The entry point for data. Receives raw job JSON from the Outbound Feed sync engine. Every pipeline starts with exactly one Input node.
| Property | Type | Description |
|---|
source | enum | jobs_feed, webhook, manual |
filters | object | Optional filters (same as feed filters) |
Liquid Template
Reshape and remap JSON fields using Liquid template syntax. The Liquid node takes the output of its parent node as input and renders a Liquid template against it. Your template must output valid JSON.
| Property | Type | Required | Description |
|---|
template | string | Yes | Liquid template string that outputs JSON |
Available Variables
The input data is available via data and dot notation:
| Variable | Type | Description |
|---|
data | object | The full JSON object from the parent node. Access nested fields with dot notation: data.company.name |
data.title | string | Job title |
data.description | string | Full job description (HTML) |
data.company | object | Company object with name, url, logo |
data.company.name | string | Company name |
data.company.url | string | Company website URL |
data.company.logo | string | Company logo URL |
data.location | object | Location object with city, state, country |
data.apply_url | string | Application URL |
data.published_at | string | Publication date |
data.tags | array | Array of tag strings |
Available Filters
All standard Shopify Liquid filters are supported, plus these custom filters:
| Filter | Description | Example |
|---|
slugify | Converts a string to a URL-safe slug | {{ data.title | slugify }} → software-engineer |
truncate_words | Truncates to a number of words | {{ data.description | truncate_words: 50 }} |
json_escape | Escapes a string for safe JSON embedding | {{ data.description | json_escape }} |
Example: Remap Fields
{
"jobTitle": "{{ data.title }}",
"employer": "{{ data.company.name }}",
"location": "{{ data.location.city }}, {{ data.location.state }}",
"applyUrl": "{{ data.apply_url }}",
"postedAt": "{{ data.published_at | date: '%Y-%m-%d' }}"
}
Example: SEO Slug Generation
{
"seo_title": "{{ data.title }} at {{ data.company.name }} - {{ data.location.city }}",
"slug": "{{ data.title | slugify }}-{{ data.company.name | slugify }}-{{ data.id | slice: 0, 8 }}",
"is_engineering": {% if data.tags contains 'engineering' %}true{% else %}false{% endif %}
}
AI (OpenRouter)
Use LLM intelligence to extract, classify, enrich, or rewrite job data. The AI node sends data to any model available on OpenRouter (GPT-4o, Claude, Gemini, Llama, etc.). Configure a system prompt for behavior and a user prompt with the {{input}} placeholder for the data.
Configuration
| Property | Type | Required | Default | Description |
|---|
model | string | Yes | — | OpenRouter model ID (e.g., google/gemini-2.0-flash-001, openai/gpt-4o-mini, anthropic/claude-3.5-sonnet) |
systemPrompt | string | Yes | — | System instructions for the AI. Should include “respond with valid JSON” for structured output. |
userPrompt | string | Yes | — | The prompt template. Use {{input}} to inject the parent node’s output as context. |
outputSchema | object | No | — | Optional JSON Schema to constrain the AI’s output structure. Strongly recommended for reliability. |
temperature | number | No | 0.3 | LLM temperature (0–2). Lower = more consistent, higher = more creative. |
The {{ input }} placeholder in the user prompt is replaced with the full
JSON output from the parent node. This is the primary way to pass data to the
AI model.
Supported Models
Any model available on OpenRouter can be used. Popular choices:
| Model | Best For |
|---|
google/gemini-2.0-flash-001 | Fast, cost-effective structured extraction |
openai/gpt-4o-mini | Balanced quality and speed |
openai/gpt-4o | Highest quality extraction and generation |
anthropic/claude-3.5-sonnet | Complex reasoning and long descriptions |
You are a job data extraction assistant. Always respond with valid JSON.
Extract technical skills, soft skills, and seniority level from the job posting.
Extract skills and seniority from this job posting:
{{input}}
Respond with:
{
"technical_skills": ["skill1", "skill2"],
"soft_skills": ["skill1"],
"seniority": "senior" | "mid" | "junior" | "lead"
}
Full Configuration Example
{
"model": "google/gemini-2.0-flash-001",
"systemPrompt": "You are a job data analyst. Extract structured requirements from job descriptions. Always respond with valid JSON.",
"userPrompt": "Extract requirements and responsibilities from this job:\n\n{{input}}",
"outputSchema": {
"type": "object",
"properties": {
"requirements": {
"type": "object",
"properties": {
"must_have": { "type": "array", "items": { "type": "string" } },
"preferred": { "type": "array", "items": { "type": "string" } }
}
},
"responsibilities": {
"type": "array",
"items": { "type": "string" }
}
}
},
"temperature": 0.2
}
JSON Merge
Combine two JSON inputs into a single unified object. The JSON Merge node has two input handles: base (original data) and patch (new data, typically AI output). This is essential for preserving original job data while adding AI-generated fields.
Configuration
| Property | Type | Required | Default | Description |
|---|
strategy | enum | No | Deep | Deep — recursively merges nested objects. Shallow — only merges top-level keys. |
conflictResolution | enum | No | PatchWins | PatchWins — patch values override base on conflict. BaseWins — base values are preserved on conflict. |
Deep Merge Example
Base (original data)
Patch (AI output)
Result (Deep, PatchWins)
{
"title": "Software Engineer",
"company": { "name": "Acme" },
"location": "San Francisco"
}
{
"skills": ["Python", "AWS"],
"seniority": "senior",
"salary_estimate": "$150k-$180k"
}
{
"title": "Software Engineer",
"company": { "name": "Acme" },
"location": "San Francisco",
"skills": ["Python", "AWS"],
"seniority": "senior",
"salary_estimate": "$150k-$180k"
}
Nested Deep Merge Example
When both base and patch have nested objects, deep merge recursively combines them:
Base
Patch
Result (Deep, PatchWins)
{
"title": "SWE",
"skills": { "hard": ["Python"] }
}
{
"skills": { "hard": ["Python", "Go"], "soft": ["Leadership"] }
}
{
"title": "SWE",
"skills": { "hard": ["Python", "Go"], "soft": ["Leadership"] }
}
Use Deep + PatchWins (the defaults) in most cases. This preserves all
original fields while letting AI-generated data override specific nested
values.
Output Node
The exit point for processed data. The final JSON from this node is what gets pushed to your destination. Every pipeline ends with exactly one Output node.
| Property | Type | Description |
|---|
destination | enum | database, webhook, file |
config | object | Destination-specific configuration |
