Package Information
Documentation
n8n-nodes-ezkielyna
n8n community nodes for the Ezkielyna AI API: chat with DeepSeek, GLM, Qwen, Kimi, Claude, GPT and other models, search the web, turn text into Indonesian speech, transcribe recordings and generate images, plus a chat model for the n8n AI Agent. One credential, one API key, pay from $0.5 with QRIS.
Bahasa Indonesia: lihat bagian di bawah.
Ezkielyna is an independent AI API gateway (OpenAI-compatible) for Indonesian developers. It is not affiliated with the model vendors.
Contents: Install · Credentials · Operations · Ezkielyna Chat Model (AI Agent) · Example workflows · Costs and limits · FAQ · Development · Bahasa Indonesia
Install
Self-hosted n8n (owner or admin account):
- Open Settings → Community Nodes → Install.
- Enter
n8n-nodes-ezkielyna, tick the risk notice, click Install. - Search for Ezkielyna in the node panel.
n8n 2.47 warns that the default of N8N_UNVERIFIED_PACKAGES_ENABLED will change to false. Until this package is verified by n8n, admins on such a version must set N8N_UNVERIFIED_PACKAGES_ENABLED=true to install it.
n8n Cloud: community nodes can be installed only after the package is verified by n8n. Until then, use the OpenAI Chat Model with a custom Base URL instead.
Manual install (Docker or queue mode): run npm install n8n-nodes-ezkielyna in ~/.n8n/nodes and restart n8n. See the n8n guide.
Requirements: a 2026 n8n release for the Ezkielyna Chat Model sub-node (it uses n8n's AI node SDK, tested with n8n 2.47.4). The main Ezkielyna node does not need the SDK; it was tested on n8n 2.47.4 only.
Credentials
Buy API Credit at ezkielyna.store/beli-api-key (from $0.5; QRIS, DANA, GoPay or USDT) and redeem it under Wallet at api.ezkielyna.store.
Create a key under API Keys. The key group decides which models the key can use:
Group Models Use it for Standard DeepSeek, GLM, Kimi K2, MiniMax, Qwen, MiMo and other text models; audio models such as edge-ttsandwhisper-large-v3Chat, search, text to speech, transcription Premium (1.5x) Claude, GPT, Grok, Gemini Pro and others; audio models Chat with those models Media Image and video models (GPT Image, Nano Banana, Flux, Imagen...) Image generation Auto Everything above One key for every operation In n8n: Credentials → Add credential → Ezkielyna API. Paste the key (
sk-...). Leave Base URL athttps://api.ezkielyna.store/v1. Click Save: n8n tests the key withGET /v1/models.
Operations
Node Ezkielyna (resources and operations):
| Resource | Operation | Endpoint | Output |
|---|---|---|---|
| Text | Message a Model | POST /v1/chat/completions |
text, model, finishReason, usage (or the raw response with Simplify Output off) |
| Web Search | Search | POST /v1/search |
One item per result: position, title, url, snippet, published (or one item with results) |
| Audio | Text to Speech | POST /v1/audio/speech |
Audio file in a binary field (default data) |
| Audio | Transcribe | POST /v1/audio/transcriptions (multipart) |
text (or SRT/VTT subtitles), language, duration |
| Image | Generate | POST /v1/images/generations |
One item per image with the file in a binary field; url when the model returns a link |
| Model | Get Many | GET /v1/models |
id, kind (chat, speech, transcription, image, video, embedding), endpointTypes |
Details:
- Message a Model: model dropdown loaded from your key (default
deepseek-v4.1-flash), Prompt, and options System Message, Sampling Temperature (default 0.7), Maximum Number of Tokens (default 1024) and JSON Output (sendsresponse_format: json_objectand returns the parsed object inparsed; mention JSON in the prompt). Models that think out loud return their reasoning inreasoning. - Search: query 1 to 400 characters, 1 to 10 results, optional time range (day, week, month, year) and language code (
id,en). - Text to Speech: default model
edge-tts, default voiceid-ID-GadisNeural(female;id-ID-ArdiNeuralis male). Up to 4,096 characters per item. Formats MP3 (default), WAV, Opus, AAC, FLAC, as far as the model supports them. - Transcribe: reads the audio file from a binary field (from Telegram, HTTP Request, Google Drive, Read Binary File...), max 25 MB. Default model
whisper-large-v3. Options: language (idfor Indonesian; empty = auto-detect), a prompt with names to spell, output as text, SRT or VTT. - Generate (image): default model
gpt-image-2(Media or Auto key). Options: number of images (1-4, each billed), size (1024x1024...), quality, file or link. Links are downloaded into a binary field unless Download Image is off; your API key is never sent to the image host. - Get Many (models): filter by type, Return All or a limit.
The node can also be used as a tool by the AI Agent (n8n shows it as Ezkielyna Tool), for example to let an agent search the web.
Model lists never show models whose id contains uncensored, abliterated or nsfw.
Ezkielyna Chat Model (AI Agent)
A sub-node that plugs into AI Agent, Basic LLM Chain, Question and Answer Chain, Summarization Chain and other nodes with a Chat Model input. Pick the model from the dropdown (default deepseek-v4.1-flash). Options: temperature, maximum number of tokens (default 1024, -1 = model default), top P, frequency and presence penalty, timeout, max retries.
It is built on n8n's @n8n/ai-node-sdk, which n8n ships with its 2026 releases, so the package has no runtime dependencies. On an older n8n the sub-node shows an error asking you to update; the main node keeps working.
For agents with tools, choose a model that supports tool calling (for example deepseek-v4.1-flash) and test it once in the chat panel.
Example workflows
Import from Workflows → Create workflow → ⋯ → Import from File:
examples/01-berita-jadi-suara.json: Manual Trigger → Web Search (today's news) → Message a Model (3-sentence summary in Indonesian) → Text to Speech (MP3,id-ID-GadisNeural).examples/02-agen-ai-pencarian-web.json: Chat Trigger → AI Agent with Ezkielyna Chat Model, Simple Memory and the Ezkielyna node as a web search tool.
After importing, select your Ezkielyna API credential in each Ezkielyna node. More Indonesian workflows: EzkielynaGroup/n8n-workflows-indonesia.
Costs and limits
Every call is billed to your Ezkielyna balance. Public list prices on 11 Nov 2026 (live price list):
| Model | Price |
|---|---|
deepseek-v4.1-flash |
$0.04 per 1M input tokens, $0.16 per 1M output tokens |
edge-tts |
$1.00 per 1M tokens |
whisper-large-v3 |
$12.00 per 1M tokens |
gpt-image-2 |
$0.0083 per image |
Prices change; the price list and Usage Logs at api.ezkielyna.store are the source of truth. Safety defaults in the nodes: max 1024 tokens per chat reply, 4,096 characters per speech item, 25 MB per audio file, 4 images per item. While building a workflow, pin the output of a node that already worked so test runs do not call the API again.
FAQ
How do I use DeepSeek in n8n?
Install n8n-nodes-ezkielyna, create an Ezkielyna API credential with a Standard or Auto key, and use Ezkielyna → Text → Message a Model with deepseek-v4.1-flash (or the Ezkielyna Chat Model under an AI Agent).
Can I use Ezkielyna in n8n without installing a community node?
Yes. In the built-in OpenAI Chat Model node, create an OpenAI credential with your Ezkielyna key and Base URL https://api.ezkielyna.store/v1, then type the model ID (for example deepseek-v4.1-flash). The HTTP Request node also works with header Authorization: Bearer sk-....
How do I turn text into Indonesian speech in n8n?
Use Ezkielyna → Audio → Text to Speech with model edge-tts and voice id-ID-GadisNeural (or id-ID-ArdiNeural). The MP3 lands in the data binary field, ready for Telegram, WhatsApp or Google Drive nodes.
How do I transcribe a Telegram voice note in n8n?
Telegram Trigger → Telegram (Get a File, download on) → Ezkielyna → Audio → Transcribe with language id. The text is in the text field.
Which key group do I need?
Standard covers chat with text models, web search, text to speech and transcription. Image generation needs Media. An Auto key works for every operation.
Is my API key safe?
n8n stores it encrypted as a credential. The node only sends it to the Base URL in the credential, never to image download links, and never writes it to the output.
Development
npm install
npm run build # n8n-node build → dist/
npm run lint # n8n community node linter
npm test # build + node --test test/ (local mock gateway, no internet, no cost)
npm run dev # n8n with this package loaded, hot reload
Tests cover request building and response mapping for every operation, binary in/out, multipart upload, errors and Continue On Fail, the credential, and the AI sub-node through the real n8n AI node SDK, all against a local mock server. The packed tarball was also installed as a community package in a local n8n 2.47.4 and both example workflows ran end to end against the same mock (including an AI Agent tool call to Web Search).
License
MIT © 2026 Ezekiel Lieberhart (Ezkielyna). Model names belong to their owners.
Bahasa Indonesia
Node komunitas n8n untuk Ezkielyna AI API: chat dengan DeepSeek, GLM, Qwen, Kimi, Claude, GPT dan model lain, cari di web, ubah teks jadi suara Bahasa Indonesia, transkrip rekaman, buat gambar, plus chat model untuk node AI Agent n8n. Satu credential, satu API key, isi saldo mulai $0.5 lewat QRIS.
Ezkielyna adalah gateway API AI independen (kompatibel OpenAI) untuk developer Indonesia, tidak berafiliasi dengan pembuat model.
Instalasi
n8n self-hosted (akun owner atau admin):
- Buka Settings → Community Nodes → Install.
- Isi
n8n-nodes-ezkielyna, centang pernyataan risiko, klik Install. - Cari Ezkielyna di panel node.
n8n 2.47 memberi peringatan bahwa default N8N_UNVERIFIED_PACKAGES_ENABLED akan berubah menjadi false. Selama paket ini belum diverifikasi n8n, admin di versi tersebut perlu mengisi N8N_UNVERIFIED_PACKAGES_ENABLED=true agar bisa memasangnya.
n8n Cloud: node komunitas baru bisa dipasang setelah paket diverifikasi n8n. Sementara itu, pakai node OpenAI Chat Model dengan Base URL kustom (lihat FAQ).
Pasang manual (Docker atau queue mode): jalankan npm install n8n-nodes-ezkielyna di folder ~/.n8n/nodes, lalu restart n8n.
Syarat: sub-node Ezkielyna Chat Model butuh n8n rilis 2026 (memakai AI node SDK n8n; diuji dengan n8n 2.47.4). Node utama Ezkielyna tidak butuh SDK itu; baru diuji di n8n 2.47.4.
Credential
Beli API Credit di ezkielyna.store/beli-api-key (mulai $0.5; QRIS, DANA, GoPay, atau USDT), lalu tukarkan di menu Wallet di api.ezkielyna.store.
Buat key di menu API Keys. Grup key menentukan model yang bisa dipakai:
Grup Model Untuk Standard DeepSeek, GLM, Kimi K2, MiniMax, Qwen, MiMo dan model teks lain; model audio seperti edge-ttsdanwhisper-large-v3Chat, pencarian web, teks ke suara, transkripsi Premium (1,5x) Claude, GPT, Grok, Gemini Pro dan lainnya; model audio Chat dengan model tersebut Media Model gambar dan video (GPT Image, Nano Banana, Flux, Imagen...) Membuat gambar Auto Semua di atas Satu key untuk semua operasi Di n8n: Credentials → Add credential → Ezkielyna API. Tempel key (
sk-...). Biarkan Base URLhttps://api.ezkielyna.store/v1. Klik Save: n8n menguji key lewatGET /v1/models.
Operasi
| Resource | Operasi | Hasil |
|---|---|---|
| Text | Message a Model | text, model, finishReason, usage (atau respons mentah kalau Simplify Output dimatikan) |
| Web Search | Search | Satu item per hasil: position, title, url, snippet, published (atau satu item berisi results) |
| Audio | Text to Speech | File audio di field biner (default data) |
| Audio | Transcribe | text (atau subtitle SRT/VTT), language, duration |
| Image | Generate | Satu item per gambar, file di field biner; url kalau model mengirim tautan |
| Model | Get Many | id, kind (chat, speech, transcription, image, video, embedding), endpointTypes |
- Message a Model: daftar model diambil dari key kamu (default
deepseek-v4.1-flash), Prompt, dan opsi System Message, Sampling Temperature (default 0,7), Maximum Number of Tokens (default 1024), JSON Output (hasil JSON yang sudah di-parse ada di fieldparsed; sebut "JSON" di prompt). - Search: kata kunci 1-400 karakter, 1-10 hasil, opsi rentang waktu (hari, minggu, bulan, tahun) dan kode bahasa (
id,en). - Text to Speech: model default
edge-tts, suara defaultid-ID-GadisNeural(perempuan;id-ID-ArdiNeurallaki-laki). Maksimal 4.096 karakter per item. Format MP3 (default), WAV, Opus, AAC, FLAC sejauh didukung model. - Transcribe: membaca file audio dari field biner (dari Telegram, HTTP Request, Google Drive...), maksimal 25 MB. Model default
whisper-large-v3. Opsi: bahasa (id; kosong = deteksi otomatis), prompt berisi nama/istilah agar ejaannya benar, hasil teks, SRT, atau VTT. - Generate (gambar): model default
gpt-image-2(key Media atau Auto). Opsi: jumlah gambar (1-4, masing-masing ditagih), ukuran, kualitas, file atau tautan. Tautan otomatis diunduh ke field biner kecuali Download Image dimatikan; API key tidak pernah dikirim ke host gambar. - Get Many (model): filter per jenis, Return All atau batas jumlah.
Node ini juga bisa dipakai sebagai tool oleh AI Agent (muncul sebagai Ezkielyna Tool), misalnya agar agen bisa mencari di web.
Daftar model tidak pernah menampilkan model yang id-nya mengandung uncensored, abliterated, atau nsfw.
Ezkielyna Chat Model (AI Agent)
Sub-node untuk AI Agent, Basic LLM Chain, Question and Answer Chain, Summarization Chain dan node lain yang punya input Chat Model. Pilih model dari daftar (default deepseek-v4.1-flash). Opsi: temperature, batas token (default 1024, -1 = default model), top P, frequency/presence penalty, timeout, jumlah retry. Dibangun di atas @n8n/ai-node-sdk bawaan n8n rilis 2026, jadi paket ini tanpa dependensi runtime. Untuk agen yang memakai tool, pilih model yang mendukung tool calling (misalnya deepseek-v4.1-flash) dan coba sekali di panel chat.
Contoh workflow
Impor lewat Workflows → Create workflow → ⋯ → Import from File:
examples/01-berita-jadi-suara.json: cari berita hari ini → ringkas 3 kalimat → jadi MP3 suara Bahasa Indonesia.examples/02-agen-ai-pencarian-web.json: chat dengan AI Agent yang memakai Ezkielyna Chat Model, Simple Memory, dan node Ezkielyna sebagai tool pencarian web.
Setelah impor, pilih credential Ezkielyna API di setiap node Ezkielyna. Workflow Indonesia lainnya: EzkielynaGroup/n8n-workflows-indonesia.
Biaya dan batas
Setiap panggilan memotong saldo Ezkielyna. Harga daftar publik per 11 November 2026: deepseek-v4.1-flash $0,04 per 1 juta token input dan $0,16 per 1 juta token output; edge-tts $1 per 1 juta token; whisper-large-v3 $12 per 1 juta token; gpt-image-2 $0,0083 per gambar. Harga bisa berubah; cek daftar harga live dan Usage Logs di api.ezkielyna.store. Saat menyusun workflow, pin output node yang sudah berhasil supaya uji coba tidak memanggil API berulang kali.
FAQ
Bagaimana cara pakai DeepSeek di n8n?
Pasang n8n-nodes-ezkielyna, buat credential Ezkielyna API dengan key Standard atau Auto, lalu pakai Ezkielyna → Text → Message a Model dengan deepseek-v4.1-flash (atau Ezkielyna Chat Model di bawah AI Agent).
Bisa pakai Ezkielyna di n8n tanpa memasang node komunitas?
Bisa. Di node bawaan OpenAI Chat Model, buat credential OpenAI berisi key Ezkielyna dan Base URL https://api.ezkielyna.store/v1, lalu ketik ID model (misalnya deepseek-v4.1-flash). Node HTTP Request juga bisa dengan header Authorization: Bearer sk-....
Bagaimana cara membuat suara Bahasa Indonesia (text to speech) di n8n?
Pakai Ezkielyna → Audio → Text to Speech dengan model edge-tts dan suara id-ID-GadisNeural (atau id-ID-ArdiNeural). File MP3 ada di field biner data, siap dikirim lewat node Telegram, WhatsApp, atau Google Drive.
Bagaimana cara mengubah voice note Telegram jadi teks di n8n?
Telegram Trigger → Telegram (Get a File, download aktif) → Ezkielyna → Audio → Transcribe dengan bahasa id. Teksnya ada di field text.
Grup key mana yang saya perlukan?
Standard cukup untuk chat model teks, pencarian web, teks ke suara, dan transkripsi. Membuat gambar butuh Media. Key Auto bisa untuk semua operasi.
Apakah API key saya aman?
n8n menyimpannya terenkripsi sebagai credential. Node hanya mengirim key ke Base URL di credential, tidak ke tautan unduhan gambar, dan tidak pernah menuliskannya ke output.
Lisensi
MIT © 2026 Ezekiel Lieberhart (Ezkielyna).