InferShield Manual Integration Guide (v0.9.0)

Use InferShield to scan code/prompts for PII and security threats before they reach your AI assistant.

Quick Start (5 minutes)

1. Get Your API Key

Go to: http://localhost:8080/dashboard.html (or your InferShield URL)
Navigate to: API Keys section
Click: CREATE NEW KEY
Name: "Manual Testing"
Copy the full key (format: isk_live_...)

2. Set Environment Variable

# Add to ~/.bashrc or ~/.zshrc
export INFERSHIELD_API_KEY="isk_live_your_key_here"
export INFERSHIELD_ENDPOINT="http://localhost:5000"  # or http://192.168.1.61:5000

# Reload shell
source ~/.bashrc  # or source ~/.zshrc

3. Test It

# Test PII detection
curl -X POST $INFERSHIELD_ENDPOINT/api/analyze \
  -H "X-API-Key: $INFERSHIELD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "const apiKey = \"sk-1234567890abcdef\"; const email = \"user@example.com\";",
    "agent_id": "manual-test"
  }' | jq '.'

Expected output:

{
  "success": true,
  "threat_detected": true,
  "risk_score": 85,
  "threats": [
    {
      "type": "pii",
      "severity": "high",
      "pattern": "api_key",
      "matched_text": "sk-1234567890abcdef"
    },
    {
      "type": "pii",
      "severity": "medium",
      "pattern": "email",
      "matched_text": "user@example.com"
    }
  ],
  "redacted_prompt": "const apiKey = \"[REDACTED_API_KEY]\"; const email = \"[REDACTED_EMAIL]\";"
}

Integration Methods

Method 1: Command-Line Scanner

Create a scanner script:

# ~/bin/infershield-scan
#!/bin/bash

if [ -z "$INFERSHIELD_API_KEY" ]; then
  echo "❌ Error: INFERSHIELD_API_KEY not set"
  exit 1
fi

FILE="${1:-/dev/stdin}"
PROMPT=$(cat "$FILE")

RESULT=$(curl -s "${INFERSHIELD_ENDPOINT:-http://localhost:5000}/api/analyze" \
  -H "X-API-Key: $INFERSHIELD_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{\"prompt\":$(echo "$PROMPT" | jq -Rs .),\"agent_id\":\"cli\"}")

THREAT=$(echo "$RESULT" | jq -r '.threat_detected')
RISK=$(echo "$RESULT" | jq -r '.risk_score')

if [ "$THREAT" = "true" ]; then
  echo "⚠️  THREAT DETECTED (Risk: $RISK/100)"
  echo "$RESULT" | jq -r '.threats[] | "  - \(.type): \(.pattern) (\(.severity))"'
  exit 1
else
  echo "✅ No threats detected (Risk: $RISK/100)"
  exit 0
fi

Make it executable:

chmod +x ~/bin/infershield-scan

Usage:

# Scan a file
infershield-scan mycode.js

# Scan from stdin
echo "const password = '123456';" | infershield-scan

# Scan before committing
git diff HEAD | infershield-scan

Method 2: Git Pre-Commit Hook

Automatically scan staged changes before every commit.

# .git/hooks/pre-commit
#!/bin/bash

echo "🛡️  InferShield: Scanning staged changes..."

# Get all staged files
STAGED_FILES=$(git diff --cached --name-only --diff-filter=ACM)

if [ -z "$STAGED_FILES" ]; then
  echo "No files to scan."
  exit 0
fi

# Scan each file
for FILE in $STAGED_FILES; do
  echo "Scanning: $FILE"
  
  CONTENT=$(git show ":$FILE")
  
  RESULT=$(curl -s "${INFERSHIELD_ENDPOINT:-http://localhost:5000}/api/analyze" \
    -H "X-API-Key: $INFERSHIELD_API_KEY" \
    -H "Content-Type: application/json" \
    -d "{\"prompt\":$(echo "$CONTENT" | jq -Rs .),\"agent_id\":\"git-hook\",\"metadata\":{\"file\":\"$FILE\"}}")
  
  THREAT=$(echo "$RESULT" | jq -r '.threat_detected')
  RISK=$(echo "$RESULT" | jq -r '.risk_score')
  
  if [ "$THREAT" = "true" ]; then
    echo "❌ THREAT DETECTED in $FILE (Risk: $RISK/100)"
    echo "$RESULT" | jq -r '.threats[] | "   - \(.type): \(.pattern) (\(.severity))"'
    echo ""
    echo "Commit blocked. Remove sensitive data and try again."
    echo "Or use: git commit --no-verify (not recommended)"
    exit 1
  else
    echo "✅ $FILE is clean (Risk: $RISK/100)"
  fi
done

echo "✅ All files passed InferShield scan!"
exit 0

Make it executable:

chmod +x .git/hooks/pre-commit

Test it:

# Create a file with PII
echo "const apiKey = 'sk-test123';" > test.js
git add test.js
git commit -m "test"  # Should be blocked!

# Remove PII
echo "const apiKey = process.env.API_KEY;" > test.js
git add test.js
git commit -m "test"  # Should succeed!

Method 3: VS Code Tasks (Manual Scan)

Add to .vscode/tasks.json:

{
  "version": "0.9.0",
  "tasks": [
    {
      "label": "InferShield: Scan Current File",
      "type": "shell",
      "command": "curl -s $INFERSHIELD_ENDPOINT/api/analyze -H 'X-API-Key: $INFERSHIELD_API_KEY' -H 'Content-Type: application/json' -d '{\"prompt\":\"'$(cat ${file} | jq -Rs .)'\",\"agent_id\":\"vscode\"}' | jq '.'",
      "problemMatcher": [],
      "presentation": {
        "reveal": "always",
        "panel": "new"
      }
    },
    {
      "label": "InferShield: Scan All Changes",
      "type": "shell",
      "command": "git diff HEAD | curl -s $INFERSHIELD_ENDPOINT/api/analyze -H 'X-API-Key: $INFERSHIELD_API_KEY' -H 'Content-Type: application/json' -d @- | jq '.'",
      "problemMatcher": []
    }
  ]
}

Usage:

Cmd+Shift+P (Mac) or Ctrl+Shift+P (Windows/Linux)
Type: "Tasks: Run Task"
Select: "InferShield: Scan Current File"

Method 4: Copilot Workflow (Manual)

Before accepting Copilot suggestions:

Copy the suggested code (don't accept yet)
Save to temp file: copilot-suggestion.txt
Scan it:
```
infershield-scan copilot-suggestion.txt
```
Review threats (if any)
Accept or reject based on results

Pro tip: Create a keyboard shortcut:

# Add to ~/.bashrc
alias scan-clipboard='pbpaste | infershield-scan'  # Mac
alias scan-clipboard='xclip -o | infershield-scan'  # Linux

Workflow:

Copilot suggests code
Copy it (Cmd+C)
Run: scan-clipboard
Accept if clean ✅

API Reference

Endpoint: `/api/analyze`

Request:

POST /api/analyze
Headers:
  X-API-Key: isk_live_...
  Content-Type: application/json
Body:
{
  "prompt": "string (required) - Code or text to analyze",
  "agent_id": "string (optional) - Identifier for tracking",
  "metadata": {
    "file": "string (optional)",
    "language": "string (optional)",
    "context": "string (optional)"
  }
}

Response:

{
  "success": true,
  "threat_detected": false,
  "risk_score": 25,
  "threats": [
    {
      "type": "pii|injection|sensitive",
      "severity": "low|medium|high|critical",
      "pattern": "email|api_key|ssn|credit_card|etc",
      "matched_text": "actual match",
      "position": { "start": 0, "end": 10 }
    }
  ],
  "redacted_prompt": "Prompt with [REDACTED_*] placeholders",
  "metadata": {
    "scanned_at": "ISO timestamp",
    "scan_duration_ms": 123
  }
}

Threat Types:

pii - Personal Identifiable Information
injection - Prompt injection attempts
sensitive - API keys, passwords, secrets
policy_violation - Custom policy rules

Severity Levels:

critical (90-100): Block immediately
high (70-89): Strong warning, likely PII
medium (40-69): Review recommended
low (0-39): Minor concern

Examples

Example 1: Scan a Python File

# test.py
import openai
openai.api_key = "sk-1234567890abcdef"  # Bad!
user_email = "john.doe@example.com"      # Bad!

# Scan it
infershield-scan test.py

Output:

⚠️  THREAT DETECTED (Risk: 85/100)
  - pii: api_key (high)
  - pii: email (medium)

Example 2: Scan Clipboard (Mac)

# Copy code with Cmd+C, then:
pbpaste | curl -s $INFERSHIELD_ENDPOINT/api/analyze \
  -H "X-API-Key: $INFERSHIELD_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{\"prompt\":\"$(pbpaste | jq -Rs .)\",\"agent_id\":\"clipboard\"}" \
  | jq '.threat_detected, .risk_score, .threats[].pattern'

Example 3: Batch Scan Directory

# Scan all .js files in src/
find src/ -name "*.js" -exec bash -c '
  echo "Scanning: $0"
  infershield-scan "$0" || echo "FAILED: $0"
' {} \;

Troubleshooting

Error: "401 Unauthorized"

Check: echo $INFERSHIELD_API_KEY (is it set?)
Verify: API key is active in dashboard
Test: curl -H "X-API-Key: $INFERSHIELD_API_KEY" $INFERSHIELD_ENDPOINT/api/usage/current

Error: "Connection refused"

Check: Is InferShield running? curl $INFERSHIELD_ENDPOINT/health
Network: Are you using the right IP? (localhost vs 192.168.x.x)
Firewall: Port 5000 open?

Error: "Quota exceeded"

Check usage: curl -H "X-API-Key: $INFERSHIELD_API_KEY" $INFERSHIELD_ENDPOINT/api/usage/current
Upgrade plan or wait for reset (free tier: 100 req/month)

Next Steps

Current Limitations (v0.9.0):

Single-instance deployment (no distributed state)
In-memory session state (no Redis)
No multi-session correlation
Rule-based detection (no ML models)

Planned next:

Redis-backed distributed sessions
Multi-session correlation + stronger org-wide policy enforcement

For now:

Use git hooks for commit-time protection
Use CLI tool for manual scans
Build it into your workflow

Support

Docs: https://infershield.io/docs
API Status: curl $INFERSHIELD_ENDPOINT/health
Usage: curl -H "X-API-Key: $INFERSHIELD_API_KEY" $INFERSHIELD_ENDPOINT/api/usage/current
GitHub: https://github.com/InferShield/infershield

Made with 🛡️ by HoZyne Inc

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

InferShield Manual Integration Guide (v0.9.0)

Quick Start (5 minutes)

1. Get Your API Key

2. Set Environment Variable

3. Test It

Integration Methods

Method 1: Command-Line Scanner

Method 2: Git Pre-Commit Hook

Method 3: VS Code Tasks (Manual Scan)

Method 4: Copilot Workflow (Manual)

API Reference

Endpoint: `/api/analyze`

Examples

Example 1: Scan a Python File

Example 2: Scan Clipboard (Mac)

Example 3: Batch Scan Directory

Troubleshooting

Error: "401 Unauthorized"

Error: "Connection refused"

Error: "Quota exceeded"

Next Steps

Support

FilesExpand file tree

MANUAL_INTEGRATION.md

Latest commit

History

MANUAL_INTEGRATION.md

File metadata and controls

InferShield Manual Integration Guide (v0.9.0)

Quick Start (5 minutes)

1. Get Your API Key

2. Set Environment Variable

3. Test It

Integration Methods

Method 1: Command-Line Scanner

Method 2: Git Pre-Commit Hook

Method 3: VS Code Tasks (Manual Scan)

Method 4: Copilot Workflow (Manual)

API Reference

Endpoint: /api/analyze

Examples

Example 1: Scan a Python File

Example 2: Scan Clipboard (Mac)

Example 3: Batch Scan Directory

Troubleshooting

Error: "401 Unauthorized"

Error: "Connection refused"

Error: "Quota exceeded"

Next Steps

Support

Endpoint: `/api/analyze`