fix: replace deprecated langchain_classic imports with modern LCEL

[Artibar]

Summary

• Remove unused PromptTemplate import from app.py
• Replace LLMChain + PromptTemplate in NewsService.py with modern LCEL syntax (prompt | llm | StrOutputParser())
• Update openai==0.27.8 to openai>=1.0.0 in requirements.txt

app.py used invalid import, however NewsService.py used deprecated
import from langchain_classic and requirements.txt pins openai==0.27.8 (2023).
This causes runtime errors and prevents running the app locally.

Description

app.py — removed invalid PromptTemplate import

• services/NewsService.py — replaced LLMChain + PromptTemplate
  from langchain_classic with modern LCEL syntax:
  chain = prompt | self.llm | StrOutputParser()
• requirements.txt — updated openai==0.27.8 to openai>=1.0.0

Related Issue

Fixes #197

Reference

Official LLMChain deprecation notice:
https://reference.langchain.com/python/langchain-classic/chains/llm/LLMChain
Assisted-By: Claude (syntax reference)

Motivation and Context

change is required due to deprecated import cause error and need to maintain up to date latest version

How Has This Been Tested?

• Cloned the repository fresh on Windows 11 with Python 3.11
• Ran pip install -r requirements.txt successfully after fix
• Ran python app.py — app starts without ImportError
• Verified app loads at http://127.0.0.1:5000 successfully
• Confirmed langchain_classic imports no longer present in
  source files using findstr /r "langchain_classic" app.py services\NewsService.py

Test environment:

• OS: Windows 11
• Python: 3.11.9
• pip: 26.0.1

Screenshots (if appropriate):

• Before fix — deprecated imports found:

• After Fix — app reloads successfully without ImportError:

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have read the CONTRIBUTING document.
• I have added tests to cover my changes.
• All new and existing tests passed.

[shahryarkhalid-cmd]

Hi @Artibar, I tested this PR locally against main and wanted to share results.

Confirmed working:

• The unused PromptTemplate import removal in app.py is correct.
• langchain_core.prompts.PromptTemplate, langchain_core.output_parsers.StrOutputParser, and langchain_community.llms.HuggingFaceEndpoint all import cleanly with this fix.
• The chain = prompt | self.llm | StrOutputParser() LCEL pattern runs correctly and returns a plain string, which correctly resolves the old output['text'] mismatch since toJSON() expects a string input.

Found while testing — not fixed by this PR:
toJSON() still has a bug independent of this fix. It unconditionally does news_list_json.pop() at the end, which drops the last parsed item every time. I tested it with realistic LLM-style output:

• 2 valid news items in → only 1 returned
• 1 valid news item in → empty list returned

This means even after this PR's fix, the /news and /news_keywords endpoints will still lose at least one article, and single-result queries will return nothing.

Separately (not related to this PR): installing the full requirements.txt as-is triggers a long pip backtracking loop across unpinned langchain/langchain-community/sentence-transformers versions — looks like the same issue described in #211.

Would you like me to open a follow-up PR for the toJSON() fix, or would you prefer to fold it into this one? Happy to help either way — let me know if this PR is still active on your end.

[shahryarkhalid-cmd]

@Artibar @hardik1408

I tested this PR locally against main. The old code used deprecated langchain_classic imports that broke at runtime — this PR's LCEL fix resolves that correctly, confirmed the imports work and the chain runs.

While testing, I found a separate bug in toJSON() that's still present: it always drops the last parsed news item (news_list_json.pop() at the end runs unconditionally). Tested with 2 items → only 1 returned; tested with 1 item → empty list returned.

Happy to open a follow-up PR for the toJSON fix if useful, or fold it into this one — let me know what you'd prefer.