n8n-compose/tests/e2e-test-scenario.md

E2E Test: Full AI Support Automation Workflow

System Setup Prerequisites

Services Health Check

• Milvus Vector DB: Accessible on port 9091
• PostgreSQL: Running with KB schema and audit tables
• n8n: Workflow engine operational
• Freescout: Help desk system with custom fields configured
• LiteLLM: LLM proxy service accessible
• Baramundi: Remote execution system (optional)

Test Environment Variables

export FREESCOUT_API_KEY="your-api-key"
export LITELLM_API_KEY="your-api-key"
export N8N_AUTH_TOKEN="your-auth-token"
export POSTGRES_PASSWORD="your-password"

---

E2E Test Scenario: Complete Workflow

Test Case ID: E2E-001

Objective: Full workflow from ticket creation through knowledge base update

---

1. Setup Phase

Pre-Test Validation

# Verify all services are running
./tests/curl-test-collection.sh

# Clear test data from previous runs (optional)
docker-compose exec postgres psql -U kb_user -d n8n_kb -c \
  "DELETE FROM knowledge_base_updates WHERE created_at > NOW() - INTERVAL '1 hour';"

docker-compose exec postgres psql -U kb_user -d n8n_kb -c \
  "DELETE FROM ticket_audit WHERE ticket_id LIKE 'TEST_%';"

Test Data Preparation

• Create test customer: test@example.com in Freescout (if not exists)
• Verify custom fields exist in Freescout:
  • AI_SUGGESTION (text field)
  • AI_SUGGESTION_STATUS (enum: PENDING, APPROVED, REJECTED, EXECUTED)
  • AI_CONFIDENCE (number field)

---

2. Workflow A: Ticket Analysis & AI Suggestion

Test Step 1: Create Test Ticket in Freescout

Action: Create new ticket via Freescout UI or API

curl -X POST https://ekshelpdesk.fft-it.de/api/v1/mailboxes/1/conversations \
  -H "Authorization: Bearer $FREESCOUT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "subject": "TEST_E2E_001: Drucker funktioniert nicht",
    "body": "Jeder Druck-Befehl wird abgelehnt. Fehlercode 5. Bitte schnell lösen!",
    "customer_email": "test@example.com",
    "mailbox_id": 1
  }'

Expected Outcome:

• HTTP Status: 201 (Created)
• Response contains conversation_id
• Ticket visible in Freescout dashboard

Validation:

# Get the created ticket ID from response
TICKET_ID="<conversation_id_from_response>"
echo "Created Ticket: $TICKET_ID"

---

Test Step 2: Wait for Workflow A Cycle (5 minutes)

Process Flow:

1. n8n Workflow A triggers every 5 minutes
2. Fetches new tickets from Freescout
3. Sends ticket text to LiteLLM for analysis
4. Updates custom fields with AI suggestion

Wait Time: 5 minutes (plus 1 min buffer = 6 minutes total)

During Wait - Monitor Logs:

# Monitor n8n logs for workflow execution
docker-compose logs -f n8n | grep -i "workflow_a\|ticket\|analysis"

# Check PostgreSQL audit log
docker-compose exec postgres psql -U kb_user -d n8n_kb -c \
  "SELECT * FROM ticket_audit WHERE ticket_id = '$TICKET_ID' ORDER BY created_at DESC;"

Expected n8n Logs:

• "Processing ticket: TEST_E2E_001"
• "Calling LiteLLM API for analysis"
• "Analysis complete: category=Hardware, confidence=0.92"
• "Updated custom fields in Freescout"

---

Test Step 3: Verify AI Suggestion in Freescout

Action: Check ticket custom fields

curl -H "Authorization: Bearer $FREESCOUT_API_KEY" \
  https://ekshelpdesk.fft-it.de/api/v1/conversations/$TICKET_ID

Expected Response Fields:

{
  "conversation": {
    "id": "TICKET_ID",
    "subject": "TEST_E2E_001: Drucker funktioniert nicht",
    "custom_fields": {
      "AI_SUGGESTION": "Hardware Problem: Drucker-Treiber fehlerhaft oder beschädigt. Empfohlene Lösung: 1) Drucker neustarten, 2) Treiber neu installieren",
      "AI_SUGGESTION_STATUS": "PENDING",
      "AI_CONFIDENCE": 0.92
    }
  }
}

Validation Checklist:

• ✅ AI_SUGGESTION is not empty
• ✅ AI_SUGGESTION_STATUS = "PENDING"
• ✅ AI_CONFIDENCE between 0.7 and 1.0
• ✅ Suggestion text contains problem category and solution

Alternative: Manual Check in UI:

• Open ticket in Freescout
• Scroll to custom fields section
• Verify all three fields populated with expected values

---

3. Workflow B: Approval & Execution

Test Step 4: Approve AI Suggestion

Action: Update custom field via API

curl -X PUT https://ekshelpdesk.fft-it.de/api/v1/conversations/$TICKET_ID \
  -H "Authorization: Bearer $FREESCOUT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "custom_fields": {
      "AI_SUGGESTION_STATUS": "APPROVED"
    }
  }'

Expected Outcome:

• HTTP Status: 200 (OK)
• Custom field updated in Freescout

Validation:

# Verify field was updated
curl -H "Authorization: Bearer $FREESCOUT_API_KEY" \
  https://ekshelpdesk.fft-it.de/api/v1/conversations/$TICKET_ID | \
  jq '.conversation.custom_fields.AI_SUGGESTION_STATUS'
# Expected output: "APPROVED"

---

Test Step 5: Wait for Workflow B Cycle (2 minutes)

Process Flow:

1. n8n Workflow B triggers every 2 minutes
2. Fetches tickets with AI_SUGGESTION_STATUS = APPROVED
3. Executes action based on category:
   • Hardware/Remote: Create Baramundi job
   • Software/Config: Send support email
   • Knowledge: Update KB directly
4. Updates status to EXECUTED

Wait Time: 2 minutes (plus 1 min buffer = 3 minutes total)

During Wait - Monitor Logs:

# Check n8n workflow B execution
docker-compose logs -f n8n | grep -i "workflow_b\|approved\|executing"

# Check for Baramundi job creation (if applicable)
curl https://baramundi-api.example.com/jobs \
  -H "Authorization: Bearer $BARAMUNDI_TOKEN" | \
  jq '.jobs[] | select(.ticket_id == "'"$TICKET_ID"'")'

---

Test Step 6: Verify Execution Status

Action: Check ticket status

curl -H "Authorization: Bearer $FREESCOUT_API_KEY" \
  https://ekshelpdesk.fft-it.de/api/v1/conversations/$TICKET_ID

Expected Response:

{
  "conversation": {
    "custom_fields": {
      "AI_SUGGESTION_STATUS": "EXECUTED",
      "EXECUTION_TIMESTAMP": "2026-03-16T14:35:00Z"
    }
  }
}

Validation Checklist:

• ✅ AI_SUGGESTION_STATUS = "EXECUTED"
• ✅ EXECUTION_TIMESTAMP is recent
• ✅ No errors in n8n logs

---

4. Workflow C: Knowledge Base Update

Test Step 7: Wait for Workflow C Cycle (1 minute)

Process Flow:

1. n8n Workflow C triggers every 1 minute
2. Fetches executed tickets
3. Extracts: Problem, Solution, Category
4. Inserts into PostgreSQL knowledge_base_updates table
5. Triggers Milvus vector DB embedding and indexing

Wait Time: 1 minute (plus 1 min buffer = 2 minutes total)

---

Test Step 8: Verify Knowledge Base Entry

Action 1: Check PostgreSQL insert

docker-compose exec postgres psql -U kb_user -d n8n_kb -c \
  "SELECT id, ticket_id, problem, solution, category, frequency, created_at
   FROM knowledge_base_updates
   WHERE ticket_id = '$TICKET_ID'
   ORDER BY created_at DESC LIMIT 1;"

Expected Output:

 id | ticket_id | problem | solution | category | frequency | created_at
----+-----------+---------+----------+----------+-----------+---------------------
 42 | TEST...   | Drucker | Treiber  | Hardware | 1         | 2026-03-16 14:35:00

Action 2: Check Milvus vector DB

# Query Milvus for the new KB entry
curl -X POST http://127.0.0.1:19530/v1/search \
  -H "Content-Type: application/json" \
  -d '{
    "collection_name": "knowledge_base",
    "vectors": [
      "vector_embedding_of_problem_text"
    ],
    "top_k": 10,
    "output_fields": ["id", "ticket_id", "problem", "solution", "similarity"]
  }' | jq '.'

Expected Response:

{
  "results": [
    {
      "id": 42,
      "ticket_id": "TEST_E2E_001",
      "problem": "Drucker funktioniert nicht - Fehlercode 5",
      "solution": "Drucker neustarten und Treiber neu installieren",
      "similarity": 0.98
    }
  ]
}

Validation Checklist:

• ✅ Record exists in knowledge_base_updates table
• ✅ Fields populated: problem, solution, category, frequency
• ✅ Record appears in Milvus search results
• ✅ Similarity score >= 0.95 for exact match

---

5. Vector DB Search Test

Test Step 9: Search Similar Problem

Objective: Test that similar problems can be found via vector similarity

Action: Query Milvus with similar but different text

# Prepare embedding for similar problem
curl -X POST http://127.0.0.1:19530/v1/search \
  -H "Content-Type: application/json" \
  -d '{
    "collection_name": "knowledge_base",
    "vectors": [
      "vector_for_query: Druckerprobleme beim Ausdrucken mit Fehler"
    ],
    "top_k": 5,
    "output_fields": ["id", "ticket_id", "problem", "solution", "category"]
  }'

Expected Outcome:

• Top result is our test KB entry
• Similarity score > 0.85
• Same category returned (Hardware)

Expected Result:

{
  "results": [
    {
      "rank": 1,
      "similarity": 0.91,
      "ticket_id": "TEST_E2E_001",
      "problem": "Drucker funktioniert nicht - Fehlercode 5",
      "solution": "Drucker neustarten und Treiber neu installieren",
      "category": "Hardware"
    },
    {
      "rank": 2,
      "similarity": 0.78,
      "ticket_id": "OTHER_TICKET",
      "problem": "Netzwerkdrucker nicht erreichbar",
      "solution": "IP-Konfiguration prüfen"
    }
  ]
}

Validation:

• ✅ Test entry ranks in top 3
• ✅ Similarity > 0.85
• ✅ Relevant results returned

---

6. Failure Scenarios

Scenario F1: Service Unavailable

Test: What happens if LiteLLM is unreachable?

Action: Stop LiteLLM service

docker-compose stop litellm

Expected Behavior:

• n8n Workflow A fails gracefully
• Error logged in PostgreSQL error_log table
• Ticket status remains PENDING
• After service recovery, retry happens automatically

Verify:

docker-compose exec postgres psql -U kb_user -d n8n_kb -c \
  "SELECT error_message, retry_count FROM error_log WHERE ticket_id = '$TICKET_ID';"

Cleanup: Restart service

docker-compose up -d litellm

---

Scenario F2: Invalid Custom Field

Test: What if custom field value is invalid?

Action: Set invalid status value

curl -X PUT https://ekshelpdesk.fft-it.de/api/v1/conversations/$TICKET_ID \
  -H "Authorization: Bearer $FREESCOUT_API_KEY" \
  -d '{
    "custom_fields": {
      "AI_SUGGESTION_STATUS": "INVALID_VALUE"
    }
  }'

Expected Behavior:

• Workflow B ignores ticket
• Error logged
• Manual intervention required

---

7. Test Cleanup

Post-Test Actions

# Archive test data
docker-compose exec postgres psql -U kb_user -d n8n_kb -c \
  "UPDATE ticket_audit
   SET archived = true
   WHERE ticket_id LIKE 'TEST_%';"

# Optional: Delete test ticket from Freescout
curl -X DELETE https://ekshelpdesk.fft-it.de/api/v1/conversations/$TICKET_ID \
  -H "Authorization: Bearer $FREESCOUT_API_KEY"

# Verify cleanup
docker-compose exec postgres psql -U kb_user -d n8n_kb -c \
  "SELECT COUNT(*) as active_test_tickets FROM ticket_audit
   WHERE ticket_id LIKE 'TEST_%' AND archived = false;"

---

8. Test Metrics & Success Criteria

Timing Metrics

Phase	Expected Duration	Tolerance
Setup	-	-
Workflow A (Analysis)	5 min	±1 min
Workflow B (Execution)	2 min	±30 sec
Workflow C (KB Update)	1 min	±30 sec
Total End-to-End	~8 min	±2 min

Success Criteria

• ✅ All services respond to health checks
• ✅ Ticket created successfully in Freescout
• ✅ AI analysis completes within 5 min
• ✅ Confidence score >= 0.7
• ✅ Approval triggers execution within 2 min
• ✅ KB entry created in PostgreSQL
• ✅ KB entry indexed in Milvus
• ✅ Vector search returns relevant results (similarity > 0.85)
• ✅ All audit logs recorded correctly
• ✅ No unhandled errors in logs

Performance Targets

• API response time: < 2 seconds
• LiteLLM inference: < 10 seconds
• Milvus embedding + indexing: < 5 seconds
• PostgreSQL inserts: < 1 second

---

9. Test Execution Checklist

# Start here:
[ ] Verify all services running: ./tests/curl-test-collection.sh
[ ] Create test ticket (capture TICKET_ID)
[ ] Wait 6 minutes for Workflow A
[ ] Verify AI_SUGGESTION populated
[ ] Approve ticket (update AI_SUGGESTION_STATUS)
[ ] Wait 3 minutes for Workflow B
[ ] Verify AI_SUGGESTION_STATUS = EXECUTED
[ ] Wait 2 minutes for Workflow C
[ ] Verify PostgreSQL kb entry exists
[ ] Verify Milvus vector search works
[ ] Review all audit logs
[ ] Clean up test data
[ ] Document any issues found

---

10. Troubleshooting Guide

Issue: AI_SUGGESTION not populated after 5 min

Check:

1. n8n logs: docker-compose logs n8n | grep -i error
2. Freescout API connectivity: curl https://ekshelpdesk.fft-it.de/healthz
3. LiteLLM service: curl http://llm.eks-ai.apps.asgard.eks-lnx.fft-it.de/health
4. Custom field exists: Check Freescout admin panel

Issue: Workflow B not executing

Check:

1. Ticket status field correct: curl -H "Auth: Bearer $KEY" https://ekshelpdesk.fft-it.de/api/v1/conversations/$TICKET_ID | jq '.custom_fields.AI_SUGGESTION_STATUS'
2. n8n Workflow B enabled: Check n8n UI
3. Error logs: docker-compose exec postgres psql -U kb_user -d n8n_kb -c "SELECT * FROM error_log ORDER BY created_at DESC LIMIT 5;"

Issue: Milvus search returns no results

Check:

1. Milvus running: curl http://127.0.0.1:19530/v1/healthz
2. KB collection exists: See Milvus documentation for collection listing
3. Vector embedding generated: Check PostgreSQL knowledge_base_updates.vector_embedding

---

Appendix A: API Endpoints Reference

Service	Endpoint	Purpose
Freescout	https://ekshelpdesk.fft-it.de/api/v1/conversations	Ticket management
LiteLLM	http://llm.eks-ai.apps.asgard.eks-lnx.fft-it.de/v1/chat/completions	AI analysis
Milvus	http://127.0.0.1:19530/v1/search	Vector DB search
PostgreSQL	localhost:5432	Data persistence
n8n	http://localhost:5678	Workflow orchestration

---

Document Version: 1.0 Last Updated: 2026-03-16 Author: QA/Testing Agent Status: Ready for Testing