nmtechsolutionstuc/botnm

Purpose and Runtime

makebufferbot is a Node.js Telegram long-polling bot that turns received media and user direction into Buffer drafts requiring manual approval. It also supports Telegram-group publishing and a separate YouTube draft flow. The process starts in src/index.js, which launches the bot plus HTTP endpoints for health checks and temporary media delivery.

Conversation State and Routing

Conversation progress is persisted per chat/user in src/state/conversationStore.js (data/state.json at runtime). The main router and photo-post workflow live in src/bot/handlers.js: collect one or more photos, choose a Sheet-defined mode, collect text/audio, select immediate or scheduled timing, optionally send to a group, then create Buffer drafts. Shared state transitions and UI/message utilities are centralized in src/bot/shared.js.

Mode Configuration Boundary

Google Sheets is the dynamic business configuration source: modes determine prompts, fixed copy, permitted Telegram users, and Buffer channel IDs. src/services/sheets.js reads and caches these rows; it allows adding brands/modes without code changes. Configuration/environment validation and service credentials are handled through src/config.js.

AI Generation and Fallback

AI operations are abstracted so Gemini is preferred but Cloudflare is a resilient fallback. src/services/ai.js orchestrates text generation, audio transcription, and trivia generation across providers. src/services/gemini.js handles Gemini image enhancement and content tasks; src/services/cloudflare.js handles Workers AI image generation/editing and fallback text/audio tasks. This boundary keeps bot flows provider-agnostic while exposing which fallback was used.

Image Command Subflow

src/bot/imageCommands.js implements /generar, /editar, and /mejorar. It creates or transforms media through Cloudflare, rate-limits requests with src/services/rateLimit.js, and requires user confirmation before adding the resulting image into the normal post-drafting conversation. Gemini image-enhancement failures similarly fall back to Cloudflare before retaining the original.

Publishing and Public-Media Boundary

src/services/buffer.js creates pending/scheduled Buffer posts through Buffer’s GraphQL API, using channel IDs selected from the mode configuration. Buffer needs publicly downloadable media rather than raw uploaded bytes; src/services/mediaHost.js temporarily exposes generated/downloaded media via the HTTP endpoint rooted at src/index.js (/media/:id). PUBLIC_BASE_URL is therefore required for media-containing Buffer drafts.

Alternative Delivery Flows

Telegram group posting is intentionally separate from Buffer delivery. src/bot/groupCommands.js parses direct group text/photo, polls, and quizzes; src/services/telegramGroup.js sends confirmed content to the configured Telegram group. src/bot/youtubeFlow.js is a separate state flow for incoming videos, producing distinct AI title/description fields and publishing through a separate YouTube Buffer token/channel configuration.

Infrastructure and External Services

The service depends on Telegram Bot API, Google Sheets API via a service account, Google Gemini, Cloudflare Workers AI, and Buffer GraphQL. It is designed for a persistent Node host such as Render; /health in src/index.js supports external uptime monitoring to prevent free-tier sleep from interrupting Telegram polling. Local persistent state is under data/, so deployments need durable storage if conversation recovery across restarts matters.