O

ONLYOFFICE Answers

Loading...
🤖
Hi! I'm the ONLYOFFICE Answers assistant. Ask me anything about ONLYOFFICE products — I'll search our support knowledge base and give you the best answer I can.

Try questions like:
  • "How to enable SSO with SAML?"
  • "Document server connection issues"
  • "JWT token configuration"
  • "How to migrate from Community Server?"

📊 Knowledge Base Analytics

Tickets by Channel

Tickets by Status

Tickets Over Time

Top Tags

About ONLYOFFICE Answers

ONLYOFFICE Answers is an AI-powered knowledge base search portal built from real support ticket data. It uses semantic search (sentence-transformers) to find the most relevant tickets based on meaning, not just keywords.

How It Works

  • AI Answer Synthesis — Ask a question in natural language. The system retrieves the most relevant support tickets using semantic search, then GPT-4o-mini reads the actual support conversations and synthesizes a clear, accurate answer. Sources are cited so you can verify.
  • Semantic Search — Uses the all-MiniLM-L6-v2 model to encode 1,781 tickets (with 6,285 conversation messages) into vector embeddings. Finds tickets by meaning, not just keywords — so "can't open documents" matches "file loading error" even without shared words.
  • Full Conversation History — Each ticket includes the complete support conversation thread. Click any result to see every message between the customer and support team, exactly as it happened.
  • Analytics — Visual breakdown of ticket volume by channel, status, timeline, and trending tags.

Data Source

Built from 1,781 ONLYOFFICE support tickets (November 2025) containing 6,285 conversation messages across chat, email, and web channels. All conversations are preserved to help find solutions to common issues faster.

Technology

  • Answer Synthesis: GPT-4o-mini (OpenAI) — reads top 5 matching ticket conversations and generates a concise answer (~$0.0003 per query)
  • Embeddings: sentence-transformers (all-MiniLM-L6-v2) — runs locally on our server, no data sent externally for search
  • Vector Search: NumPy cosine similarity over ticket + comment embeddings
  • Database: SQLite with full-text ticket and comment storage
  • Backend: Python stdlib HTTP server
  • Frontend: Single-file HTML + CSS + Chart.js
  • Hosting: Self-hosted on dedicated infrastructure

Privacy

This portal is for internal use. Semantic search and embeddings run entirely on our own server — no ticket data leaves the infrastructure for search operations. Answer synthesis sends relevant ticket excerpts to OpenAI's API for response generation only.