This workflow follows the HTTP Request → Postgres recipe pattern — see all workflows that pair these two integrations.
The workflow JSON
Copy or download the full n8n JSON below. Paste it into a new n8n workflow, add your credentials, activate. Full import guide →
{
"name": "Continuous Learning Pipeline",
"nodes": [
{
"parameters": {
"rule": {
"interval": [
{
"field": "cronExpression",
"expression": "0 2 * * 0"
}
]
}
},
"id": "learning-trigger-001",
"name": "Weekly Learning Trigger",
"type": "n8n-nodes-base.scheduleTrigger",
"typeVersion": 1.2,
"position": [
240,
300
]
},
{
"parameters": {
"operation": "executeQuery",
"query": "SELECT * FROM voice_interactions_v2 WHERE created_at >= NOW() - INTERVAL '7 days' AND user_feedback IS NOT NULL ORDER BY created_at DESC LIMIT 5000;",
"options": {}
},
"id": "collect-training-data-002",
"name": "Collect Training Data",
"type": "n8n-nodes-base.postgres",
"typeVersion": 2.5,
"position": [
460,
300
],
"credentials": {
"postgres": {
"name": "<your credential>"
}
}
},
{
"parameters": {
"language": "javaScript",
"jsCode": "// Enhanced training data preparation with quality filtering\nconst rawData = $input.all().map(item => item.json);\nconsole.log(`Processing ${rawData.length} raw interactions`);\n\n// Quality filtering criteria\nfunction filterQualityData(interactions) {\n return interactions.filter(interaction => {\n const feedback = interaction.user_feedback || {};\n const metrics = interaction.processing_metrics || {};\n \n // Quality criteria\n const hasPositiveFeedback = feedback.rating >= 4 || feedback.helpful === true;\n const hasGoodConfidence = metrics.stt_confidence >= 0.8;\n const hasReasonableLength = interaction.transcription && interaction.transcription.length >= 10;\n const hasValidResponse = interaction.llm_response && interaction.llm_response.length >= 10;\n const noErrors = !metrics.errors || metrics.errors.length === 0;\n \n return hasPositiveFeedback && hasGoodConfidence && hasReasonableLength && hasValidResponse && noErrors;\n });\n}\n\n// Extract conversation pairs\nfunction extractTrainingPairs(interactions) {\n return interactions.map(interaction => {\n const userMessage = interaction.transcription;\n const assistantResponse = interaction.user_feedback?.corrected_response || interaction.llm_response;\n const context = interaction.processing_metrics?.context || '';\n \n return {\n messages: [\n {\n role: 'system',\n content: 'You are a helpful AI assistant specialized in voice interactions. Keep responses natural and conversational.'\n },\n {\n role: 'user', \n content: userMessage\n },\n {\n role: 'assistant',\n content: assistantResponse\n }\n ],\n metadata: {\n interaction_id: interaction.id,\n confidence: interaction.processing_metrics?.stt_confidence,\n rating: interaction.user_feedback?.rating,\n language: interaction.processing_metrics?.language || 'en',\n timestamp: interaction.created_at\n }\n };\n });\n}\n\n// Group by language and quality\nfunction groupAndBalance(trainingPairs) {\n const grouped = {};\n \n trainingPairs.forEach(pair => {\n const lang = pair.metadata.language;\n if (!grouped[lang]) grouped[lang] = [];\n grouped[lang].push(pair);\n });\n \n // Balance datasets - ensure we don't have too much of one language\n const maxPerLanguage = 1000;\n Object.keys(grouped).forEach(lang => {\n if (grouped[lang].length > maxPerLanguage) {\n // Sort by rating and take best examples\n grouped[lang].sort((a, b) => (b.metadata.rating || 0) - (a.metadata.rating || 0));\n grouped[lang] = grouped[lang].slice(0, maxPerLanguage);\n }\n });\n \n return grouped;\n}\n\n// Generate training dataset\nconst qualityData = filterQualityData(rawData);\nconst trainingPairs = extractTrainingPairs(qualityData);\nconst groupedData = groupAndBalance(trainingPairs);\n\n// Flatten and shuffle\nconst allPairs = Object.values(groupedData).flat();\nconst shuffled = allPairs.sort(() => 0.5 - Math.random());\n\n// Split train/validation\nconst splitIndex = Math.floor(shuffled.length * 0.85);\nconst trainData = shuffled.slice(0, splitIndex);\nconst validationData = shuffled.slice(splitIndex);\n\n// Create dataset metadata\nconst datasetId = `voice_ai_${new Date().toISOString().split('T')[0].replace(/-/g, '')}`;\nconst datasetMetadata = {\n id: datasetId,\n created: new Date().toISOString(),\n source: 'voice_interactions_feedback',\n quality: {\n total_raw: rawData.length,\n quality_filtered: qualityData.length,\n final_training: trainData.length,\n final_validation: validationData.length,\n filter_rate: Math.round((qualityData.length / rawData.length) * 100)\n },\n languages: Object.keys(groupedData),\n distribution: Object.fromEntries(\n Object.entries(groupedData).map(([lang, data]) => [lang, data.length])\n )\n};\n\nconsole.log('Dataset prepared:', datasetMetadata);\n\nreturn {\n json: {\n dataset: {\n metadata: datasetMetadata,\n train: trainData,\n validation: validationData\n },\n readyForTraining: trainData.length >= 50\n }\n};"
},
"id": "prepare-dataset-003",
"name": "Prepare Enhanced Dataset",
"type": "n8n-nodes-base.code",
"typeVersion": 2,
"position": [
680,
300
]
},
{
"parameters": {
"conditions": {
"boolean": [
{
"value1": "={{ $json.readyForTraining }}",
"operation": "equal",
"value2": true
}
],
"number": [
{
"value1": "={{ $json.dataset.train.length }}",
"operation": "largerEqual",
"value2": 50
}
]
}
},
"id": "check-training-threshold-004",
"name": "Check Training Threshold",
"type": "n8n-nodes-base.if",
"typeVersion": 1,
"position": [
900,
300
]
},
{
"parameters": {
"language": "javaScript",
"jsCode": "// Create fine-tuning configuration for Ollama\nconst dataset = $json.dataset;\nconst baseModel = 'llama3.2:8b';\n\n// Advanced training configuration\nconst trainingConfig = {\n model: baseModel,\n dataset_id: dataset.metadata.id,\n training_data: dataset.train,\n validation_data: dataset.validation,\n hyperparameters: {\n learning_rate: 0.0001,\n batch_size: 4,\n epochs: 3,\n warmup_steps: Math.floor(dataset.train.length * 0.1),\n weight_decay: 0.01,\n gradient_accumulation_steps: 2,\n max_grad_norm: 1.0,\n lr_scheduler: 'cosine',\n save_steps: Math.floor(dataset.train.length / 4)\n },\n model_config: {\n context_length: 4096,\n rope_scaling: 1.0,\n attention_dropout: 0.1,\n hidden_dropout: 0.1\n },\n training_arguments: {\n fp16: true,\n dataloader_num_workers: 4,\n remove_unused_columns: false,\n logging_steps: 10,\n evaluation_strategy: 'steps',\n eval_steps: Math.floor(dataset.train.length / 8),\n save_strategy: 'steps',\n load_best_model_at_end: true,\n metric_for_best_model: 'eval_loss',\n greater_is_better: false\n },\n job_metadata: {\n job_id: `finetune_${dataset.metadata.id}`,\n created_at: new Date().toISOString(),\n languages: dataset.metadata.languages,\n data_quality: dataset.metadata.quality\n }\n};\n\nreturn {\n json: trainingConfig\n};"
},
"id": "create-training-config-005",
"name": "Create Training Configuration",
"type": "n8n-nodes-base.code",
"typeVersion": 2,
"position": [
1120,
240
]
},
{
"parameters": {
"method": "POST",
"url": "http://ollama:11434/api/fine-tune",
"sendHeaders": true,
"headerParameters": {
"parameters": [
{
"name": "Content-Type",
"value": "application/json"
}
]
},
"sendBody": true,
"bodyContentType": "json",
"jsonBody": "={{ JSON.stringify($json) }}",
"options": {
"timeout": 7200000
}
},
"id": "start-finetuning-006",
"name": "Start Enhanced Fine-tuning",
"type": "n8n-nodes-base.httpRequest",
"typeVersion": 4.2,
"position": [
1340,
240
],
"continueOnFail": true,
"retryOnFail": true,
"maxTries": 2
},
{
"parameters": {
"operation": "executeQuery",
"query": "INSERT INTO training_jobs (job_id, dataset_id, base_model, training_config, job_status, started_at, training_samples, validation_samples, languages) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9)",
"options": {
"queryParameters": "={{ $('Create Training Configuration').item.json.job_metadata.job_id }},{{ $('Create Training Configuration').item.json.dataset_id }},{{ $('Create Training Configuration').item.json.model }},{{ JSON.stringify($('Create Training Configuration').item.json) }},started,{{ new Date().toISOString() }},{{ $('Prepare Enhanced Dataset').item.json.dataset.train.length }},{{ $('Prepare Enhanced Dataset').item.json.dataset.validation.length }},{{ JSON.stringify($('Prepare Enhanced Dataset').item.json.dataset.metadata.languages) }}"
}
},
"id": "log-training-job-007",
"name": "Log Training Job",
"type": "n8n-nodes-base.postgres",
"typeVersion": 2.5,
"position": [
1560,
240
],
"credentials": {
"postgres": {
"name": "<your credential>"
}
}
},
{
"parameters": {
"method": "POST",
"url": "={{ $vars.SLACK_WEBHOOK_URL }}",
"sendHeaders": true,
"headerParameters": {
"parameters": [
{
"name": "Content-Type",
"value": "application/json"
}
]
},
"sendBody": true,
"bodyContentType": "json",
"jsonBody": "={\n \"text\": \"\ud83e\udd16 Enhanced Fine-tuning Started\",\n \"blocks\": [\n {\n \"type\": \"header\",\n \"text\": {\n \"type\": \"plain_text\",\n \"text\": \"Voice AI Model Training\"\n }\n },\n {\n \"type\": \"section\",\n \"text\": {\n \"type\": \"mrkdwn\",\n \"text\": \"*Enhanced Fine-tuning Job Started*\\n\\n*Job ID:* {{ $('Create Training Configuration').item.json.job_metadata.job_id }}\\n*Base Model:* {{ $('Create Training Configuration').item.json.model }}\\n*Training Samples:* {{ $('Prepare Enhanced Dataset').item.json.dataset.train.length }}\\n*Validation Samples:* {{ $('Prepare Enhanced Dataset').item.json.dataset.validation.length }}\\n*Languages:* {{ $('Prepare Enhanced Dataset').item.json.dataset.metadata.languages.join(', ') }}\\n*Data Quality Rate:* {{ $('Prepare Enhanced Dataset').item.json.dataset.metadata.quality.filter_rate }}%\\n*Started:* {{ new Date().toISOString() }}\"\n }\n },\n {\n \"type\": \"section\",\n \"text\": {\n \"type\": \"mrkdwn\",\n \"text\": \"*Training Configuration:*\\n\u2022 Learning Rate: {{ $('Create Training Configuration').item.json.hyperparameters.learning_rate }}\\n\u2022 Batch Size: {{ $('Create Training Configuration').item.json.hyperparameters.batch_size }}\\n\u2022 Epochs: {{ $('Create Training Configuration').item.json.hyperparameters.epochs }}\\n\u2022 Context Length: {{ $('Create Training Configuration').item.json.model_config.context_length }}\"\n }\n }\n ]\n}"
},
"id": "notify-training-start-008",
"name": "Notify Training Start",
"type": "n8n-nodes-base.httpRequest",
"typeVersion": 4.2,
"position": [
1780,
240
],
"continueOnFail": true
},
{
"parameters": {
"method": "POST",
"url": "={{ $vars.SLACK_WEBHOOK_URL }}",
"sendHeaders": true,
"headerParameters": {
"parameters": [
{
"name": "Content-Type",
"value": "application/json"
}
]
},
"sendBody": true,
"bodyContentType": "json",
"jsonBody": "={\n \"text\": \"\u26a0\ufe0f Insufficient training data\",\n \"blocks\": [\n {\n \"type\": \"section\",\n \"text\": {\n \"type\": \"mrkdwn\",\n \"text\": \"*Fine-tuning Skipped - Insufficient Data*\\n\\n*Available Samples:* {{ $('Prepare Enhanced Dataset').item.json.dataset.train.length }}\\n*Required Minimum:* 50\\n*Quality Filter Rate:* {{ $('Prepare Enhanced Dataset').item.json.dataset.metadata.quality.filter_rate }}%\\n*Raw Interactions:* {{ $('Prepare Enhanced Dataset').item.json.dataset.metadata.quality.total_raw }}\\n\\n*Recommendations:*\\n\u2022 Collect more user feedback\\n\u2022 Improve interaction quality\\n\u2022 Review feedback collection process\\n\\n*Next Check:* Next Sunday at 2 AM\"\n }\n }\n ]\n}"
},
"id": "notify-insufficient-data-009",
"name": "Notify Insufficient Data",
"type": "n8n-nodes-base.httpRequest",
"typeVersion": 4.2,
"position": [
1120,
400
],
"continueOnFail": true
},
{
"parameters": {
"rule": {
"interval": [
{
"field": "minutes",
"minutesInterval": 30
}
]
}
},
"id": "training-monitor-trigger-010",
"name": "Training Monitor Trigger",
"type": "n8n-nodes-base.scheduleTrigger",
"typeVersion": 1.2,
"position": [
240,
600
]
},
{
"parameters": {
"operation": "executeQuery",
"query": "SELECT * FROM training_jobs WHERE job_status IN ('started', 'running') AND started_at >= NOW() - INTERVAL '24 hours';",
"options": {}
},
"id": "check-active-jobs-011",
"name": "Check Active Training Jobs",
"type": "n8n-nodes-base.postgres",
"typeVersion": 2.5,
"position": [
460,
600
],
"credentials": {
"postgres": {
"name": "<your credential>"
}
}
},
{
"parameters": {
"language": "javaScript",
"jsCode": "// Monitor training job progress\nconst activeJobs = $input.all().map(item => item.json);\n\nif (activeJobs.length === 0) {\n return {\n json: {\n hasActiveJobs: false,\n message: 'No active training jobs to monitor'\n }\n };\n}\n\nconst jobUpdates = [];\n\nfor (const job of activeJobs) {\n try {\n // Check job status via Ollama API\n const response = await fetch(`http://ollama:11434/api/fine-tune/status/${job.job_id}`);\n const status = await response.json();\n \n jobUpdates.push({\n job_id: job.job_id,\n current_status: status.status,\n progress: status.progress || 0,\n metrics: status.metrics || {},\n last_updated: new Date().toISOString(),\n needs_update: status.status !== job.job_status\n });\n } catch (error) {\n console.log(`Error checking job ${job.job_id}:`, error.message);\n jobUpdates.push({\n job_id: job.job_id,\n current_status: 'error',\n error: error.message,\n last_updated: new Date().toISOString(),\n needs_update: true\n });\n }\n}\n\nreturn {\n json: {\n hasActiveJobs: true,\n jobUpdates,\n completedJobs: jobUpdates.filter(j => j.current_status === 'completed'),\n failedJobs: jobUpdates.filter(j => j.current_status === 'failed' || j.current_status === 'error')\n }\n};"
},
"id": "monitor-job-progress-012",
"name": "Monitor Job Progress",
"type": "n8n-nodes-base.code",
"typeVersion": 2,
"position": [
680,
600
]
},
{
"parameters": {
"conditions": {
"boolean": [
{
"value1": "={{ $json.hasActiveJobs }}",
"operation": "equal",
"value2": true
}
]
}
},
"id": "check-has-active-jobs-013",
"name": "Check Has Active Jobs",
"type": "n8n-nodes-base.if",
"typeVersion": 1,
"position": [
900,
600
]
},
{
"parameters": {
"operation": "executeQuery",
"query": "UPDATE training_jobs SET job_status = $1, progress = $2, metrics = $3, updated_at = $4 WHERE job_id = $5",
"options": {
"queryParameters": "={{ $item(0).json.current_status }},{{ $item(0).json.progress || 0 }},{{ JSON.stringify($item(0).json.metrics || {}) }},{{ new Date().toISOString() }},{{ $item(0).json.job_id }}"
}
},
"id": "update-job-status-014",
"name": "Update Job Status",
"type": "n8n-nodes-base.postgres",
"typeVersion": 2.5,
"position": [
1120,
600
],
"credentials": {
"postgres": {
"name": "<your credential>"
}
}
}
],
"connections": {
"Weekly Learning Trigger": {
"main": [
[
{
"node": "Collect Training Data",
"type": "main",
"index": 0
}
]
]
},
"Collect Training Data": {
"main": [
[
{
"node": "Prepare Enhanced Dataset",
"type": "main",
"index": 0
}
]
]
},
"Prepare Enhanced Dataset": {
"main": [
[
{
"node": "Check Training Threshold",
"type": "main",
"index": 0
}
]
]
},
"Check Training Threshold": {
"main": [
[
{
"node": "Create Training Configuration",
"type": "main",
"index": 0
}
],
[
{
"node": "Notify Insufficient Data",
"type": "main",
"index": 0
}
]
]
},
"Create Training Configuration": {
"main": [
[
{
"node": "Start Enhanced Fine-tuning",
"type": "main",
"index": 0
}
]
]
},
"Start Enhanced Fine-tuning": {
"main": [
[
{
"node": "Log Training Job",
"type": "main",
"index": 0
}
]
]
},
"Log Training Job": {
"main": [
[
{
"node": "Notify Training Start",
"type": "main",
"index": 0
}
]
]
},
"Training Monitor Trigger": {
"main": [
[
{
"node": "Check Active Training Jobs",
"type": "main",
"index": 0
}
]
]
},
"Check Active Training Jobs": {
"main": [
[
{
"node": "Monitor Job Progress",
"type": "main",
"index": 0
}
]
]
},
"Monitor Job Progress": {
"main": [
[
{
"node": "Check Has Active Jobs",
"type": "main",
"index": 0
}
]
]
},
"Check Has Active Jobs": {
"main": [
[
{
"node": "Update Job Status",
"type": "main",
"index": 0
}
]
]
}
},
"settings": {
"executionOrder": "v1",
"saveManualExecutions": true
},
"staticData": null,
"tags": [
"learning",
"fine-tuning",
"enhanced",
"monitoring"
],
"triggerCount": 2,
"updatedAt": "2025-06-27T00:00:00.000Z",
"versionId": "2.0"
}
Credentials you'll need
Each integration node will prompt for credentials when you import. We strip credential IDs before publishing — you'll add your own.
postgres
Pro
For the full experience including quality scoring and batch install features for each workflow upgrade to Pro
About this workflow
Continuous Learning Pipeline. Uses postgres, httpRequest. Scheduled trigger; 14 nodes.
Source: https://github.com/161sam/n8n-voice-ai/blob/main/workflows/Continuous-Learning-Pipeline.json — original creator credit. Request a take-down →
Related workflows
Workflows that share integrations, category, or trigger type with this one. All free to copy and import.
Disparador 1.8. Uses itemLists, postgres, emailSend, httpRequest. Scheduled trigger; 85 nodes.
Item Lists, Postgres, Email Send +1
공유회_알림톡_크론. Uses postgres, httpRequest, n8n-nodes-solapi. Scheduled trigger; 39 nodes.
Postgres, HTTP Request, N8N Nodes Solapi
QuepasaAutomatic. Uses postgres, postgresTrigger, httpRequest. Scheduled trigger; 39 nodes.
Postgres, Postgres Trigger, HTTP Request
QuepasaAutomatic. Uses postgres, postgresTrigger, httpRequest. Scheduled trigger; 39 nodes.
Postgres, Postgres Trigger, HTTP Request
QuepasaAutomatic. Uses postgres, postgresTrigger, httpRequest. Scheduled trigger; 39 nodes.
Postgres, Postgres Trigger, HTTP Request