.invoke(seed_q)
# retrieved may be list of docs depending on LC version
docs = retrieved if isinstance(retrieved, list) else retrieved.get("context", [])
parts: List[str] = []
for d in docs[:10]:
try:
parts.append(getattr(d, "page_content", "")[:400])
except Exception:
pass
doc_snippets = "\n".join(parts)[:4000]
except Exception as e:
logger.warning(f"Supabase retrieval failed: {e}")
# 3) Pull public info (Wikipedia summary on user focus)
public_summary = ""
try:
if wikipediaapi:
wiki = wikipediaapi.Wikipedia(language='en', user_agent='DubswayVideoAI/1.0 (contact: support@dubsway.ai)')
page = wiki.page(user_focus)
if page and page.exists():
public_summary = page.summary[:1200]
except Exception as e:
logger.warning(f"Wikipedia fetch failed: {e}")
# 4) Use Groq LLM to synthesize 25 feeds
try:
llm = ChatGroq(groq_api_key=os.getenv("GROQ_API_KEY"), model_name=os.getenv("GROQ_MODEL", "llama-3.3-70b-versatile"))
system = (
"You are a career coach assistant. Create 25 short personalized feed items that help the user grow "
"in their career. Each item should be practical and contextual to the user focus, using the given "
"context notes and public info. Output strictly JSON array with objects having keys: title, content, "
"tags (array of strings), category, source."
)
prompt = (
f"User name: {user.name or 'User'}\n"
f"User focus: {user_focus}\n\n"
f"Context from user's vector docs (may be empty):\n{doc_snippets}\n\n"
f"Public info (may be empty):\n{public_summary}\n\n"
f"Generate 25 items. Keep title <= 120 chars and content 1-2 sentences."
)
resp = llm.invoke([{"role": "system", "content": system + " Respond ONLY with a JSON array."}, {"role": "user", "content": prompt}])
text = resp.content if hasattr(resp, "content") else str(resp)
# Normalize potential markdown code fences and extract JSON array
text_stripped = text.strip()
if text_stripped.startswith("```)" ):
text_stripped = text_stripped.strip('`')
if text_stripped.startswith("```json"):
text_stripped = text_stripped[7:]
if text_stripped.startswith("```") and text_stripped.endswith("```"):
text_stripped = text_stripped[3:-3]
# Attempt to find a JSON array inside
try:
data = json.loads(text_stripped)
except Exception:
start = text_stripped.find('[')
end = text_stripped.rfind(']')
if start != -1 and end != -1 and end > start:
data = json.loads(text_stripped[start:end+1])
else:
raise
items: List[dict] = []
for it in data[:25]:
tags_joined = ",".join(it.get("tags", [])[:6]) if isinstance(it.get("tags"), list) else None
items.append({
"title": (it.get("title") or "").strip()[:200] or "Career insight",
"content": (it.get("content") or "").strip() or "Practical career tip.",
"media_url": None,
"tags": tags_joined,
"category": (it.get("category") or user_focus)[:64],
"source": (it.get("source") or "agentic-ai")[:64],
"relevance_score": random.randint(80, 100),
"likes": 0,
"comments_count": 0,
"shares": 0,
})
# Ensure we always return 25
if len(items) < 25:
items.extend(_fallback_generate(user)[: 25 - len(items)])
return items[:25]
except Exception as e:
logger.error(f"Agentic generation failed: {e}")
return _fallback_generate(user)
@router.get("/feeds", response_model=List[FeedOut])
async def get_feeds(current_user: User = Depends(get_current_user),
db: AsyncSession = Depends(get_db)):
# Auto-refresh: clear previous feeds for this user
try:
await db.execute(delete(Feed).where(Feed.user_id == current_user.id))
await db.commit()
except Exception as e:
await db.rollback()
logger.error(f"Failed deleting old feeds: {e}")
raise HTTPException(status_code=500, detail="Failed refreshing feeds")
# Generate new 25 items (agentic with fallback)
items = _agentic_generate(current_user)
# Insert into DB
try:
feed_rows = [
Feed(user_id=current_user.id, **item) for item in items
]
db.add_all(feed_rows)
await db.commit()
except Exception as e:
await db.rollback()
logger.error(f"Failed inserting feeds: {e}")
raise HTTPException(status_code=500, detail="Failed storing feeds")
# Return top 25 ordered by relevance desc, then recent
result = await db.execute(
select(Feed).where(Feed.user_id == current_user.id).order_by(Feed.relevance_score.desc(), Feed.id.desc()).limit(25)
)
rows = result.scalars().all()
# Convert comma tags to list for response
out: List[FeedOut] = []
for r in rows:
out.append(FeedOut(
id=r.id,
title=r.title,
content=r.content,
media_url=r.media_url,
tags=(r.tags.split(",") if r.tags else None),
category=r.category,
source=r.source,
relevance_score=r.relevance_score or 0,
likes=r.likes or 0,
comments_count=r.comments_count or 0,
shares=r.shares or 0,
))
return out