facebook-sources

Facebook Messenger Import#

Import your Facebook Messenger conversation history to build an AI agent that understands your communication style and past conversations. Perfect for customer support agents, personal assistants, or continuing existing relationships.

Overview#

Facebook Messenger imports allow your agent to:

  • Learn your communication style - English-Tagalog code-switching, emoji usage, tone
  • Remember past conversations - Context from months/years of chat history
  • Extract knowledge - Automatically identify Q&A pairs from conversations
  • Generate persona - Auto-create AI personality based on your style

This feature is especially powerful for Filipino businesses using mixed English-Tagalog communication.

What Gets Imported#

AlonChat processes your entire Facebook data export:

1. Conversation History#

  • All text messages
  • Timestamps and participants
  • Conversation threads
  • Storage: conversation namespace (lower priority, contextual)

2. Images (Optional, Premium)#

  • Photos shared in conversations
  • AI Vision analysis for image descriptions
  • Extracted visual context
  • Note: Requires premium plan or credits

3. Extracted Knowledge (Auto-Generated)#

  • Q&A pairs identified from conversations
  • Common questions you've answered
  • Product recommendations you've made
  • Storage: examples namespace (high priority)

4. AI Persona (Optional)#

  • Communication style analysis
  • Language mixing patterns (English-Tagalog)
  • Emoji usage frequency
  • Tone and formality preferences
  • Storage: persona namespace (always included)

How to Import Facebook Messenger Data#

Step 1: Request Your Facebook Data#

  1. Go to Facebook Settings
  2. Click Your Facebook Information
  3. Click Download Your Information
  4. Configure your download:
    • Date Range: Choose how far back (e.g., Last 1 year, All time)
    • Format: JSON (required!)
    • Media Quality: Low (faster download)
    • Sections: Check only Messages
  5. Click Create File
  6. Wait for email notification (usually 10 minutes to 1 hour)
  7. Download the .zip file

File size: Usually 10MB - 500MB depending on history length

Step 2: Upload to AlonChat#

  1. Go to your agent's Settings page
  2. Click Knowledge Base tab
  3. Click Add Facebook Import
  4. Drag and drop your Facebook .zip file
  5. Configure import options (see below)

Step 3: Configure Import Options#

Include Images (Optional, Premium)#

  • ✅ Enabled: AI Vision analyzes shared photos (costs credits)
  • ❌ Disabled: Text-only import (free)

When to enable:

  • Product photos with descriptions
  • Memes with context
  • Visual tutorials
  • Skip if: Mostly personal photos, large image count

Generate Persona (Recommended)#

  • ✅ Enabled: Auto-generate AI communication style
  • ❌ Disabled: Use default agent personality

Persona includes:

  • Language mixing patterns (e.g., 70% English, 30% Tagalog)
  • Emoji usage frequency
  • Tone preferences (formal/casual)
  • Common phrases and greetings

Extract Q&A Knowledge#

  • ✅ Enabled: Auto-extract question-answer pairs (recommended)
  • ❌ Disabled: Conversations only, no knowledge extraction

Q&A extraction finds:

  • Product inquiries with your answers
  • Common "How to..." questions
  • Pricing questions with responses
  • Troubleshooting conversations

Step 4: Wait for Processing#

Processing steps:

  1. Streaming extraction (1-5 minutes) - Unzips and reads messages
  2. Image analysis (5-30 minutes, if enabled) - AI Vision processing
  3. Q&A extraction (2-10 minutes) - Identifies knowledge pairs
  4. Persona generation (1-3 minutes) - Analyzes communication style
  5. Chunking (1-5 minutes) - Splits conversations into searchable pieces
  6. Embedding (5-20 minutes) - Generates vector embeddings

Total time: 10-60 minutes depending on file size and options

Progress tracking:

  • Real-time progress bar
  • Current step indicator
  • Processing statistics

How Facebook Data Is Used in Chat#

When a user asks a question, AlonChat retrieves relevant content from:

1. Persona (Always Included)#

Priority: 1.0 (highest) Limit: 1 chunk per agent

Your AI responds using your communication style:

Code
User: Magkano po ito?
Agent: Ay, yan 500 pesos lang 😊
       Super worth it yan promise! 💯

The agent naturally code-switches and uses emojis like you do.

2. Conversation History (Contextual)#

Priority: 0.1-0.5 (lower priority) Limit: 3-6 chunks

Provides relationship context:

  • "Remember when we discussed...?"
  • "You mentioned before that..."
  • Previous preferences and choices

3. Extracted Q&A (High Priority)#

Priority: 0.7 Limit: 2-6 chunks

Direct answers to common questions:

Code
User: Do you ship to Cebu?
Agent: Yes! We ship nationwide via LBC.
       Delivery takes 3-5 days to Cebu.

Understanding Namespaces#

AlonChat organizes Facebook data into 3 namespaces:

NamespaceContainsPriorityUse Case
personaCommunication style, language patterns1.0 (always)How the AI talks
examplesExtracted Q&A pairs0.7 (high)Product info, FAQs
conversationRaw chat history0.1-0.5 (contextual)Relationship memory

This separation ensures:

  • ✅ AI always uses your style (persona)
  • ✅ Common questions get direct answers (examples)
  • ✅ Past context available when relevant (conversation)

Re-Importing Facebook Data#

Update your agent with new conversations:

  1. Request new Facebook data export (see Step 1)
  2. Upload new .zip file to same agent
  3. AlonChat appends new messages (doesn't delete old data)
  4. Updates persona if communication style changed
  5. Extracts new Q&A pairs

Incremental update: Only new messages since last import are added.

Image Processing (Premium Feature)#

When enabled, AlonChat uses AI Vision to analyze images:

What gets analyzed:

  • Product photos → "iPhone 13 Pro Max, 256GB, Pacific Blue"
  • Screenshots → Extracted text content
  • Memes → Context and humor understanding
  • Receipts → Transaction details

Cost: ~₱0.50-2.00 per image (depends on AI model used)

Optimization: Images are deduplicated (same photo sent twice = analyzed once)

Persona Generation Details#

AlonChat analyzes 100+ messages to generate your communication style:

Language Analysis#

  • Code-switching frequency: "Uses Tagalog 30% of the time"
  • Common phrases: "Ay oo!", "Sige po", "Thank you po!"
  • Formality: Formal vs casual tone

Emoji Usage#

  • Frequency: High/Medium/Low emoji user
  • Favorite emojis: 😊 ❤️ 👍 💯
  • Placement: Start, middle, or end of messages

Tone & Style#

  • Friendliness: Warm, professional, casual
  • Response length: Short replies vs detailed explanations
  • Punctuation: Casual (no periods) vs formal

Example Generated Persona#

Code
You are a friendly customer support agent for an online store.

Communication Style:
- Mix English and Tagalog naturally (70% English, 30% Tagalog)
- Use emojis frequently, especially: 😊 ❤️ 👍
- Keep responses warm and personal
- Address customers with "po" for respect
- Use casual, conversational tone

Common Phrases:
- "Ay oo!" (agreement)
- "Sige po" (acknowledgment)
- "Thank you po!" (gratitude)
- "Pwede po yan" (affirmation)

Response Pattern:
- Start with warm greeting
- Keep answers concise (2-3 sentences)
- End with emoji to show friendliness

Best Practices#

1. Choose the Right Date Range#

  • 1 year: Good balance of recency and volume
  • All time: Best for businesses with long history
  • Last 6 months: Faster processing, recent context

2. Review Extracted Q&A#

After import:

  1. Go to Knowledge Base
  2. Review auto-extracted Q&A pairs
  3. Edit inaccurate answers
  4. Delete irrelevant pairs
  5. Add missing important questions

3. Test Persona Accuracy#

Chat with your agent to verify:

  • Does it code-switch like you?
  • Are emojis used appropriately?
  • Is the tone correct?

Adjust if needed: Edit the System Prompt in Agent Settings

4. Privacy Considerations#

  • Review conversation content before upload
  • Remove sensitive personal information
  • Delete conversations with confidential data
  • Use a test agent first for private exports

5. Combine with Other Sources#

Facebook imports work best when combined with:

  • File Sources: Product catalogs, price lists
  • Text Sources: Company policies, FAQs
  • Website Sources: Official documentation

This creates a comprehensive knowledge base.

Limitations#

  • File size: Maximum 2GB (most exports are < 500MB)
  • Format: JSON only (not HTML format)
  • Language: Optimized for English-Tagalog, works with other languages
  • Images: Premium feature, credit cost applies
  • Group chats: All conversations imported (filter by reviewing Q&A)

Troubleshooting#

"Upload failed - Invalid format"#

  • Ensure you downloaded JSON format, not HTML
  • Extract the .zip if you uploaded the extracted folder

"Processing stuck"#

  • Large files (> 500MB) take 30-60 minutes
  • Refresh page to check current status
  • Contact support if stuck > 2 hours

"Q&A extraction found nothing"#

  • Conversations may be too casual (no clear questions)
  • Try enabling persona generation only
  • Manually add Q&A pairs from conversations

"Images not processed"#

  • Verify premium plan or sufficient credits
  • Check image processing is enabled in settings
  • Some images may fail (corrupted, unsupported format)

"Persona doesn't match my style"#

  • Agent analyzed too few messages (< 50)
  • Upload more conversation history
  • Manually edit System Prompt to adjust tone

Privacy & Security#

  • ✅ Data stays private: Only you and your agent can access
  • ✅ Encrypted storage: All data encrypted at rest
  • ✅ No training: Your data NEVER used to train AI models
  • ✅ Deletable: Permanently delete anytime from Knowledge Base

Next Steps#

facebook-sources | AlonChat Docs