connect-users

name	connect-users
description	Find community members with relevant expertise using vector search
routing	{"use_when":"The user wants to FIND, MEET, or BE INTRODUCED to community members by expertise, interest, or background: \"anyone who…\", \"who knows…\", \"connect me with…\", \"looking for a mentor/teammate/cofounder in…\".\n","avoid_when":"Connecting SERVICES or integrations (github-integration). Questions about a topic itself rather than about people (\"explain X\" is chat).\n","examples":[{"text":"do you know anyone in AI research?","action":"search"},{"text":"anyone in the community working with medical imaging?","action":"search"},{"text":"connect me with someone who writes blog content","action":"search"},{"text":"my ecg project needs a teammate, anyone interested in medical AI?","action":"search"},{"text":"looking for a mentor in data engineering, any suggestions?","action":"search"}],"negative_examples":[{"text":"connect github again for mlai.au","instead":"github-integration"},{"text":"explain what a vector database is","instead":"respond_in_chat"}]}
actions	[{"name":"search","description":"Search the community for members matching an expertise/interest query.","params":{"query":{"type":"string","description":"The expertise or topic to search for."},"limit":{"type":"integer","description":"Max suggestions (default 5)."}}}]

Connect Users Skill

This skill enables Claude to find and recommend MLAI community members based on their expertise, interests, and what they're working on.

Capabilities

• Search for users with specific expertise using vector similarity
• Match users based on topics, skills, and interests
• Provide warm introductions and connection suggestions

Parameters

• query: The expertise or topic the user is looking for (required)
• exclude_user_id: Slack user ID to exclude from results (optional - usually the requester)
• limit: Maximum number of users to suggest (default: 5)

Workflow

Step 1: Extract Topics

Parse the user's query to identify the specific expertise areas they're looking for.

Common patterns to recognize:

• "who knows about X" → extract X
• "anyone working on Y" → extract Y
• "expert in Z" → extract Z
• "looking for someone in [field]" → extract field
• "recommend someone for [topic]" → extract topic

Step 2: Vector Search

Use the vector_search function to find users with matching expertise.

The search uses cosine similarity on embeddings stored in the user_expertise table.

SELECT u.id, u.name, u.slack_id, e.topic, e.relationship,
       1 - (e.embedding <=> $query_embedding) as similarity
FROM users u
JOIN user_expertise e ON u.id = e.user_id
WHERE 1 - (e.embedding <=> $query_embedding) > 0.7
ORDER BY similarity DESC
LIMIT 5;

Step 3: Format Response

Generate a warm, friendly response suggesting the matched users.

Include for each user:

• Their name (with Slack mention if available)
• Their expertise area that matched
• The relationship type (expert, working on, interested)

Response Style

• Start with an acknowledgment of what they're looking for
• Use casual Australian language occasionally (mate, no worries, legend, etc.)
• Be encouraging about making connections
• If no users found, offer alternative suggestions or ask for clarification

Example Responses

Successful Match

G'day! Looking for folks in AI research, eh? Here are some legends who might help:

• **@sam** - Expert in machine learning and neural networks
• **@jane** - Currently working on computer vision projects
• **@bob** - Interested in deep learning applications

Feel free to reach out to them! 🦘

No Matches Found

Hmm, I couldn't find anyone specifically matching "quantum computing" in our community yet.

A few things we could try:
• Broaden the search - maybe "physics" or "advanced computing"?
• Post in #introductions asking if anyone's into this space
• Check out our upcoming events - might meet someone there!

Want me to try a different search? 🤔

Error Handling

If the database search fails:

1. Apologize briefly
2. Suggest alternative ways to find help (Slack channels, events)
3. Offer to try again