Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.tryfltr.com/llms.txt

Use this file to discover all available pages before exploring further.

​
n8n Integration

n8n is an open-source workflow automation tool that you can self-host or use in the cloud. Perfect for developers who want full control over their automation workflows.

​
Why n8n?

Self-Hosted

Deploy on your own infrastructure for complete control

Open Source

Inspect, modify, and extend the platform

Developer-Friendly

Code nodes, JavaScript expressions, and custom nodes

Fair Pricing

Free self-hosted or affordable cloud plans

​
Installation

​
n8n Cloud

Fastest way to get started:
  1. Sign up at n8n.cloud
  2. Create your first workflow
  3. No setup required

​
Self-Hosted (Docker)

# Using Docker
docker run -it --rm \
  --name n8n \
  -p 5678:5678 \
  -v ~/.n8n:/home/node/.n8n \
  n8nio/n8n

# Access at http://localhost:5678

​
Self-Hosted (npm)

npm install n8n -g
n8n start
For production deployments, see the n8n deployment guide.

​
Quick Start: FLTR Search Workflow

​
Step 1: Create New Workflow

  1. Open n8n interface
  2. Click New workflow
  3. Name it “FLTR Knowledge Search”

​
Step 2: Add HTTP Request Node

  1. Click + to add node
  2. Search for HTTP Request
  3. Configure:
Method: POST URL: https://api.fltr.com/v1/mcp/query Authentication: Generic Credential Type → Header Auth
  • Name: Authorization
  • Value: Bearer YOUR_API_KEY
Body Parameters (JSON): 
{
  "query": "{{ $json.query }}",
  "dataset_id": "ds_abc123",
  "limit": 5
}

​
Step 3: Test the Node

  1. Click Execute Node
  2. View results in the output panel
  3. Inspect the JSON response

​
Step 4: Process Results

Add a Code node to transform results: 
// Access FLTR results
const results = $input.first().json.results;

// Format for output
const formatted = results.map(result => ({
  title: result.metadata?.title || 'Untitled',
  content: result.content,
  score: Math.round(result.score * 100),
  chunk_id: result.chunk_id
}));

return formatted.map(item => ({ json: item }));

​
Common Workflows

Nodes:
  1. Email Trigger (IMAP) - Watch inbox
  2. HTTP Request - FLTR Query
  3. Code - Format results
  4. Send Email - Reply with results
FLTR HTTP Request: 
{
  "query": "{{ $json.subject }} {{ $json.text }}",
  "dataset_id": "ds_support_kb",
  "limit": 3,
  "rerank": true
}
Code Node (Format Response): 
const results = $input.first().json.results;

const response = `
Hi,

Based on your email, here are some helpful resources:

${results.map((r, i) => `
${i + 1}. ${r.metadata.title} (${Math.round(r.score * 100)}% match)
${r.content.substring(0, 200)}...
`).join('\n')}

Hope this helps!
`;

return [{ json: { response, to: $('Email Trigger').first().json.from } }];

​
2. Slack Q&A Bot

Nodes:
  1. Slack Trigger - New message in channel
  2. IF - Check if question
  3. HTTP Request - FLTR Query
  4. Slack - Post reply
IF Condition: 
{{ $json.text.includes('?') }}
FLTR Query: 
{
  "query": "{{ $json.text }}",
  "dataset_id": "ds_company_kb",
  "limit": 1
}
Slack Reply: 
{{ $json.results[0].content }}

_Source: {{ $json.results[0].metadata.title }}_
_Confidence: {{ Math.round($json.results[0].score * 100) }}%_

​
3. Document Indexer Pipeline

Nodes:
  1. Webhook - Receive document
  2. Code - Extract metadata
  3. HTTP Request - Upload to FLTR
  4. Slack - Send notification
Code Node (Extract Metadata): 
const doc = $input.first().json;

return [{
  json: {
    content: doc.text,
    metadata: {
      title: doc.title,
      author: doc.author,
      source: 'API Upload',
      created_at: new Date().toISOString(),
      tags: doc.tags || []
    }
  }
}];
HTTP Request (Upload): 
Method: POST
URL: https://api.fltr.com/v1/datasets/ds_abc123/documents

Body:
{
  "content": "{{ $json.content }}",
  "metadata": {{ JSON.stringify($json.metadata) }}
}

​
4. Customer Support Automation

Nodes:
  1. Webhook Trigger - New support ticket
  2. HTTP Request - FLTR Query
  3. Switch - Route by confidence
  4. Multiple endpoints - Different actions per route
Switch Routes: Route 1: {{ $json.results[0].score >= 0.8 }} → Auto-respond with answer Route 2: {{ $json.results[0].score >= 0.5 && $json.results[0].score < 0.8 }} → Create ticket with suggestions Route 3: {{ $json.results[0].score < 0.5 }} → Escalate to human

​
Advanced Techniques

​
Using Code Nodes

n8n’s Code node supports full JavaScript: Example: Filter and rank results 
// Get FLTR results
const results = $input.first().json.results;

// Filter by score threshold
const filtered = results.filter(r => r.score >= 0.7);

// Sort by score (descending)
const sorted = filtered.sort((a, b) => b.score - a.score);

// Add ranking
const ranked = sorted.map((r, i) => ({
  ...r,
  rank: i + 1,
  score_percent: Math.round(r.score * 100)
}));

return ranked.map(item => ({ json: item }));
Example: Combine multiple queries 
// Run multiple FLTR queries and combine results
const query1 = $('HTTP Request 1').first().json.results;
const query2 = $('HTTP Request 2').first().json.results;

// Merge and deduplicate by chunk_id
const allResults = [...query1, ...query2];
const unique = allResults.reduce((acc, curr) => {
  if (!acc.find(r => r.chunk_id === curr.chunk_id)) {
    acc.push(curr);
  }
  return acc;
}, []);

// Sort by score
const sorted = unique.sort((a, b) => b.score - a.score);

return sorted.map(item => ({ json: item }));

​
Loop Over Results

Use Split in Batches to process each result:
  1. Add Split in Batches after FLTR query
  2. Set batch size to 1
  3. Connect to action node
  4. Each result processed individually
Example: Create Notion page for each result 
Input: {{ $json.results }}
Batch size: 1

Notion node:
Title: {{ $json.metadata.title }}
Content: {{ $json.content }}
Tags: {{ $json.metadata.category }}

​
Error Handling

Add error workflows:
  1. On any node, click Settings (gear icon)
  2. Enable Continue On Fail
  3. Add IF node to check for errors
  4. Route failures to error handler
Error Check: 
{{ $json.error !== undefined }}
Error Notification: 
⚠️ Workflow Error

Node: {{ $node.name }}
Error: {{ $json.error }}
Input: {{ JSON.stringify($json) }}

​
Scheduling Workflows

Add Schedule Trigger for recurring tasks: Example: Daily document sync
  • Trigger: Cron - 0 2 * * * (2 AM daily)
  • Action: Fetch new documents from source
  • Upload to FLTR
  • Send summary email

​
Using Credentials

Store API keys securely:
  1. Go to Credentials → New
  2. Select Header Auth
  3. Name: FLTR API Key
  4. Header name: Authorization
  5. Header value: Bearer YOUR_API_KEY
Use in HTTP Request nodes:
  • Authentication: FLTR API Key

​
All FLTR Endpoints

​
Query Dataset

Node: HTTP Request 
Method: POST
URL: https://api.fltr.com/v1/mcp/query

Authentication: Header Auth
Header: Authorization
Value: Bearer YOUR_API_KEY

Body (JSON):
{
  "query": "{{ $json.query }}",
  "dataset_id": "ds_abc123",
  "limit": 5,
  "rerank": false
}

​
Batch Query

URL: https://api.fltr.com/v1/mcp/batch-query

Body:
{
  "queries": {{ JSON.stringify($json.queries) }},
  "dataset_id": "ds_abc123",
  "limit": 3
}

​
Upload Document

URL: https://api.fltr.com/v1/datasets/DATASET_ID/documents

Body:
{
  "content": "{{ $json.content }}",
  "metadata": {{ JSON.stringify($json.metadata) }}
}

​
Create Dataset

URL: https://api.fltr.com/v1/datasets

Body:
{
  "name": "{{ $json.name }}",
  "description": "{{ $json.description }}",
  "is_public": false
}

​
List Datasets

Method: GET
URL: https://api.fltr.com/v1/datasets

Authentication: Header Auth

​
Expressions and Functions

​
Accessing Data

// Current node data
{{ $json.field }}

// Previous node data
{{ $('Node Name').first().json.field }}

// All items from previous node
{{ $('Node Name').all() }}

// Item by index
{{ $('Node Name').item.json.field }}

​
String Operations

// Substring
{{ $json.content.substring(0, 100) }}

// Uppercase/lowercase
{{ $json.title.toUpperCase() }}

// Replace
{{ $json.text.replace('old', 'new') }}

// Includes
{{ $json.text.includes('keyword') }}

​
Array Operations

// Map
{{ $json.results.map(r => r.title) }}

// Filter
{{ $json.results.filter(r => r.score > 0.7) }}

// Join
{{ $json.tags.join(', ') }}

// Length
{{ $json.results.length }}

​
Math Functions

// Round
{{ Math.round($json.score * 100) }}

// Max/Min
{{ Math.max(...$json.scores) }}

// Random
{{ Math.random() }}

​
Rate Limiting

Handle FLTR’s rate limits in n8n:

​
Method 1: Wait Node

Add Wait node between iterations: 
Amount: 1
Unit: Seconds

​
Method 2: Code Node Delay

// Add delay before returning
await new Promise(resolve => setTimeout(resolve, 1000));

return $input.all();

​
Method 3: Error Handling

// In error workflow
if ($json.error?.includes('rate limit')) {
  // Wait 1 hour
  await new Promise(resolve => setTimeout(resolve, 3600000));

  // Retry
  return [{ json: { retry: true } }];
}

​
Webhook Workflows

​
Create Webhook Endpoint

  1. Add Webhook node
  2. Set HTTP Method: POST
  3. Copy webhook URL
  4. Configure response
Webhook Response: 
{
  "status": "success",
  "results": {{ JSON.stringify($('HTTP Request').first().json.results) }}
}

​
Secure Webhooks

Validate incoming requests:
  1. Add IF node after webhook
  2. Check secret header:
{{ $json.headers.authorization === 'Bearer YOUR_SECRET' }}
  1. Route invalid requests to error response

​
Production Best Practices

​
1. Use Environments

Set environment variables: 
# In docker-compose.yml
environment:
  - N8N_ENCRYPTION_KEY=your-key
  - FLTR_API_KEY=your-api-key
  - FLTR_DATASET_ID=ds_abc123
Access in workflows: 
{{ $env.FLTR_API_KEY }}

​
2. Enable Execution Data

Keep execution history: Settings → Workflow Settings:
  • Save execution progress: Yes
  • Save manual executions: Yes
  • Save error executions: Yes

​
3. Add Monitoring

Create monitoring workflow:
  • Trigger: Schedule (every 5 minutes)
  • Check: Last execution status
  • Alert: Send notification if failed

​
4. Optimize Performance

  • Use Merge nodes to combine data
  • Enable Continue On Fail for non-critical nodes
  • Use Split in Batches for large datasets
  • Cache results with Set node

​
Complete Example: Support Ticket System

Here’s a production-ready workflow: Node 1: Webhook 
HTTP Method: POST
Path: support-ticket
Response Mode: When Last Node Finishes
Node 2: Validate Input 
// Code node
const required = ['email', 'subject', 'description'];
const missing = required.filter(f => !$json[f]);

if (missing.length > 0) {
  throw new Error(`Missing fields: ${missing.join(', ')}`);
}

return [$input.first()];
Node 3: FLTR Query 
{
  "query": "{{ $json.subject }} {{ $json.description }}",
  "dataset_id": "{{ $env.FLTR_SUPPORT_KB }}",
  "limit": 5,
  "rerank": true
}
Node 4: Switch (Route by Confidence) Route 1: {{ $json.results[0].score >= 0.8 }} Route 2: {{ $json.results[0].score >= 0.5 }} Route 3: {{ $json.results[0].score < 0.5 }} Node 5a: Auto-Response Email 
To: {{ $('Webhook').first().json.email }}
Subject: Re: {{ $('Webhook').first().json.subject }}

Hi,

{{ $json.results[0].content }}

Source: {{ $json.results[0].metadata.title }}

If this doesn't help, our team will follow up.
Node 5b: Create Zendesk Ticket 
{
  "subject": "{{ $('Webhook').first().json.subject }}",
  "description": "{{ $('Webhook').first().json.description }}",
  "requester": "{{ $('Webhook').first().json.email }}",
  "tags": ["fltr-suggested"],
  "comment": {
    "body": "Suggested docs:\n{{ $json.results.map(r => r.metadata.title).join('\n') }}"
  }
}
Node 5c: Slack Alert 
⚠️ Low-confidence ticket

From: {{ $('Webhook').first().json.email }}
Score: {{ Math.round($json.results[0].score * 100) }}%
Subject: {{ $('Webhook').first().json.subject }}
Node 6: Webhook Response 
{
  "status": "processed",
  "confidence": {{ $('FLTR Query').first().json.results[0].score }},
  "action": "{{ $runIndex === 0 ? 'auto-responded' : $runIndex === 1 ? 'ticket-created' : 'escalated' }}"
}

​
Resources

n8n Documentation

Official n8n docs

FLTR API Reference

Complete API documentation

n8n Community

Get help from the community

Deployment Guide

Self-host n8n in production

​
Next Steps