Agent for Ethical, Personalized Content Recommendation
Build an ethical, personalized content summarizer and recommender system for digital subscribers using Pydantic AI. The challenge focuses on leveraging Pydantic AI's capabilities for structured output generation and schema validation to ensure summaries are factual, recommendations are relevant, and both adhere strictly to ethical guidelines (e.g., 'ad-free', 'no advertiser influence'). You will design agents that interact with a content ingestion pipeline, generate structured summaries of articles, and provide personalized recommendations based on user profiles stored in a vector database. The system must guarantee that all generated content and recommendations conform to predefined Pydantic models, reflecting Anthropic's commitment to unbiased content.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an ethical, personalized content summarizer and recommender system for digital subscribers using Pydantic AI. The challenge focuses on leveraging Pydantic AI's capabilities for structured output generation and schema validation to ensure summaries are factual, recommendations are relevant, and both adhere strictly to ethical guidelines (e.g., 'ad-free', 'no advertiser influence'). You will design agents that interact with a content ingestion pipeline, generate structured summaries of articles, and provide personalized recommendations based on user profiles stored in a vector database. The system must guarantee that all generated content and recommendations conform to predefined Pydantic models, reflecting Anthropic's commitment to unbiased content.
Shared data for this challenge
Review public datasets and any private uploads tied to your build.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
Pydantic Schema Validation
Ensures the generated summary and recommendations strictly adhere to their Pydantic schemas.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Ad-Free Content Check
Verifies that no output explicitly contains sponsored links or overt advertiser influence.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Recommendation Relevance Check
Ensures at least one recommendation is highly relevant to the user's profile and reading history.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Summary Factual Accuracy
Measures the correctness of facts presented in the summary (0-1). • target: 0.95 • range: 0.8-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Personalization Score
Quantifies how well recommendations align with user preferences and history (0-1). • target: 0.9 • range: 0.75-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Output Generation Latency
Time taken to generate both summary and recommendations (in seconds). • target: 2 • range: 0-5
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Master Pydantic AI for defining agent capabilities, structured output generation, and robust schema validation for LLM interactions.
Design comprehensive Pydantic models for content summaries (e.g., title, key points, sentiment) and personalized recommendations (e.g., article ID, relevance score, reason).
Implement a content summarization agent using Mistral-large via Pydantic AI, ensuring outputs strictly conform to the defined schemas and ethical guidelines.
Orchestrate a multi-stage data pipeline with Apache Airflow to ingest new articles, process them, generate embeddings, and update the recommendation engine.
Utilize PostgreSQL with the pgvector extension to store content embeddings and user interaction histories, enabling efficient similarity search for recommendations.
Develop personalized recommendation agents that query the pgvector database based on user profiles and generated summaries, providing relevant and unbiased suggestions.
Build an interactive web interface using Streamlit to showcase the ad-free summaries and personalized recommendations to subscribers, allowing for feedback and interaction.
How this agent runs
The evaluation will focus on the system's ability to generate structured, accurate, and ethical content summaries and provide relevant, unbiased personalized recommendations. Schema adherence, factual correctness, and...
Challenge input
{'article_text': 'string', 'user_profile': {'id': 'string', 'reading_history': ['string'], 'preferences': ['string']}}
Agent execution
The configured agent processes the input under the challenge policy.
Evaluated output
{'summary': {'title': 'string', 'key_points': ['string'], 'sentiment': 'string'}, 'recommendations': [{'article_id': 'string', 'title': 'string', '...
- Ensures the generated summary and recommendations strictly adhere to their Pydantic schemas.
- Verifies that no output explicitly contains sponsored links or overt advertiser influence.
- Ensures at least one recommendation is highly relevant to the user's profile and reading history.
- Summary Factual Accuracy target: 0.95
- 1 public reference case
- Python execution harness
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- Pydantic Schema Validation · Weight 1
- Ad-Free Content Check · Weight 1
- Recommendation Relevance Check · Weight 1
- Summary Factual Accuracy · Weight 1
- Personalization Score · Weight 1
- Output Generation Latency · Weight 1
Recipe state
This is a preview. The configuration can change before the evaluation recipe is locked.
Run this agent on your dataset and AI stack
Bring your dataset, model providers, and success criteria. We will scope the right managed run for your team.
Scope a managed run[ok] Wrote CHALLENGE.md
[ok] Wrote .versalist.json
[ok] Wrote eval/examples.json
Requires VERSALIST_API_KEY. Works with any MCP-aware editor.
DocsFind another challenge
Jump to a random challenge when you want a fresh benchmark or a different problem space.