Every workflow behind FluxChat

The main conversational agent — routes to the right agent (Shopify, WooCommerce, or a custom agent), calls your sub-agents and tools, and streams the reply back to the SDK or widget.

• Realtime chat over the streaming response
• Automatic routing to a Shopify, WooCommerce-specific, or custom agent and its MCP client/tool
• Image input handling for photo-based questions (supports resizing before sending)
• Multiple LLM providers — OpenAI, Gemini, and custom models
• Model fallback if your primary LLM provider errors out
• Pluggable memory backend — Simple, Postgres, MongoDB, or Redis chat memory

A lighter-weight text-completion workflow for prompt-driven features — separate from the full chat agent, backed by a filterable prompt library.

• One-shot and streaming text generation endpoints
• A prompt library your app can pull suggested prompts from

Turns a text prompt into images, with interchangeable backends so you're not locked into one image model provider.

• OpenAI image generation
• Google Gemini ("Nano Banana") image generation
• Cloudflare Workers AI image generation
• Pollinations free image generation

Semantic product search — your agent answers "do you have anything like..." questions from a real vector index of your catalog, not keyword matching.

• Text-embedding index of your product catalog
• Text-in, matching-products-out search tool
• Shopify/WooCommerce catalog sync with pagination
• Incremental sync so new/changed products stay indexed
• A dedicated API for product search

Visual product search — a shopper uploads a photo and your agent finds catalog items that look like it.

• Image-embedding index of your photos
• Photo-in, matching-products-out search tool
• Shopify/WooCommerce catalog sync with pagination
• Incremental sync so new/changed images stay indexed
• A dedicated API for image search

Upload PDFs, spreadsheets, CSVs, Markdown, or plain text and your agent answers from them — your policies, FAQs, and docs, not just your product catalog.

• Multi-source ingestion: file upload, Google Drive, GitHub
• Chunking, embeddings, and your choice of 7 vector stores
• A callable Knowledge Sub-Agent (RAG) for retrieval-augmented answers

A lighter alternative to full RAG for teams whose knowledge base already lives in a database like Notion — queries your pages and databases directly.

• Direct page and database lookups, no vector store to maintain

Handles "where's my order" questions by looking up real order data from your store.

• Order lookup by order ID or customer email
• Shopify (GraphQL/REST/MCP) and WooCommerce (REST/MCP) support

Answers detailed "tell me about this product" questions, pulling from your catalog API, image search, or document search — whichever fits the question.

• Shopify (REST/GraphQL/MCP) and WooCommerce (REST/MCP) product detail lookups
• Automatically calls Image Search or Document Search when relevant

When the bot can't help, escalate the full conversation to a human in a channel instead of leaving the customer stuck.

• Posts conversation history and context into a channel (Discord, Slack, Telegram, or email)
• Callable as a sub-workflow from the main AI Chat Agent

Turns your store or custom tools into MCP tools the agent can call directly — and can connect to external MCP servers too.

• MCP server + client wiring
• Custom tools