Flowise V2

docs.flowiseai.com covers chatflow building guide, API reference, vector store integrations, Docker deployment, and environment configuration. Getting started (self-hosted): docker run -d -p 3000:3000 flowiseai/flowise — UI at localhost:3000 in under 2 minutes. Flowise Cloud: free tier. GitHub: FlowiseAI/Flowise (38k+ stars). Community via Discord (very active). Good documentation; visual flow building guide particularly thorough.

Flowise lets non-engineers build LLM chatbots and agents visually, then exposes them as REST API endpoints for programmatic use. For agents: POST /api/v1/prediction/{chatflowId} sends a question to a Flowise chatflow and receives an LLM answer — Flowise handles the LangChain/LlamaIndex orchestration, vector retrieval, memory, and model calls internally. Supports RAG flows (upload documents to built-in vector stores), agent flows (tool-using agents), and conversational flows with memory. Self-hosted Docker or Flowise Cloud. 38k+ GitHub stars. Confidence is docs-derived.

REST API at {flowise_host}/api/v1. Prediction: POST /api/v1/prediction/{chatflowId} body: {"question": "user question", "overrideConfig": {}, "history": []}. Response: {"text": "LLM answer", "chatId": "...", "chatMessageId": "...", "sourceDocuments": [...]}. Streaming: POST /api/v1/prediction/{chatflowId} with Accept: text/event-stream returns SSE. Vector store upsert: POST /api/v1/vector/upsert/{chatflowId} body: FormData with files to add to flow's vector store. Get chatflows: GET /api/v1/chatflows lists all flows. Message history: GET /api/v1/chatmessage/{chatflowId} retrieves conversation history. Feedback: POST /api/v1/feedback/{chatflowId}/{chatmessageId} adds thumbs up/down feedback. Webhook: configure webhook URL in flow settings for async completion notification.

Auth options: (1) No auth — local/internal dev use. (2) Username/password: FLOWISE_USERNAME + FLOWISE_PASSWORD env vars (Basic Auth). (3) API key: Create in Flowise Settings → API Keys; pass Authorization: Bearer {api_key} header. (4) Flowise Cloud: managed auth. HTTPS: deploy behind reverse proxy for TLS (not included in default Docker setup). Chatflow-level API key override: individual flows can require specific API keys. No OAuth2 or OIDC support natively.

Prediction latency: depends on underlying LLM and flow complexity (1–30s typical). Streaming: SSE reduces perceived latency for long responses. Memory: conversational flows store session history in-process (ephemeral) or persistent DB (configure PostgreSQL/MySQL for Flowise). Vector stores: external (Pinecone, Weaviate, etc.) or in-memory (Faiss, Chroma local). Cold start: first prediction may be slower due to model loading. Self-hosted reliability: Docker single-container (stateless; use PostgreSQL for persistence). Flowise Cloud uptime: status.flowiseai.com. Error format: {message: "Error: ...", stack: "..."} for uncaught flow errors.