API explorer

Converts images to favicon.ico format using ImageMagick. Accepts a base64-encoded image (PNG, JPG, SVG, etc.) and returns a multi-resolution favicon.ico file as base64. Generates sizes: 64x64, 48x48, 32x32, 24x24, 16x16.

A text completion model (Codestral 2405) that fills in the missing part of the prompt. Mostly used for code completion.

Generates an image from a text prompt using Z.AI with Groq content moderation. Automatically polls for completion (up to 5 seconds), downloads the image, and saves it to sources with binary content for deduplication. Returns the source ID and image URL.

Generates an image from a text prompt using Z.AI with Groq content moderation and returns the URL directly (without saving to sources). Faster but no deduplication.

Executes an AI agent with custom system prompt and user input for complex reasoning tasks. Use for: decision making, content analysis, data interpretation, creative writing, problem-solving, summarization with specific requirements. Don't use for: simple text transformation (use dedicated nodes), or when deterministic rules suffice. Example: system="You are a financial analyst" + user="Analyze this stock data and provide investment recommendation" returns detailed analysis with reasoning.

Performs inference tasks using a high performance vision model (Qwen 3.6 27B). Returns the vision model's response regarding the image. Use for image analysis, OCR, object detection, and visual understanding tasks. Features comprehensive error handling and detailed metadata.

Search the web for multiple queries in parallel using Exa + Perplexity, then summarize results with an AI subagent. Each query gets parallel search + AI summarization. Max 10 queries per call.

Extracts text from images using Mistral OCR, with image enhancement. Returns the extracted text content as a string. Supports various image formats and provides detailed processing metadata.

Utilize perplexity web search for fast and accurate information retrieval

Creates a new memory with optional vector embedding, keywords, and rich metadata. Returns the created memory with detailed information.

Soft deletes a memory (marks as deleted but keeps in database for audit purposes). Returns detailed deletion status.

Retrieves a specific memory by ID with full details including keywords and embedding status. Returns the memory if found and accessible.

Retrieves the full knowledge graph structure for a memory, including all connected entities and their inter-relations.

Walks the knowledge graph from a starting entity to find related memories through entity relationships. Enables contextual recall by traversing the entity-memory graph.

Lists all memories for a subscriber with advanced filtering, sorting, and pagination options. Returns paginated results with summary statistics.

Creates a new memory and links it to one or more entities in one operation. Enables knowledge graph construction by associating memories with entities.

Searches for memories using various methods (vector similarity, keywords, hybrid). Returns matching memories with relevance scores and search metadata.

Updates an existing memory with new content, metadata, or keywords. Returns the updated memory with change tracking.

Creates a public news article or blog post with structured metadata. Use for: publishing news content, creating knowledge base articles, storing research findings, building content libraries. Don't use for: private user content (use Source Create), or temporary data (use Memory). Articles are searchable by content, topics, and title. Example: Create articles about industry trends for a news aggregator or research summaries for a knowledge base.

Automatically fetches webpage content, extracts the main article text, and creates a structured article. Use for: converting web content to structured articles, building content libraries from web sources, archiving articles for research. Auto-extracts title/content, optionally creates AI summary and topics. Don't use for: PDFs or documents (use Document processing), or when you already have the content (use Article Create). Fails on sites with anti-scraping. Example: Convert news articles or blog posts into searchable, structured content.

Fetches a single article by ID. Returns the complete article with all metadata including content, topics, and timestamps.

Gets recent articles from the last N days. Useful for news aggregation and content curation.

Finds articles relevant to a given topic or query by searching through recent articles. Uses content matching across title, content, summary, and topics. Useful for finding related news or building topic-specific content feeds.

Searches articles by title, content, or summary. Supports pagination and returns matching articles with relevance scoring.

Uses AI to create a summary of an article. Automatically updates the article with the new summary. Supports different summary lengths: short (2-3 sentences), medium (1-2 paragraphs), or long (3-4 paragraphs).

Uses AI to extract relevant topics from an article. Automatically updates the article with extracted topics. Returns 3-7 relevant topics that reflect the main themes of the article.

Converts documents between formats using pandoc. Supports various output formats like PDF, HTML, DOCX, etc. Input is a base64 encoded string of the document, with a data:{doctype};base64 prefix, and returns the converted file as a base64 encoded string.

Generate Digest From Base64

Generate Digest From Buffer

Generate Digest From String

Extracts text from PDF documents using Mistral OCR. Returns structured text content with page breakdown, markdown formatting for tables, and optional image extraction. Supports deduplication via source digest and can auto-embed content for semantic search.

Searches the web for information using Exa.

Extract content from an HTML string using a css selector. Returns the matched content as a string.

Convert HTML to markdown. Returns a markdown string.

Control the user's real browser via Kimi WebBridge daemon. Navigate, snapshot, click, fill, evaluate, and more using the user's actual login sessions.

Downloads raw HTML from a webpage and saves to cloud storage. Use for: web scraping, HTML parsing, archiving content, extracting specific elements with CSS selectors. Don't use for: getting summarized content (use Web Search instead), or when you only need readable text. Returns file_id that can be used with Extract HTML Selector or HTML to Markdown converter. Example: fetch product pages for price monitoring or news articles for content extraction.

Performs AI-powered web search with browser automation and returns human-readable, markdown-formatted results with citations. Use for: research summaries, current events, fact-checking, getting overview of topics. Don't use for: bulk data collection, raw HTML scraping, or when you need specific page content. Example: "latest AI breakthroughs in 2024" returns summarized findings with source links.

Convert a file to CSV format using ssconvert

Converts addresses into latitude/longitude coordinates with structured location details using OpenCage API.

Extract JSON from a string. Useful for getting JSON out of agent outputs that may contain other text. Returns an array of the found JSON objects; includes inner JSON objects as well. Use the keys parameter to filter the results to only include objects that have the specified keys.

Generates professional diagrams and flowcharts from Mermaid markup code. Use for: creating workflow diagrams, system architecture charts, process flow maps, organizational charts, data flow diagrams. Returns base64-encoded SVG image suitable for embedding in reports or web pages. Don't use for: complex technical drawings (use specialized tools), or when you need editable vector graphics. Supports flowcharts, sequence diagrams, Gantt charts, class diagrams. Example: Generate a flowchart showing "graph TD; A[Start] --> B[Process]; B --> C[End]" for documentation.

Runs probabilistic simulations to model uncertainty and risk in financial or business scenarios. Use for: investment risk analysis, portfolio modeling, business forecasting, option pricing, uncertainty quantification. Supports geometric Brownian motion for stock prices and custom formulas. Returns percentiles (P10, P25, P50, P75, P90) and downside probability. Don't use for: deterministic calculations (use simple math), or non-probabilistic scenarios. Example: Model stock price ranges over 12 months with 5% expected return and 20% volatility to estimate downside risk.

Transform row-based or visual data into tabular format. Converts data where related values are spread across rows into a proper table structure. Supports CSV input/output, automatic header detection, and flexible column grouping. Use for: converting invoice line items, transforming survey responses, normalizing log entries, consolidating multi-row records.

Query a CSV file using DuckDB. You can use this on tempfiles, string output, or sources.

Converts spreadsheets between formats using gnumeric's ssconvert command. Supports Excel, OpenOffice, LaTeX, HTML, PDF, CSV, etc. Accepts a base64 encoded string of the spreadsheet, and returns the converted file as a base64 encoded string.

Supabase Query CountRows

Supabase Query InsertRow

Supabase Query SelectAll

Supabase Query SelectRows

Supabase Query UpdateRow

Chunk text for RAG applications. Supports fixed-size, sentence, paragraph, and semantic chunking strategies. Optional overlap creates context continuity between chunks.

Create a new saved script for later execution. Supports multiple programming languages with validation. Scripts can be public or private, tagged for organization, and include metadata. Use for: saving reusable code, creating automation scripts, storing utility functions, building script libraries.

Soft-deletes a saved script (marks as deleted without removing from database). Only the script owner can delete their scripts. Deleted scripts are not returned in searches or lists. Use for: cleaning up old scripts, removing test code, managing script library.

Execute a saved script with secure sandboxing. Supports JavaScript, Python, TypeScript, and Bash. Provides execution logs, timing, and memory usage. Scripts run in isolated context with limited access to system resources. Use for: automating tasks, data processing, calculations, API integrations. Scripts can access Math, Date, JSON, and basic console logging.

Lists all saved scripts for the current subscriber with pagination and filtering options. Supports filtering by language, visibility, and tags. Use for: browsing script library, managing code collection, finding reusable scripts.

Match a regular expression pattern against text. Returns all matches including capture groups. Use for: extracting structured data from text, parsing log files, validating input formats, extracting emails/URLs/dates, tokenizing text. Supports global matching (find all), case-insensitive mode, and multiline mode. Returns an array of match objects with full match and capture groups.

Search and replace text using regular expressions. Supports capture group references ($1, $2, etc.), named group references ($<name>), and special replacement patterns ($& for matched text, $` for before, $' for after). Use for: text cleanup, data normalization, redacting sensitive information, formatting transformations, template substitution, removing unwanted content.

Updates an existing saved script. Only the script owner can update their scripts. Supports partial updates - only provide the fields you want to change. Use for: fixing bugs, improving code, updating descriptions, managing script metadata.

Fetches real-time stock quotes and comprehensive financial metrics from FinViz for multiple ticker symbols. Use for: portfolio analysis, stock screening, financial research, investment decisions. Returns 60+ financial metrics including price, market cap, P/E ratio, volume, margins, growth rates, and performance data. Don't use for: historical data analysis (limited to current data), or non-US stocks. Example: ["AAPL", "GOOGL", "MSFT"] returns current price, financial ratios, analyst estimates, and performance metrics for each stock.

Extract financial data from SEC inline XBRL filings. Parses XBRL documents to extract financial metrics, context information, and facts.

Fetch SEC filing data by ticker or direct URL. Gets filing metadata and optionally fetches the full content for XBRL extraction.

Claim a pending job for execution by an edge node. Called by the Rainfall DevKit daemon to pull available work from the queue. Returns job details if a matching job is available, or null if no work is pending.

Get the current status of a distributed job. Returns job details including status (pending/running/completed/failed), result (if completed), and error (if failed).

Advanced human-in-the-loop with dynamic schema generation. Combines UserInput schema building with workflow suspension. Use for: complex approval workflows, dynamic form generation, conditional user input, schema-driven human review. Generates frontend-compatible forms and manages workflow suspension with timeout handling.

Queue a tool execution job for distributed processing across edge nodes. The job will be picked up by a capable edge node and executed asynchronously.

Schedule a workflow to run at specific times or intervals. Use for: recurring tasks (daily reports, weekly summaries), periodic data syncs, scheduled maintenance, time-based workflows. Returns: schedule ID, next run time, status. Don't use for: delays within a workflow (use WorkflowTrigger node), executing workflows immediately (use ExecuteWorkflow node). Supports cron expressions ("0 9 * * 1-5" for weekdays at 9am), interval-based scheduling ("every 86400 seconds" for daily), and one-time scheduled runs.

Basic approval workflow with yes/no decision. Suspends workflow and waits for human approval. Use for: simple review processes, binary decisions, approval checkpoints. Provides approve/reject options with optional comments.

Submit the result of a distributed job execution. Called by the Rainfall DevKit daemon after completing a job to report success or failure.

Fetch a file from cloud storage. Pass in the file id. Returns the file as a base64 encoded string.

Writes a base64-encoded parquet file to a temporary file and returns the path. Use this for parquet file operations.

Download a subscriber stored file from cloud storage. Pass in the subscriber id and the file id. Returns the file as a base64 encoded string.

Create a temporary file. Pass in the file name and the file content. Returns the file path. Use this to pass a binary object between nodes.

Execute another workflow, optionally waiting for completion. Use for: workflow modularity, reusing common workflows, creating workflow pipelines, parallel processing. Returns: execution ID, status, and workflow result. Don't use for: scheduling workflows (use Schedule node), adding delays (use WorkflowTrigger node). Supports synchronous execution (wait for completion) and asynchronous execution (fire-and-forget).

Generate Policy

Policy Generator

Creates a new named source (text content generated by or on behalf of a user). Returns the new source. This is saved in the subscriber's private collection of content. Ensure you have a subscriber ID for this (for public content, use Articles instead). Use for: storing user-generated content, saving private notes, building personal knowledge bases, storing workflow results. Supports namespacing for organizing sources into logical groups.

Soft-deletes a source (marks as deleted without removing from database). Only the source owner can delete their sources. Deleted sources are not returned in searches or lists.

Fetches a source (text content generated by or on behalf of a user). Returns the complete source with all metadata. Requires subscriber authentication.

Lists all sources for the current subscriber with pagination. Supports filtering by namespace. Useful for: browsing saved content, content management, finding sources without searching.

Searches for sources (text content generated by or on behalf of a user). Searches across both name and content fields. Returns a list of matching sources with content previews. Use for: finding saved notes, searching through personal content, content discovery. Supports filtering by namespace.

Updates an existing source with new content or metadata. Only the source owner can update their sources. Useful for: editing saved content, updating notes, revising workflow results.

Tools For Subscriber

Collects structured input from users with customizable forms. Supports workflow suspension for human-in-the-loop processes. Use for: gathering user preferences, collecting form data, approval workflows, configuration wizards, human review checkpoints. Can either generate dynamic schemas for immediate use or suspend workflows for later input. Provides frontend-compatible JSON Schema for UI rendering.

Generates a workflow from a user prompt using available proc nodes. Creates a JSON workflow definition with nodes, conditions, and state management.

Optimizes workflow traversal paths and dependencies using MILP

Pause workflow execution until a condition is met, time elapses, or external event occurs. Use for: adding delays between steps, waiting for external approvals, polling for state changes, creating continuation points for later resumption. Returns: trigger status, triggered timestamp, and any result data. Don't use for: scheduling workflows (use Schedule node), executing sub-workflows (use ExecuteWorkflow node). Supports delays, condition polling with configurable intervals and timeouts, and continuation IDs for external resumption.

Stripe Billing Meters Create Event

Stripe Checkout Create Session

Stripe Checkout Retrieve Session

Stripe Customers Create

Stripe Customers List Payment Methods

Stripe Customers Retrieve

Stripe Customers Update

Stripe Invoices Create

Stripe Invoices Finalize

Stripe Invoices List

Stripe Invoices Pay

Stripe Invoices Retrieve

Stripe Payment Intents Cancel

Stripe Payment Intents Confirm

Stripe Payment Intents Create

Stripe Payment Intents Retrieve

Stripe Payment Methods Attach

Stripe Payment Methods Detach

Stripe Payment Methods Retrieve

Stripe Prices Create

Stripe Prices List

Stripe Prices Retrieve

Stripe Prices Update

Stripe Products Create

Stripe Products List

Stripe Products Retrieve

Stripe Products Update

Stripe Refunds Create

Stripe Refunds List

Stripe Refunds Retrieve

Stripe Subscriptions Cancel

Stripe Subscriptions Create

Stripe Subscriptions List

Stripe Subscriptions Retrieve

Stripe Subscriptions Update

Idempotency store for Stripe webhook events (olympic.stripe_webhook_events): status (dedup check by Stripe event id), record (insert-on-conflict-ignore), markProcessed (update processed flag + result). Fail-open on DB errors — idempotency is a guard against duplicate delivery, never a gate on real payments.

Send email notification using a template

Send a heartbeat from an edge node to maintain its active registration. Called periodically by the Rainfall DevKit daemon. Updates last seen timestamp, refreshes TTL, and reports current load/queue status.

Register a new edge compute node with the Rainfall backend. Called by the Rainfall DevKit daemon on startup to announce its presence, capabilities, and network endpoints. Returns a unique edgeNodeId for subsequent operations.

Register a node schema in the central Olympic registry. Used by edge-exposed functions to make their schemas discoverable via /params and node_list.

Register proc nodes for edge execution. Associates specific proc node capabilities with an existing edge node, enabling cross-node job routing.

Unregister an edge node from the Rainfall backend. Called by the Rainfall DevKit daemon on graceful shutdown to clean up resources and release any claimed jobs.

Perform fast cosine similarity search on embedded chunks using DuckDB in-memory.

A database node. Returns a result from the database.

Schedule an email notification for future delivery

Resolves scheduling conflicts using Mixed Integer Linear Programming (MILP) with HiGHS solver (high-performance WASM-based optimizer). Handles people, rooms, and resources with time slots. Optimizes for priority, conflict minimization, or load balancing.

[DEPRECATED - Use Jina Image Embedding instead] Generates CLIP embeddings for images. This node is redundant with Jina Image Embedding and will be removed. Use Jina Image Embedding for all image embedding needs.

Classifies documents into predefined categories using Jina AI. Returns classification scores for each document-label pair.

Converts text documents into numerical vectors (embeddings) for semantic search and similarity comparison. Use for: building search systems, document clustering, content recommendation, finding similar articles. Don't use for: PDF files (use Jina PDF Embedding), images (use Jina Image Embedding), or short queries (use Jina Query Embedding). Each document becomes a 768-dimensional vector that can be compared using cosine similarity. Example: ["Annual report 2023", "Q3 earnings call"] creates embeddings for similarity search.

Reorders documents by relevance to a search query using AI ranking. Use for: improving search results, filtering relevant content, prioritizing documents by query match. Returns relevance scores (higher = more relevant) that you use to reorder your original document list. Don't use for: initial document retrieval (use similarity search), or when you need the documents returned. Example: query="machine learning" + documents=["AI article", "cooking recipe"] returns scores [0.95, 0.12] to reorder by relevance.

Converts images into numerical vectors for visual similarity search and image retrieval. Use for: finding similar images, visual content moderation, image clustering, building visual search engines. Don't use for: text documents (use Jina Document Embedding), PDFs (use Jina PDF Embedding), or when you need text from images (use OCR Text Extraction). Images must be base64-encoded without data prefix. Creates 768-dimensional vectors for cosine similarity comparison. Example: product images for visual recommendation system.

Generates embeddings for PDF documents using Jina AI. Returns array of embedding vectors. Each PDF must be base64-encoded WITHOUT the data: prefix (e.g., "JVBERi0xLjQK...").

Generates embeddings for search queries using Jina AI. Returns array of embedding vectors.

Segments long text documents into smaller, semantically coherent chunks. Useful for processing large documents that exceed token limits.

Slack Core AddReaction

Slack Core ListChannels

Slack Core ListMessages

Slack Core ListUsers

Slack Core PostMessage

GitHub integration using Octokit SDK for repository and issue management

Google Docs Batch Update

Google Docs Create Document

Google Docs Get Document

Google Drive Get File

Google Drive List Files

Google Oauth Authorize

Google Oauth Refresh

Google Oauth Token

Google Sheets Append Values

Google Sheets Batch Update

Google Sheets Clear Values

Google Sheets Create Sheet

Google Sheets Get Spreadsheet

Google Sheets Get Values

Google Sheets Update Values

Figma Auth Oauth

Figma Comments DeleteComment

Figma Comments GetFileComments

Figma Comments PostComment

Figma Components GetComponentSet

Figma Components GetFileComponents

Figma Files GetFile

Figma Files GetFileImage

Figma Files GetFileNodes

Figma Projects GetProjectFiles

Figma Projects GetTeamProjects

Figma Styles GetFileStyles

Figma Teams GetTeamMembers

Figma Users GetMe

Figma Versions GetFileVersions

Notion Auth CreateToken

Notion Auth IntrospectToken

Notion Auth RefreshToken

Notion Auth RevokeToken

Notion Blocks AppendChildren

Notion Blocks Delete

Notion Blocks Retrieve

Notion Blocks RetrieveChildren

Notion Blocks Update

Notion Databases Create

Notion Databases Query

Notion Databases Retrieve

Notion Databases Update

Notion Pages Create

Notion Pages Retrieve

Notion Pages RetrieveProperty

Notion Pages Update

Notion Search Search

Notion Users List

Notion Users Me

Notion Users Retrieve

Linear Core Issue

Linear Core IssueArchive

Linear Core IssueCreate

Linear Core IssueUpdate

Linear Core Issues

Linear Core Teams

Linear Core Viewer

Xai Chat Completions

Xai Images Generations

Xai Models List

Xai Models Retrieve

Analyzes an image for blur using Laplacian variance (native @napi-rs/canvas). Runs the crash-prone canvas decode inside an isolated child process so a poison image cannot kill the parent worker. Returns sharpness percentage (0-100, Michaelis-Menten normalized), raw variance score, and pass/fail. Use for: jewelry photo quality checks, image vetting, blur pre-filtering before AI analysis.

Supabase Storage CreateSignedUrl

Supabase Storage GetPublicUrl

Supabase Storage ListObjects

Image zoom/crop/upscale for detail inspection. Divides an image into a grid, crops specified cells at multiple magnification levels, and upscales each crop to 500x500px for detailed examination. Ideal for jewelry appraisal and artifact inspection.