Profiling and Recommending Workflow

This example contains two agents and two tasks:

ProfilerAgent: It is used to create/update the profile of the user
RecommenderAgent: It is used to recommend the title of the article to the user
GenerateUpdatePersonaTask: It is used to generate/update the profile of the user
RecommendNewsArticlesTask: It is used to recommend the title of the article to the user

Workflow Details

Input Schema:
- user_ppid: User's unique identifier
- articles_read: Array of read articles Required fields: Both fields are mandatory
Tools:
- get_user_from_ppid: User data retrieval system
- list_user_docs: User document listing system
- create_user_doc: User document creation system
Main Steps:
1. User Data Retrieval
  - Fetch user information by PPID
  - Sort by recent activity
2. Document Analysis
  - Retrieve existing persona documents
  - Extract current persona data
3. Persona Generation
  - Analyze demographics
  - Evaluate psychographics
  - Map content interests
  - Assess sports preferences
4. Document Creation
  - Generate embedding instruction
  - Create new persona document
  - Store with proper formatting

Perfect for platforms requiring sophisticated user profiling and content personalization.

Input Schema:
- user_ppid: User's unique identifier
- mmr_strength: MMR (Maximal Marginal Relevance) strength for content diversity
- agent_id: Agent identifier for document search Required fields: All fields are mandatory
Tools:
- get_user_from_ppid: Retrieves user information from PPID
- get_user_docs: Fetches user documents and persona
- search_agent_docs: Searches and ranks agent documents using embedding similarity
Input Schema:
- user_ppid: The user's unique identifier
- mmr_strength: MMR (Maximal Marginal Relevance) strength for content diversity
- agent_id: Agent identifier for document search
Tools:
- get_user_from_ppid: Retrieves user information from PPID
- get_user_docs: Fetches user documents and persona
- search_agent_docs: Searches and ranks agent documents using embedding similarity
Main Steps:
1. User Data Retrieval
  - Fetch latest user information using PPID
  - Retrieve most recent user documents
2. Persona Analysis
  - Extract user embeddings and persona
  - Use embeddings for content similarity ranking
3. Content Selection
  - Search and rank documents using MMR
  - Select top 50 articles for consideration
4. Article Diversification
  - Rank articles based on user persona
  - Select top 5 articles with sport diversity
5. Newsletter Generation
  - Create personalized newsletter title
  - Format final article selection

This workflow is ideal for personalized news recommendation systems requiring: