Llm Implementation

[ahmedalkawaz]

Context
This PR integrates an LLM-based translation service into NodeBB, enabling automatic detection and translation of non-English posts at creation time.

Description
When a user creates a post, NodeBB calls a separate Python Flask translator service powered by Ollama. The service classifies the language of the post and translates it to English if needed. Non-English posts display a "Click here to view the translated message" button that reveals the translation inline.

Changes
Created translator/ directory containing a standalone Flask translation microservice (src/translator.py, app.py, requirements.txt)
LLM integration uses Ollama with llama3.1:8b model for language classification and translation via structured prompts
Robust error handling: validates LLM output against a known languages list, falls back to assuming English on gibberish responses, empty translations, or service errors
Updated src/translate/index.js to call the Flask service asynchronously with graceful fallback if the service is unreachable
Unit tests in translator/test/unit/test_translator.py covering normal responses, English detection, gibberish handling, timeouts, empty translations, and empty input

Testing
Unit tests: cd translator && uv run pytest (7 tests passing)
Integration: requires three processes running simultaneously:
ollama serve (LLM backend)
cd translator && uv run flask run (translator service on port 5000)
NodeBB (./nodebb build && ./nodebb start)
Create a non-English post (e.g. "Dies ist eine deutsche Nachricht.") and verify the translate button appears with the English translation