LLM

Blocks for interacting with Large Language Models including AI conversations, image generation, video creation, and intelligent condition evaluation.

AI Ad Maker Video Creator

What it is

Creates an AI‑generated 30‑second advert (text + images)

How it works

This block generates video advertisements by combining AI-generated visuals with narrated scripts. Line breaks in the script create scene transitions. Choose from various voices and background music options.

Optionally provide your own images via input_media_urls, or let the AI generate visuals. The finished video is returned as a URL for download or embedding.

Inputs

Input

Description

Type

Required

script

Short advertising copy. Line breaks create new scenes.

str

Yes

ratio

Aspect ratio

str

target_duration

Desired length of the ad in seconds.

int

voice

Narration voice

background_music

Background track

input_media_urls

List of image URLs to feature in the advert.

List[str]

use_only_provided_media

Restrict visuals to supplied images only.

bool

Outputs

Output

Description

Type

error

Error message if the operation failed

str

video_url

URL of the finished advert

str

Possible use case

Product Marketing: Create quick promotional videos for products or services.

Social Media Ads: Generate short video ads for social media advertising campaigns.

Content Automation: Automatically create video ads from product descriptions or scripts.

AI Condition

What it is

Uses AI to evaluate natural language conditions and provide conditional outputs

How it works

This block uses an LLM to evaluate natural language conditions that can't be expressed with simple comparisons. Describe the condition in plain English, and the AI determines if it's true or false for the given input.

The result routes data to yes_output or no_output, enabling intelligent branching based on meaning, sentiment, or complex criteria.

Inputs

Input

Description

Type

Required

input_value

The input value to evaluate with the AI condition

Input Value

Yes

condition

A plaintext English description of the condition to evaluate

str

Yes

yes_value

(Optional) Value to output if the condition is true. If not provided, input_value will be used.

Yes Value

no_value

(Optional) Value to output if the condition is false. If not provided, input_value will be used.

No Value

model

The language model to use for evaluating the condition.

Outputs

Output

Description

Type

error

Error message if the AI evaluation is uncertain or fails

str

result

The result of the AI condition evaluation (True or False)

bool

yes_output

The output value if the condition is true

Yes Output

no_output

The output value if the condition is false

No Output

Possible use case

Sentiment Routing: Route messages differently based on whether they express frustration or satisfaction.

Content Moderation: Check if content contains inappropriate material or policy violations.

Intent Detection: Determine if a user message is a question, complaint, or request.

AI Conversation

What it is

A block that facilitates multi-turn conversations with a Large Language Model (LLM), maintaining context across message exchanges.

How it works

The block sends the entire conversation history to the chosen LLM, including system messages, user inputs, and previous responses. It then returns the LLM's response as the next part of the conversation.

Inputs

Input

Description

Type

Required

prompt

The prompt to send to the language model.

str

messages

List of messages in the conversation.

List[Any]

Yes

model

The language model to use for the conversation.

max_tokens

The maximum number of tokens to generate in the chat completion.

int

ollama_host

Ollama host for local models

str

Outputs

Output

Description

Type

error

Error message if the operation failed

str

response

The model's response to the conversation.

str

prompt

The prompt sent to the language model.

List[Any]

Possible use case

Creating an interactive chatbot that can maintain context over multiple exchanges, such as a customer service assistant or a language learning companion.

AI Image Customizer

What it is

Generate and edit custom images using Google's Nano-Banana models from Gemini. Provide a prompt and optional reference images to create or modify images.

How it works

This block uses Google's Gemini Nano-Banana models for image generation and editing. Provide a text prompt describing the desired image, and optionally include reference images for style guidance or modification.

Configure aspect ratio to match your needs and choose between JPG or PNG output format. The generated image is returned as a URL.

Inputs

Input

Description

Type

Required

prompt

A text description of the image you want to generate

str

Yes

model

The AI model to use for image generation and editing

"google/nano-banana" | "google/nano-banana-pro" | "google/nano-banana-2"

images

Optional list of input images to reference or modify

List[str (file)]

aspect_ratio

Aspect ratio of the generated image

"match_input_image" | "1:1" | "2:3" | "3:2" | "3:4" | "4:3" | "4:5" | "5:4" | "9:16" | "16:9" | "21:9"

output_format

Format of the output image

"jpg" | "png"

Outputs

Output

Description

Type

error

Error message if the operation failed

str

image_url

URL of the generated image

str (file)

Possible use case

Product Visualization: Generate product images with different backgrounds or settings.

Creative Content: Create unique images for marketing, social media, or presentations.

Image Modification: Edit existing images by providing them as references with modification prompts.

AI Image Editor

What it is

Edit images using Flux Kontext or Google Nano Banana models. Provide a prompt and optional reference image to generate a modified image.

How it works

This block uses BlackForest Labs' Flux Kontext or Google's Nano Banana models for context-aware image editing. Describe the desired edit in the prompt, and optionally provide an input image to modify.

Choose between Flux Kontext Pro, Max, or Nano Banana models for different quality/speed tradeoffs. Set a seed for reproducible results across multiple runs (Flux Kontext only).

Inputs

Input

Description

Type

Required

prompt

Text instruction describing the desired edit

str

Yes

input_image

Reference image URI (jpeg, png, gif, webp)

str (file)

aspect_ratio

Aspect ratio of the generated image

"match_input_image" | "1:1" | "16:9" | "9:16" | "4:3" | "3:4" | "3:2" | "2:3" | "4:5" | "5:4" | "21:9" | "9:21" | "2:1" | "1:2"

seed

Random seed. Set for reproducible generation (Flux Kontext only; ignored by Nano Banana models)

int

model

Model variant to use

"Flux Kontext Pro" | "Flux Kontext Max" | "Nano Banana Pro" | "Nano Banana 2"

Outputs

Output

Description

Type

error

Error message if the operation failed

str

output_image

URL of the transformed image

str (file)

Possible use case

Style Transfer: Transform images to match different artistic styles or moods.

Object Editing: Add, remove, or modify specific elements in existing images.

Background Changes: Replace or modify image backgrounds while preserving subjects.

AI Image Generator

What it is

Generate images using various AI models through a unified interface

How it works

This block generates images from text prompts using your choice of AI models including Flux and Recraft. Select the image size (square, landscape, portrait, wide, or tall) and visual style to match your needs.

The unified interface allows switching between models without changing your workflow, making it easy to compare results or adapt to different use cases.

Inputs

Input

Description

Type

Required

prompt

Text prompt for image generation

str

Yes

model

The AI model to use for image generation

size

Format of the generated image: - Square: Perfect for profile pictures, icons - Landscape: Traditional photo format - Portrait: Vertical photos, portraits - Wide: Cinematic format, desktop wallpapers - Tall: Mobile wallpapers, social media stories

"square" | "landscape" | "portrait" | "wide" | "tall"

style

Visual style for the generated image

Outputs

Output

Description

Type

error

Error message if the operation failed

str

image_url

URL of the generated image

str

Possible use case

Content Creation: Generate images for blog posts, articles, or social media.

Marketing Visuals: Create product images, banners, or promotional graphics.

Illustration: Generate custom illustrations for presentations or documents.

AI List Generator

What it is

A block that creates lists of items based on prompts using a Large Language Model (LLM), with optional source data for context.

How it works

The block formulates a prompt based on the given focus or source data, sends it to the chosen LLM, and then processes the response to ensure it's a valid Python list. It can retry multiple times if the initial attempts fail.

Inputs

Input

Description

Type

Required

focus

The focus of the list to generate.

str

source_data

The data to generate the list from.

str

model

The language model to use for generating the list.

max_retries

Maximum number of retries for generating a valid list.

int

force_json_output

Whether to force the LLM to produce a JSON-only response. This can increase the block's reliability, but may also reduce the quality of the response because it prohibits the LLM from reasoning before providing its JSON response.

bool

max_tokens

The maximum number of tokens to generate in the chat completion.

int

ollama_host

Ollama host for local models

str

Outputs

Output

Description

Type

error

Error message if the operation failed

str

generated_list

The generated list.

List[str]

list_item

Each individual item in the list.

str

prompt

The prompt sent to the language model.

List[Any]

Possible use case

Automatically generating a list of key points or action items from a long meeting transcript or summarizing the main topics discussed in a series of documents.

AI Music Generator

What it is

This block generates music using Meta's MusicGen model on Replicate.

How it works

This block uses Meta's MusicGen model to generate original music from text descriptions. Describe the desired music style, mood, and instruments in the prompt, and the AI creates a matching audio track.

Configure duration, temperature (for variety), and output format. Higher temperature produces more diverse results, while lower values stay closer to typical patterns.

Inputs

Input

Description

Type

Required

prompt

A description of the music you want to generate

str

Yes

music_gen_model_version

Model to use for generation

"stereo-large" | "melody-large" | "large"

duration

Duration of the generated audio in seconds

int

temperature

Controls the 'conservativeness' of the sampling process. Higher temperature means more diversity

float

top_k

Reduces sampling to the k most likely tokens

int

top_p

Reduces sampling to tokens with cumulative probability of p. When set to 0 (default), top_k sampling is used

float

classifier_free_guidance

Increases the influence of inputs on the output. Higher values produce lower-variance outputs that adhere more closely to inputs

int

output_format

Output format for generated audio

"wav" | "mp3"

normalization_strategy

Strategy for normalizing audio

"loudness" | "clip" | "peak" | "rms"

Outputs

Output

Description

Type

error

Error message if the operation failed

str

result

URL of the generated audio file

str

Possible use case

Video Soundtracks: Generate background music for videos, podcasts, or presentations.

Content Creation: Create original music for social media or marketing content.

Prototyping: Quickly generate music concepts for creative projects.

AI Screenshot To Video Ad

What it is

Turns a screenshot into an engaging, avatar‑narrated video advert.

How it works

This block creates video advertisements featuring a screenshot with AI-generated narration. Provide the screenshot URL and narration script, and the block generates a video with voice and background music.

Choose from various voices and music tracks. The video showcases the screenshot while the AI narrator reads your script.

Inputs

Input

Description

Type

Required

script

Narration that will accompany the screenshot.

str

Yes

screenshot_url

Screenshot or image URL to showcase.

str

Yes

ratio

str

target_duration

int

voice

background_music

Outputs

Output

Description

Type

error

Error message if the operation failed

str

video_url

Rendered video URL

str

Possible use case

App Demos: Create narrated demonstrations of software features from screenshots.

Product Tours: Turn product screenshots into engaging video walkthroughs.

Tutorial Videos: Generate instructional videos from step-by-step screenshots.

AI Shortform Video Creator

What it is

Creates a shortform video using revid.ai

How it works

This block creates short-form videos from scripts using revid.ai. Format scripts with line breaks for scene changes and use [brackets] to guide visual generation. Text outside brackets becomes narration.

Choose video style (stock video, moving images, or AI-generated), voice, background music, and generation presets. The finished video URL is returned for download or sharing.

Inputs

Input

Description

Type

Required

script

1. Use short and punctuated sentences 2. Use linebreaks to create a new clip 3. Text outside of brackets is spoken by the AI, and [text between brackets] will be used to guide the visual generation. For example, [close-up of a cat] will show a close-up of a cat.

str

Yes

ratio

Aspect ratio of the video

str

resolution

Resolution of the video

str

frame_rate

Frame rate of the video

int

generation_preset

Generation preset for visual style - only affects AI-generated visuals

background_music

Background music track

voice

AI voice to use for narration

video_style

Type of visual media to use for the video

"stockVideo" | "movingImage" | "aiVideo"

Outputs

Output

Description

Type

error

Error message if the operation failed

str

video_url

The URL of the created video

str

Possible use case

Social Media Content: Create TikTok, Reels, or Shorts content automatically.

Explainer Videos: Generate short educational or promotional videos.

Content Repurposing: Convert written content into engaging short-form video.

AI Structured Response Generator

What it is

A block that generates structured JSON responses using a Large Language Model (LLM), with schema validation and format enforcement.

How it works

The block sends the input prompt to a chosen LLM, along with any system prompts and expected response format. It then processes the LLM's response, ensuring it matches the expected format, and returns the structured data.

Inputs

Input

Description

Type

Required

prompt

The prompt to send to the language model.

str

Yes

expected_format

Expected format of the response. If provided, the response will be validated against this format. The keys should be the expected fields in the response, and the values should be the description of the field.

Dict[str, str]

Yes

list_result

Whether the response should be a list of objects in the expected format.

bool

model

The language model to use for answering the prompt.

force_json_output

bool

sys_prompt

The system prompt to provide additional context to the model.

str

conversation_history

The conversation history to provide context for the prompt.

List[Dict[str, Any]]

retry

Number of times to retry the LLM call if the response does not match the expected format.

int

prompt_values

Values used to fill in the prompt. The values can be used in the prompt by putting them in a double curly braces, e.g. {{variable_name}}.

Dict[str, str]

max_tokens

The maximum number of tokens to generate in the chat completion.

int

compress_prompt_to_fit

Whether to compress the prompt to fit within the model's context window.

bool

ollama_host

Ollama host for local models

str

Outputs

Output

Description

Type

error

Error message if the operation failed

str

response

The response object generated by the language model.

Dict[str, Any] | List[Dict[str, Any]]

prompt

The prompt sent to the language model.

List[Any]

Possible use case

Extracting specific information from unstructured text, such as generating a product description with predefined fields (name, features, price) from a lengthy product review.

AI Text Generator

What it is

A block that produces text responses using a Large Language Model (LLM) based on customizable prompts and system instructions.

How it works

The block sends the input prompt to a chosen LLM, processes the response, and returns the generated text.

Inputs

Input

Description

Type

Required

prompt

The prompt to send to the language model. You can use any of the {keys} from Prompt Values to fill in the prompt with values from the prompt values dictionary by putting them in curly braces.

str

Yes

model

The language model to use for answering the prompt.

sys_prompt

The system prompt to provide additional context to the model.

str

retry

Number of times to retry the LLM call if the response does not match the expected format.

int

prompt_values

Values used to fill in the prompt. The values can be used in the prompt by putting them in a double curly braces, e.g. {{variable_name}}.

Dict[str, str]

ollama_host

Ollama host for local models

str

max_tokens

The maximum number of tokens to generate in the chat completion.

int

Outputs

Output

Description

Type

error

Error message if the operation failed

str

response

The response generated by the language model.

str

prompt

The prompt sent to the language model.

List[Any]

Possible use case

Generating creative writing, such as short stories or poetry, based on a given theme or starting sentence.

AI Text Summarizer

What it is

A block that summarizes long texts using a Large Language Model (LLM), with configurable focus topics and summary styles.

How it works

The block splits the input text into smaller chunks, sends each chunk to an LLM for summarization, and then combines these summaries. If the combined summary is still too long, it repeats the process until a concise summary is achieved.

Inputs

Input

Description

Type

Required

text

The text to summarize.

str

Yes

model

The language model to use for summarizing the text.

focus

The topic to focus on in the summary

str

style

The style of the summary to generate.

"concise" | "detailed" | "bullet points" | "numbered list"

max_tokens

The maximum number of tokens to generate in the chat completion.

int

chunk_overlap

The number of overlapping tokens between chunks to maintain context.

int

ollama_host

Ollama host for local models

str

Outputs

Output

Description

Type

error

Error message if the operation failed

str

summary

The final summary of the text.

str

prompt

The prompt sent to the language model.

List[Any]

Possible use case

Summarizing lengthy research papers or articles to quickly grasp the main points and key findings.

Claude Code

What it is

Execute tasks using Claude Code in an E2B sandbox. Claude Code can create files, install tools, run commands, and perform complex coding tasks autonomously.

How it works

When activated, the block:

Creates or connects to an E2B sandbox (a secure, isolated Linux environment)
Installs the latest version of Claude Code in the sandbox
Optionally runs setup commands to prepare the environment
Executes your prompt using Claude Code, which can create/edit files, install dependencies, run terminal commands, and build applications
Extracts all files created/modified during execution (text files and binary files like images, PDFs, etc.)
Returns the response and files, optionally keeping the sandbox alive for follow-up tasks

The block supports conversation continuation through three mechanisms:

Same sandbox continuation (via session_id + sandbox_id): Resume on the same live sandbox
Fresh sandbox continuation (via conversation_history): Restore context on a new sandbox if the previous one timed out
Dispose control (dispose_sandbox flag): Keep sandbox alive for multi-turn conversations

Inputs

Input

Description

Type

Required

prompt

The task or instruction for Claude Code to execute. Claude Code can create files, install packages, run commands, and perform complex coding tasks.

str

timeout

Sandbox timeout in seconds. Claude Code tasks can take a while, so set this appropriately for your task complexity. Note: This only applies when creating a new sandbox. When reconnecting to an existing sandbox via sandbox_id, the original timeout is retained.

int

setup_commands

Optional shell commands to run before executing Claude Code. Useful for installing dependencies or setting up the environment.

List[str]

working_directory

Working directory for Claude Code to operate in.

str

session_id

Session ID to resume a previous conversation. Leave empty for a new conversation. Use the session_id from a previous run to continue that conversation.

str

sandbox_id

Sandbox ID to reconnect to an existing sandbox. Required when resuming a session (along with session_id). Use the sandbox_id from a previous run where dispose_sandbox was False.

str

conversation_history

Previous conversation history to continue from. Use this to restore context on a fresh sandbox if the previous one timed out. Pass the conversation_history output from a previous run.

str

dispose_sandbox

Whether to dispose of the sandbox immediately after execution. Set to False if you want to continue the conversation later (you'll need both sandbox_id and session_id from the output).

bool

Outputs

Output

Description

Type

error

Error message if execution failed

str

response

The output/response from Claude Code execution

str

files

List of files created/modified by Claude Code during this execution. Includes text files and binary files (images, PDFs, etc.). Each file has 'path', 'relative_path', 'name', 'content', and 'workspace_ref' fields. workspace_ref contains a workspace:// URI for workspace storage. For binary files, content contains a placeholder; use workspace_ref to access the file.

List[SandboxFileOutput]

conversation_history

Full conversation history including this turn. Pass this to conversation_history input to continue on a fresh sandbox if the previous sandbox timed out.

str

session_id

Session ID for this conversation. Pass this back along with sandbox_id to continue the conversation.

str

sandbox_id

ID of the sandbox instance. Pass this back along with session_id to continue the conversation. This is None if dispose_sandbox was True (sandbox was disposed).

str

Possible use case

API Documentation to Full Application: A product team wants to quickly prototype applications based on API documentation. They fetch API docs with Firecrawl, pass them to Claude Code with a prompt like "Create a web app that demonstrates all the key features of this API", and Claude Code builds a complete application with HTML/CSS/JS frontend, proper error handling, and example API calls. The Files output can then be pushed to GitHub.

Multi-turn Development: A developer uses Claude Code to scaffold a new project iteratively - Turn 1: "Create a Python FastAPI project with user authentication" (dispose_sandbox=false), Turn 2: Uses the returned session_id + sandbox_id to ask "Add rate limiting middleware", Turn 3: Continues with "Add comprehensive tests". Each turn builds on the previous work in the same sandbox environment.

Automated Code Review and Fixes: An agent receives code from a PR, sends it to Claude Code with "Review this code for bugs and security issues, then fix any problems you find", and Claude Code analyzes the code, makes fixes, and returns the corrected files ready to commit.

Code Generation

What it is

Generate or refactor code using an OpenAI API key or a connected ChatGPT plan through Codex App Server.

How it works

Choose openai_api to call the selected Codex model through the OpenAI Responses API with an OpenAI API key. This path applies model, max_output_tokens (up to 128,000), and every reasoning effort except ultra; none omits the reasoning configuration. It rejects non-OpenAI API-key credentials and unsupported ultra requests before making an API call.

Choose codex_app_server to use a connected ChatGPT plan. This path requires matching Codex OAuth credentials and an active credential lease, selects the subscription model from the live App Server catalog, and uses the plan's output-token limit. The model and max_output_tokens inputs therefore do not affect App Server requests, while all listed reasoning efforts are forwarded through the App Server mapping. Missing or mismatched credentials fail validation, and transport failures are returned through the block's standard error output. Successful responses include the generated text, any reasoning summary, and the transport response ID; usage may be absent without failing the response.

Inputs

Input

Description

Type

Required

prompt

Primary coding request passed to the Codex model.

str

Yes

system_prompt

Optional instructions passed to the selected Codex transport.

str

transport

Use an OpenAI API key or your connected ChatGPT plan through Codex App Server.

"openai_api" | "codex_app_server"

model