Docs

Node Reference

Every flow is built by wiring nodes together. Here is the complete list with config, inputs, outputs and emitted events.

27 nodes

Agents

Conversational agents with memory, tools and reasoning.

ai_agent

AI Agent

Conversational agent with session memory, retrieved-context injection and sub-flow tool-calling.
collect_data

Collect Data

Conversational agent that gathers structured fields and remembers facts via hidden tools.

Actions

Call APIs, search your knowledge, transcribe audio, extract text from images.

a_api

HTTP Request

Calls an external HTTP endpoint and routes on OK vs error status.
a_kb_search

Knowledge Search

Embeds the query and retrieves top-K chunks from a knowledge base.
a_structured_search

Structured Search

Runs parametrised SQL against searchable records for filter/sort/range queries.
a_scrape

Web Scraper

Fetches and extracts clean content from a URL (Firecrawl or basic fetch).
a_transcribe_stream

Live Transcription

Streams session audio through a real-time STT provider and emits transcripts.
a_transcribe_file

Transcribe File

One-shot transcription of a pre-recorded audio file (URL or base64).
a_ocr_extract

Extract Text (OCR)

Extracts text from a PDF or image via Azure Document Intelligence.
a_form_ui

Form (Widget UI)

Collects structured data from the user via a graphical form rendered in the widget.
a_form_chat

Form (Conversational)

Collects structured data over a text channel by asking the next missing field as a natural-language question.
a_order_lookup

a_order_lookup

Decisions

Route the flow based on conditions evaluated over memory or input.

condition

Condition

Evaluates dot-path clauses against memory/input and routes to a matching branch.
word_counter

Word Counter

Accumulates words across turns and fires a branch once a threshold is reached.

Transforms

LLM, embeddings, rerank, vision, image, voice and video generation.

a_parse_template

Parse Template

Turns a [[field: description]] template string into an OpenAI JSON schema for collect_data.
a_llm

LLM Call

Single-shot LLM completion. Pair with memory_key to persist the reply.
a_embed_text

Embed Text

Embeds a batch of texts with the configured provider, returns vectors + dim.
a_rerank

Rerank

Cross-encoder rerank of candidates against a query.
a_recommend

Recommend

LLM-reasoned recommendations from a candidate list with a greeting line.
a_tts_stream

Text-to-Speech Stream

Streams text-to-speech audio chunks via a TTS provider (Deepgram).
a_kb_ingest

Ingest to knowledge base

Upserts pre-embedded chunks into the target knowledge base.
a_vision

Vision Analysis

Analyses an image (URL/base64) against a natural-language prompt.
a_generate_image

Generate Image

Text-to-image or image-edit via nano-banana / DALL-E. Persists to R2 by default.
a_tts

Text to Speech

Synthesises speech audio from text via Deepgram / OpenAI / ElevenLabs.
a_generate_video

Generate Video

Text-to-video or image-to-video via LTX / Runway. Persists to R2 by default.

Merge

Fan-in barriers that wait for several branches before merging.

wait_for_n

Wait For N

Fan-in barrier: waits for N upstream branches, merges payloads, fires once.

Split

Split the flow into multiple parallel branches.

fan_out

Fan Out

Explicit parallel split: broadcasts the same input to every outgoing edge.