LLM Chatbot Using DeepSeek, Google Docs, and Telegram for Personalized Conversations

This workflow builds a persistent AI chatbot with long-term memory and note storage, replacing fragmented chat histories in apps like Slack or basic bots that forget context, leading to repetitive queries and 10-15 hours weekly lost on recaps for support teams handling 200+ interactions. It triggers on messages, validates inputs, fetches memories/notes from Google Docs, aggregates context, uses session buffers for recency, routes to OpenAI/DeepSeek models via Langchain agent for personalized replies, stores new memories/notes via tools, and responds via Telegram. Key nodes: Chat Trigger for inputs, IF for validation, Google Docs for retrieval/insert (memories URL, notes URL), Merge/Aggregate for context, another IF for Telegram mode, lmChatOpenAi (GPT-4o-mini/DeepSeek, temp=0.7), MemoryBufferWindow (50-token window), GoogleDocsTools for storage (descriptions: 'Save important info', 'Save notes'), Agent with prompt for memory rules/privacy, Set for formatting, Telegram for output. This empowers customer support or internal teams at SMBs (20-100 users) managing ongoing chats, enhancing personalization and efficiency without custom coding.\n\nDeployment saves 8-12 hours/week on 150+ chats by retaining context, boosting satisfaction 30% via relevant responses; ROI in 1 month for service desks. Fits remote teams or VAs tracking client prefs. Needs OpenAI ($0.02/1k tokens, GPT-4o-mini), DeepSeek (similar), Google Workspace ($6/user/mo Docs), Telegram Bot (free), n8n Cloud ($20/mo). Scales to 500 chats/day but cap Docs edits (100/min); add Redis for high-volume memory.\n\nDocker n8n (n8n.io/download: docker run -p 5678:5678 n8nio/n8n) or cloud.n8n.io (5-min). OpenAI: platform.openai.com/api-keys (creds for lmChatOpenAi); DeepSeek: platform.deepseek.com (separate creds). Google Docs OAuth: n8n credentials, create two docs (Memories/Notes) as tables, get URLs, set $vars.GOOGLE_DOCS_MEMORY_URL etc. Telegram: BotFather for token, set $vars.TELEGRAM_ENABLED=true, CHAT_ID. Import JSON, connect models to Agent (system prompt for rules), tools to Agent. Test: Chat input 'Remember my birthday is Jan 1', verify Docs insert; follow-up 'What's my birthday?' for recall.\n\nTest: Trigger manual, input via Chat Trigger, check aggregation/response. Errors: 401 Docs (re-auth), invalid input (IF fallback), 429 OpenAI (throttle). Activate, monitor vars. Maintain: Backup Docs monthly, prune old entries; update prompts for new privacy regs. Optimize: Switch models dynamically; scale with n8n queues for concurrent chats.", "businessValue": "Saves 10 hours/week on 200 chats, retains 95% context, improves response relevance 40%", "setupTime": "45-60 minutes", "difficulty": "Intermediate", "requirements": ["OpenAI API key (~$0.02/1k tokens)", "Google Docs OAuth2 for memory/notes", "Telegram bot token and chat ID", "n8n Cloud or self-hosted"], "useCase": "Persistent chatbot for customer support with memory to track user preferences and notes"