AI & LLM

AI model integration, text generation, embeddings, and autonomous agents.

18 modules

Module	Description
Autonomous Agent	Self-directed AI agent with memory and goal-oriented behavior
Chain Agent	Sequential AI processing chain with multiple steps
Tool Use Agent	AI Agent that can call tools/functions
Text Embeddings	Generate vector embeddings from text using AI models
AI Extract	Extract structured data from text using AI
Local Ollama Chat	Chat with local LLM via Ollama (completely offline)
AI Memory	Conversation memory for AI Agent
Entity Memory	Extract and track entities (people, places, concepts) from conversations
Redis Memory	Persistent conversation memory using Redis storage
Vector Memory	Semantic memory using vector embeddings for relevant context retrieval
AI Model	LLM model configuration for AI Agent
AI Tool	Expose a module as a tool for AI Agent
Vision Analyze	Analyze images using AI vision models
Claude Chat	Send a chat message to Anthropic Claude AI and get a response
Google Gemini Chat	Send a chat message to Google Gemini AI and get a response
OpenAI Chat	Send a chat message to OpenAI GPT models
DALL-E Image Generation	Generate images using DALL-E
AI Agent	Autonomous AI agent with multi-port connections (model, memory, tools)

Modules

Autonomous Agent

agent.autonomous

Self-directed AI agent with memory and goal-oriented behavior

Parameters:

Name	Type	Required	Default	Description
goal	string	Yes	-	The goal for the agent to achieve
context	string	No	-	The goal for the agent to achieve
max_iterations	number	No	5	Additional context or constraints
llm_provider	select (openai, ollama)	No	openai	Maximum reasoning steps
model	string	No	gpt-4-turbo-preview	Model name (e.g., gpt-4, llama2, mistral)
ollama_url	string	No	http://localhost:11434	Model name (e.g., gpt-4, llama2, mistral)
temperature	number	No	0.7	Ollama server URL (only for ollama provider)

Output:

Field	Type	Description
result	string	Creativity level (0-2)
thoughts	array	The operation result
iterations	number	The operation result
goal_achieved	boolean	Agent reasoning steps

Example: Research task

goal: Research the latest trends in AI and summarize the top 3
max_iterations: 5
model: gpt-4

Example: Problem solving

goal: Find the best approach to optimize database queries
context: PostgreSQL database with 10M records
max_iterations: 10

Chain Agent

agent.chain

Sequential AI processing chain with multiple steps

Parameters:

Name	Type	Required	Default	Description
input	string	Yes	-	Initial input for the chain
chain_steps	array	Yes	-	Initial input for the chain
llm_provider	select (openai, ollama)	No	openai	Array of processing steps (each is a prompt template)
model	string	No	gpt-4-turbo-preview	Model name (e.g., gpt-4, llama2, mistral)
ollama_url	string	No	http://localhost:11434	Model name (e.g., gpt-4, llama2, mistral)
temperature	number	No	0.7	Ollama server URL (only for ollama provider)

Output:

Field	Type	Description
result	string	Creativity level (0-2)
intermediate_results	array	The operation result
steps_completed	number	The operation result

Example: Content pipeline

input: AI and machine learning trends
chain_steps: ["Generate 5 blog post ideas about: {input}", "Take the first idea and write a detailed outline: {previous}", "Write an introduction paragraph based on: {previous}"]
model: gpt-4

Example: Data analysis chain

input: User behavior data shows 60% bounce rate
chain_steps: ["Analyze what might cause this issue: {input}", "Suggest 3 solutions based on: {previous}", "Create an action plan from: {previous}"]

Tool Use Agent

agent.tool_use

AI Agent that can call tools/functions

Parameters:

Name	Type	Required	Default	Description
prompt	string	Yes	-	The goal or task for the agent
tools	array	Yes	-	List of tool definitions [{name, description, parameters}]
provider	select (openai, anthropic)	No	openai	LLM provider for the agent
model	string	No	gpt-4o	Model to use
api_key	string	No	-	API key (falls back to environment variable)
max_iterations	number	No	10	Maximum number of tool call rounds
system_prompt	string	No	-	Optional system prompt to guide the agent

Output:

Field	Type	Description
result	string	The agent final response
tool_calls	array	All tool calls made during execution
iterations	number	Number of iterations completed
model	string	Model used

Example: File Processing Agent

prompt: Read the config file and update the version number
tools: [{"name": "read_file", "description": "Read contents of a file", "parameters": {"type": "object", "properties": {"path": {"type": "string", "description": "File path"}}, "required": ["path"]}}, {"name": "write_file", "description": "Write contents to a file", "parameters": {"type": "object", "properties": {"path": {"type": "string", "description": "File path"}, "content": {"type": "string", "description": "File content"}}, "required": ["path", "content"]}}]
provider: openai
model: gpt-4o
max_iterations: 5

Text Embeddings

ai.embed

Generate vector embeddings from text using AI models

Parameters:

Name	Type	Required	Default	Description
text	string	Yes	-	Text to embed
provider	select (openai, local)	No	openai	AI provider for embeddings
model	string	No	text-embedding-3-small	Embedding model to use
api_key	string	No	-	API key (falls back to environment variable)
dimensions	number	No	-	Embedding dimensions (for models that support it)

Output:

Field	Type	Description
embeddings	array	Vector embedding array
model	string	Model used for embedding
dimensions	number	Number of dimensions in the embedding vector
token_count	number	Number of tokens processed

Example: Single Text Embedding

text: The quick brown fox jumps over the lazy dog
provider: openai
model: text-embedding-3-small

Example: Reduced Dimensions

text: Semantic search query
provider: openai
model: text-embedding-3-small
dimensions: 256

AI Extract

ai.extract

Extract structured data from text using AI

Parameters:

Name	Type	Required	Default	Description
text	string	Yes	-	Text to extract data from
schema	object	Yes	-	JSON schema defining the fields to extract
instructions	string	No	-	Additional extraction instructions
provider	select (openai, anthropic)	No	openai	AI provider to use
model	string	No	gpt-4o-mini	Model to use for extraction
api_key	string	No	-	API key (falls back to environment variable)
temperature	number	No	0	Sampling temperature (0-2)

Output:

Field	Type	Description
extracted	object	Extracted structured data
model	string	Model used for extraction
raw_response	string	Raw model response

Example: Extract Contact Info

text: John Smith is a senior engineer at Acme Corp. Email: john@acme.com
schema: {"type": "object", "properties": {"name": {"type": "string"}, "title": {"type": "string"}, "company": {"type": "string"}, "email": {"type": "string"}}}
provider: openai
model: gpt-4o-mini

Example: Extract Invoice Data

text: Invoice #1234 from Acme Corp. Total: $500.00. Due: 2024-03-01
schema: {"type": "object", "properties": {"invoice_number": {"type": "string"}, "vendor": {"type": "string"}, "total": {"type": "number"}, "due_date": {"type": "string"}}}
instructions: Extract all invoice fields. Parse amounts as numbers.

Local Ollama Chat

ai.local_ollama.chat

Chat with local LLM via Ollama (completely offline)

Parameters:

Name	Type	Required	Default	Description
prompt	string	Yes	-	The message to send to the local LLM
model	select (llama2, llama2:13b, llama2:70b, mistral, mixtral, codellama, codellama:13b, phi, neural-chat, starling-lm)	No	llama2	The message to send to the local LLM
temperature	number	No	0.7	Sampling temperature (0-2)
system_message	string	No	-	System role message (optional)
ollama_url	string	No	http://localhost:11434	System role message (optional)
max_tokens	number	No	-	Ollama server URL

Output:

Field	Type	Description
response	string	Maximum tokens in response (optional, depends on model)
model	string	Response from the operation
context	array	Response from the operation
total_duration	number	Model name or identifier
load_duration	number	Conversation context for follow-up requests
prompt_eval_count	number	Total processing duration
eval_count	number	Model loading duration

Example: Simple local chat

prompt: Explain quantum computing in 3 sentences
model: llama2

Example: Code generation with local model

prompt: Write a Python function to calculate fibonacci numbers
model: codellama
temperature: 0.2
system_message: You are a Python programming expert. Write clean, efficient code.

Example: Local reasoning task

prompt: What are the pros and cons of microservices architecture?
model: mistral
temperature: 0.7

AI Memory

ai.memory

Conversation memory for AI Agent

Parameters:

Name	Type	Required	Default	Description
memory_type	select (buffer, window, summary)	Yes	buffer	Type of memory storage
window_size	number	No	10	Number of recent messages to keep (for window memory)
session_id	string	No	-	Unique identifier for this conversation session
initial_messages	array	No	[]	Pre-loaded conversation history

Output:

Field	Type	Description
memory_type	string	Pre-loaded conversation history
session_id	string	Pre-loaded conversation history
messages	array	Type of memory
config	object	Session identifier

Example: Simple Buffer Memory

memory_type: buffer

Example: Window Memory (last 5 messages)

memory_type: window
window_size: 5

Entity Memory

ai.memory.entity

Extract and track entities (people, places, concepts) from conversations

Parameters:

Name	Type	Required	Default	Description
entity_types	multiselect	No	['person', 'organization', 'location']	Types of entities to extract and track
extraction_model	select (llm, spacy, regex)	Yes	llm	Model for entity extraction
session_id	string	No	-	Unique identifier for this memory session
track_relationships	boolean	No	True	Track relationships between entities
max_entities	number	No	100	Maximum number of entities to remember

Output:

Field	Type	Description
memory_type	string	Maximum number of entities to remember
session_id	string	Maximum number of entities to remember
entities	object	Type of memory (entity)
relationships	array	Session identifier
config	object	Tracked entities by type

Example: People & Organizations

entity_types: ["person", "organization"]
extraction_model: llm

Example: Full Entity Tracking

entity_types: ["person", "organization", "location", "concept"]
track_relationships: true
max_entities: 200

Redis Memory

ai.memory.redis

Persistent conversation memory using Redis storage

Parameters:

Name	Type	Required	Default	Description
redis_url	string	Yes	redis://localhost:6379	Redis connection URL
key_prefix	string	No	flyto:memory:	Prefix for all Redis keys
session_id	string	Yes	-	Unique identifier for this memory session
ttl_seconds	number	No	86400	Time-to-live for memory entries (0 = no expiry)
max_messages	number	No	100	Maximum messages to store per session
load_on_start	boolean	No	True	Load existing messages from Redis on initialization

Output:

Field	Type	Description
memory_type	string	Load existing messages from Redis on initialization
session_id	string	Load existing messages from Redis on initialization
messages	array	Type of memory (redis)
connected	boolean	Session identifier
config	object	Loaded message history

Example: Local Redis

redis_url: redis://localhost:6379
session_id: my-session
ttl_seconds: 3600

Example: Cloud Redis with Auth

redis_url: redis://:password@redis-cloud.example.com:6379
session_id: user-session
ttl_seconds: 86400
max_messages: 500

Vector Memory

ai.memory.vector

Semantic memory using vector embeddings for relevant context retrieval

Parameters:

Name	Type	Required	Default	Description
embedding_model	select (text-embedding-3-small, text-embedding-3-large, text-embedding-ada-002, local)	Yes	text-embedding-3-small	Model to use for generating embeddings
top_k	number	No	5	Number of most relevant memories to retrieve
similarity_threshold	number	No	0.7	Minimum similarity score (0-1) for retrieval
session_id	string	No	-	Unique identifier for this memory session
include_metadata	boolean	No	True	Include timestamp and other metadata with memories

Output:

Field	Type	Description
memory_type	string	Include timestamp and other metadata with memories
session_id	string	Include timestamp and other metadata with memories
embedding_model	string	Type of memory (vector)
config	object	Session identifier

Example: Default Vector Memory

embedding_model: text-embedding-3-small
top_k: 5

Example: High Precision Memory

embedding_model: text-embedding-3-large
top_k: 10
similarity_threshold: 0.85

AI Model

ai.model

LLM model configuration for AI Agent

Parameters:

Name	Type	Required	Default	Description
provider	select (openai, anthropic, ollama)	No	openai	AI model provider
model	string	No	gpt-4o	Specific model to use
temperature	number	No	0.7	Creativity level (0=deterministic, 1=creative)
api_key	string	No	-	API key (defaults to provider env var)
base_url	string	No	-	Custom API base URL (for Ollama or proxies)
max_tokens	number	No	4096	Maximum tokens in response

Output:

Field	Type	Description
provider	string	Maximum tokens in response
model	string	LLM provider name
config	object	LLM provider name

Example: OpenAI GPT-4

provider: openai
model: gpt-4o
temperature: 0.7

Example: Anthropic Claude

provider: anthropic
model: claude-3-5-sonnet-20241022
temperature: 0.5

AI Tool

ai.tool

Expose a module as a tool for AI Agent

Parameters:

Name	Type	Required	Default	Description
module_id	string	Yes	-	Module ID to expose as tool (e.g. http.request, data.json_parse)
tool_description	string	No	-	Custom description for the agent (overrides module default)

Output:

Field	Type	Description
module_id	string	Module ID exposed as tool

Example: HTTP Request Tool

module_id: http.request

Example: JSON Parse Tool

module_id: data.json_parse

Vision Analyze

ai.vision.analyze

Analyze images using AI vision models

Parameters:

Name	Type	Required	Default	Description
image_path	string	No	-	Local path to image file
image_url	string	No	-	URL of the image to analyze
prompt	string	No	Describe this image in detail	What to analyze or ask about the image
provider	select (openai, anthropic)	No	openai	AI provider for vision analysis
model	string	No	gpt-4o	Vision model to use
api_key	string	No	-	API key (falls back to environment variable)
max_tokens	number	No	1000	Maximum tokens in response
detail	select (low, high, auto)	No	auto	Image detail level (low/high/auto)

Output:

Field	Type	Description
analysis	string	AI analysis of the image
model	string	Model used for analysis
provider	string	Provider used for analysis
tokens_used	number	Number of tokens used

Example: Analyze Screenshot

image_path: /tmp/screenshot.png
prompt: Describe what you see in this UI screenshot
provider: openai
model: gpt-4o

Example: Analyze from URL

image_url: https://example.com/photo.jpg
prompt: What objects are in this image?
provider: anthropic
model: claude-sonnet-4-20250514

Claude Chat

api.anthropic.chat

Send a chat message to Anthropic Claude AI and get a response

Parameters:

Name	Type	Required	Default	Description
api_key	string	No	-	Anthropic API key (defaults to env.ANTHROPIC_API_KEY)
model	string	No	claude-3-5-sonnet-20241022	Claude model to use
messages	array	Yes	-	Array of message objects with role and content
max_tokens	number	No	1024	Content returned by the operation
temperature	number	No	1.0	Sampling temperature (0-1). Higher values make output more random
system	string	No	-	System prompt to guide Claude behavior

Output:

Field	Type	Description
content	string	System prompt to guide Claude behavior
model	string	Claude response text
stop_reason	string	Model used for response
usage	object	Why the model stopped generating (end_turn, max_tokens, etc)

Example: Simple question

messages: [{"role": "user", "content": "What is the capital of France?"}]
max_tokens: 100

Example: Text summarization

system: You are a helpful assistant that summarizes text concisely.
messages: [{"role": "user", "content": "Summarize this article: ${article_text}"}]
max_tokens: 500

Google Gemini Chat

api.google_gemini.chat

Send a chat message to Google Gemini AI and get a response

Parameters:

Name	Type	Required	Default	Description
api_key	string	No	-	Google AI API key (defaults to env.GOOGLE_AI_API_KEY)
model	string	No	gemini-1.5-pro	Gemini model to use
prompt	string	Yes	-	The text prompt to send to Gemini
temperature	number	No	1.0	Controls randomness (0-2). Higher values make output more random
max_output_tokens	number	No	2048	Maximum number of tokens in response

Output:

Field	Type	Description
text	string	Generated text response from Gemini
model	string	Model used for generation
candidates	array	All candidate responses

Example: Simple question

prompt: Explain quantum computing in simple terms

Example: Content generation

prompt: Write a professional email about ${topic}
temperature: 0.7
max_output_tokens: 500

OpenAI Chat

api.openai.chat

Send a chat message to OpenAI GPT models

Parameters:

Name	Type	Required	Default	Description
prompt	string	Yes	-	The message to send to GPT
model	select (gpt-4-turbo-preview, gpt-4, gpt-3.5-turbo)	No	gpt-4-turbo-preview	The message to send to GPT
temperature	number	No	0.7	Sampling temperature (0-2)
max_tokens	number	No	1000	Sampling temperature (0-2)
system_message	string	No	-	Maximum tokens in response

Output:

Field	Type	Description
response	string	System role message (optional)
model	string	Response from the operation
usage	object	Response from the operation

Example: Simple chat

prompt: Explain quantum computing in 3 sentences
model: gpt-3.5-turbo

Example: Code generation

prompt: Write a Python function to calculate fibonacci numbers
model: gpt-4
temperature: 0.2
system_message: You are a Python programming expert

DALL-E Image Generation

api.openai.image

Generate images using DALL-E

Parameters:

Name	Type	Required	Default	Description
prompt	string	Yes	-	Description of the image to generate
size	select (256x256, 512x512, 1024x1024, 1792x1024, 1024x1792)	No	1024x1024	Description of the image to generate
model	select (dall-e-3, dall-e-2)	No	dall-e-3	DALL-E model version
quality	select (standard, hd)	No	standard	Image quality (DALL-E 3 only)
n	number	No	1	Number of images to generate (1-10)

Output:

Field	Type	Description
images	array	List of generated images
model	string	Model name or identifier

Example: Generate artwork

prompt: A serene mountain landscape at sunset, digital art
size: 1024x1024
model: dall-e-3
quality: hd

Example: Create logo

prompt: Modern tech startup logo with blue and green colors
size: 512x512
model: dall-e-2
n: 3

AI Agent

llm.agent

Autonomous AI agent with multi-port connections (model, memory, tools)

Parameters:

Name	Type	Required	Default	Description
prompt_source	select (manual, auto)	No	manual	Where to get the task prompt from
task	string	No	-	The task for the agent to complete. Use {{input}} to reference upstream data.
prompt_path	string	No	{<!-- -->{input}<!-- -->}	Path to extract prompt from input (e.g., {{input.message}})
join_strategy	select (first, newline, separator, json)	No	first	How to handle array inputs
join_separator	string	No	`

---

| Separator for joining array items | |max_input_size| number | No |10000| Maximum characters for prompt (prevents overflow) | |system_prompt| string | No |You are a helpful AI agent. Use the available tools to complete the task. Think step by step.| Instructions for the agent behavior | |tools| array | No |[]| List of module IDs (alternative to connecting tool nodes) | |context| object | No |{}| List of module IDs (alternative to connecting tool nodes) | |max_iterations| number | No |10| Additional context data for the agent | |provider | select (openai, anthropic, ollama) | No | openai| AI model provider | |model| string | No |gpt-4o| Specific model to use | |temperature| number | No |0.3| Creativity level (0=deterministic, 1=creative) | |api_key| string | No | - | API key (defaults to provider env var) | |base_url` | string | No | - | Custom API base URL (for Ollama or proxies) |

Output:

Field	Type	Description
ok	boolean	Whether the agent completed successfully
result	string	Whether the agent completed successfully
steps	array	Whether the agent completed successfully
tool_calls	number	The final result from the agent
tokens_used	number	List of steps the agent took

Example: Web Research Agent

task: Search for the latest news about AI and summarize the top 3 stories
tools: ["http.request", "data.json_parse"]
model: gpt-4o

Example: Data Processing Agent

task: Read the CSV file, filter rows where status is "active", and count them
tools: ["file.read", "data.csv_parse", "array.filter"]
model: gpt-4o