Skip to main content

Eden AI integration

Run OCR, document parsing, image and speech AI, web research, and 500+ language models for your clients from a single Eden AI key.

What it does

Eden AI is one API in front of 50+ AI providers. The integration gives your workflows a single connection that reaches all of it: parse a client's invoices and resumes, read scanned documents, generate and edit images, transcribe calls, translate documents while keeping their layout, research the open web, and call any of the 500+ language models Eden AI routes to. You pick the provider per action from a live dropdown, and you can name up to three fallbacks so one provider's outage does not stop the workflow.

Because the billing, the model catalogue, and the spend controls all sit in the client's own Eden AI account, you can run the same workflow for every client and let each one pay for their own usage.

Connect an Eden AI account

  1. In TaskJuice, open the workflow you are building and add an Eden AI node.
  2. On the node, open the connection dropdown and choose Add connection.
  3. In a new tab, sign in to the Eden AI dashboard as the client, or as your agency if the client has delegated key creation to you.
  4. Go to Settings, then API Keys, and copy the key.
  5. Paste the key into TaskJuice and save the connection.

Eden AI keys are unscoped, so one key reaches every action in this integration. To rotate later, create a new key on the same page, update the TaskJuice connection, then revoke the old one.

Test without spending credits

Eden AI issues two kinds of key. An api_token makes real provider calls and bills the account. A sandbox_api_token validates the request for real and answers at no cost, but the result is a mock. Both are created on the same API Keys page. Connect a sandbox key while you are wiring a workflow up, then swap the connection to a production key when you go live.

Treat the sandbox result as a shape, not an answer. It confirms your field names and the response envelope, and it will not tell you whether the model did anything useful. It can also differ from the real response in detail: as of September 2026 the sandbox answers eden-ai/map-site with a list of bare URL strings, where the production API returns { url, title, description } objects. Test a branch that reads deep into an output against a production key before you ship it.

Triggers

Eden AI does not publish a webhook subscription API, so this integration is action-only. Start the workflow from a Schedule trigger, from another app's trigger, or from an inbound webhook, and call Eden AI inside it.

The long-running actions marked "as a job" below accept a Webhook Receiver URL, which Eden AI posts the finished result to. That is a callback for one specific call, not an account-level event stream, so it does not appear here as a trigger.

Actions

Every action takes a Model field backed by a live dropdown of exactly the providers Eden AI serves for that feature right now. There is no default, which is deliberate: models are retired continuously, and a hardcoded default would eventually point at one that no longer exists.

Text

  • eden-ai/detect-ai-text Detect AI-Generated Text: score whether a passage was written by an LLM.
  • eden-ai/moderate-text Moderate Text: flag hateful, sexual, violent or self-harm content in text.
  • eden-ai/check-spelling Check Spelling and Grammar: find spelling and grammar mistakes, with corrections.
  • eden-ai/extract-topics Extract Topics: pull the recurring themes out of a body of text.
  • eden-ai/extract-entities Extract Named Entities: identify people, organisations, places and dates in text.
  • eden-ai/detect-plagiarism Detect Plagiarism: check a passage against published sources.

Documents and OCR

  • eden-ai/ocr Extract Text from Document: read the text off a single image or page.
  • eden-ai/start-multipage-ocr Start Multipage OCR: read a whole multi-page PDF, as a job.
  • eden-ai/start-table-extraction Start Table Extraction: pull tables out of a document as structured rows, as a job.
  • eden-ai/parse-identity-document Parse ID Document: extract the fields from a passport, ID card or licence.
  • eden-ai/parse-invoice Parse Invoice or Receipt: extract line items, totals and party details from an invoice or receipt.
  • eden-ai/parse-resume Parse Resume: turn a CV into structured name, jobs, education and skills.

Images

  • eden-ai/generate-image Generate Image: generate an image from a prompt through any expert provider.
  • eden-ai/generate-image-openai Generate Image (OpenAI-compatible): generate an image through the OpenAI-compatible endpoint.
  • eden-ai/edit-image Edit Image: edit existing images against a prompt, optionally through a mask.
  • eden-ai/remove-image-background Remove Image Background: cut the subject out of a photo.
  • eden-ai/anonymize-image Anonymize Image: blur or remove personal detail from an image.
  • eden-ai/detect-ai-image Detect AI-Generated Image: score whether an image was generated by AI.
  • eden-ai/detect-image-deepfake Detect Image Deepfake: detect a face swap in a real photo.
  • eden-ai/detect-objects Detect Objects in Image: locate objects in an image with bounding boxes.
  • eden-ai/detect-faces Detect Faces in Image: locate faces and their attributes in an image.
  • eden-ai/detect-logos Detect Logos in Image: recognise brand logos in an image.
  • eden-ai/detect-explicit-content Detect Explicit Content: score an image for adult or graphic content.
  • eden-ai/compare-faces Compare Two Faces: decide whether two images show the same person.
  • eden-ai/recognize-face Recognize Face in Collection: match a face against an enrolled collection.

Face collections

  • eden-ai/create-face-collection Create Face Collection: create a collection faces can be enrolled into.
  • eden-ai/list-face-collections List Face Collections: list the collections on the account.
  • eden-ai/get-face-collection Get Face Collection: read one collection.
  • eden-ai/delete-face-collection Delete Face Collection: delete a collection and every face in it.
  • eden-ai/add-face-to-collection Add Face to Collection: enrol a face image into a collection.
  • eden-ai/list-faces-in-collection List Faces in Collection: list the faces enrolled in a collection.
  • eden-ai/delete-face-from-collection Delete Face from Collection: remove one enrolled face.

Translation

  • eden-ai/translate Translate Text: translate text between languages.
  • eden-ai/translate-document Translate Document: translate a whole document while keeping its layout.

Audio and video

  • eden-ai/synthesize-speech Convert Text to Speech: turn text into speech through any expert provider.
  • eden-ai/generate-speech-openai Generate Speech (OpenAI-compatible): turn text into speech through the OpenAI-compatible endpoint.
  • eden-ai/start-speech-to-text Start Speech-to-Text: transcribe audio with optional speaker diarization, as a job.
  • eden-ai/transcribe-audio-openai Transcribe Audio (OpenAI-compatible): transcribe audio synchronously.
  • eden-ai/start-video-generation Start Video Generation: generate video from text or an image, as a job.
  • eden-ai/start-video-deepfake-detection Start Video Deepfake Detection: score a video for synthetic manipulation, as a job.

Web

  • eden-ai/scrape-page Scrape Web Page: fetch one URL and return clean markdown, text or HTML.
  • eden-ai/map-site Map Site URLs: list the URLs on a site.
  • eden-ai/search-web Search the Web: search the web and return ranked results.
  • eden-ai/start-web-crawl Start Web Crawl: crawl a site from a starting URL, as a job.
  • eden-ai/start-batch-scrape Start Batch Web Scrape: scrape many URLs in one request, as a job.
  • eden-ai/start-structured-extraction Start Structured Extraction: extract structured data from pages against a schema, as a job.
  • eden-ai/start-deep-research Start Deep Research: run multi-step research and return a sourced answer, as a job.

Language models

  • eden-ai/chat Generate Chat Completion: generate a chat completion against any of the models Eden AI routes to.
  • eden-ai/embeddings Create Embeddings: turn text into vectors for search, RAG or clustering.
  • eden-ai/moderate-content-openai Moderate Content (OpenAI-compatible): classify text or images through the OpenAI-compatible endpoint.
  • eden-ai/create-response Create Response: create a stored response so the next turn only needs its id.
  • eden-ai/get-response Retrieve Response: read back a stored response.
  • eden-ai/delete-response Delete Response: delete a stored response.
  • eden-ai/create-anthropic-message Create Anthropic Message: call any model using Anthropic's native request format.
  • eden-ai/count-anthropic-tokens Count Anthropic Tokens: count a request's input tokens without running the model.

Async jobs

  • eden-ai/list-async-jobs List Async Jobs: list the async jobs on the account.
  • eden-ai/get-async-job Get Async Job: read an async job's status, and its output once it finishes.
  • eden-ai/delete-async-job Delete Async Job: delete an async job and its stored result.

Files

  • eden-ai/upload-file Upload File: upload a file and get an id any action can reference.
  • eden-ai/list-files List Files: list the uploaded files on the account.
  • eden-ai/delete-files Delete Files by ID: delete specific uploaded files by id.
  • eden-ai/delete-all-files Delete All Files: delete every uploaded file on the account.

Account and discovery

  • eden-ai/list-subfeatures List Subfeatures: list what Eden AI serves under one feature.
  • eden-ai/get-feature-info Get Feature Info: read a feature's live schema and its provider models with pricing.
  • eden-ai/monitor-consumption Monitor Consumption: read spend over a date range.
  • eden-ai/get-credits Get Credit Balance: read the remaining credit balance.
  • eden-ai/inspect-api-key Inspect API Key: read the connected key's label, budget and balance.

Working with files

Most document, image, audio and video actions take a File field that accepts either an uploaded file id or a direct URL. Use eden-ai/upload-file to push a file from an upstream node into Eden AI storage, then pick it from the File dropdown in every downstream action. Set Expires In Days on the upload so the client's storage cleans itself up.

eden-ai/edit-image is the exception. Give it an image_url, either an https URL or a base64 data URL, not a file_id. Eden AI accepts both in the request, but as of September 2026 a stored file fails at the provider with a 400 and the message "Invalid image file or mode for image 1". The image is not the problem: the same bytes sent as a data URL succeed.

Running a long job

Actions labelled "as a job" return a job id instead of a result, because the underlying work can take minutes. Wire them like this:

  1. Call the start action. Read public_id from its output.
  2. Add a Delay node sized to the work, then eden-ai/get-async-job with that id.
  3. Branch on status. While it is not finished, loop back to the delay.

If you would rather not poll, set the start action's Webhook Receiver to an inbound webhook URL and let Eden AI call you when the job finishes.

Known limitations

  • Every request is billed to the connected key's Eden AI account, which in turn pays the upstream provider. Rate limits, spend caps and model entitlements belong to the client's Eden AI organization, not to TaskJuice. The default limit is 10 requests per second across all of an account's keys, and Eden AI raises it on request.
  • When credit runs out, Eden AI answers HTTP 402 and TaskJuice surfaces it as a terminal payment-required error rather than retrying. Use eden-ai/get-credits to check the balance before an expensive branch.
  • Streaming is not exposed. Every action resolves once, with the full response.
  • Organization management is out of scope. Creating members, minting inference keys, and reading org-wide usage need an Eden AI management key carrying manage: scopes, which is a different credential from the inference key this connection holds.
  • Managing the account's API keys from a workflow is also out of scope, deliberately: a step that rotates or revokes its own connection credential would break the connection running it. eden-ai/inspect-api-key reads the connected key, which is the safe half of that surface.
  • Output shapes are normalized by Eden AI, but the depth of the result still varies by provider. Two providers behind the same action can return the same fields with different amounts of detail, so prefer the normalized fields over provider-specific ones when you map the output downstream.
Was this helpful?
Dropbox integrationEventbrite integration