User Hackathon: Module 4 Mastery Challenge (The Question Paper)

[!CAUTION] This is a Hands-Off challenge. Do not manually write code. Your role is as a Commander. You will orchestrate a high-fidelity environment for an AI agent.

---

1. Challenge Overview

Objective: Architect a "Context-Rich" agent environment that allows an AI to solve a complex, data-heavy task with zero hallucinations.

Time Limit: 4-8 hours Tooling: Any RAG-based tool (Vector stores, Local database agents, or LangChain/LlamaIndex orchestrators).

---

2. Choose Your Mission

Choose ONE of the following scenarios to execute.

Option A: Context-Rich E-commerce Recommendation Agent

Command the agent to:

• Requirement 1: Analyze a dataset of user preferences, stock inventory, and review sentiment.
• Requirement 2: Architect a "Context Filter" that only recommends items in stock and within the user's budget.
• Requirement 3: Include a "Why this was recommended" section based on high-fidelity user data.
• Goal: Show its Data Integrity and Context Filtering capabilities.

Option B: Dynamic Technical Documentation Assistant

Develop a system while entirely in the terminal:

• Requirement 1: Ingest a large documentation set (e.g., React or Next.js docs).
• Requirement 2: Filter context to only show results relevant to the current version of the library.
• Requirement 3: Set up "Conflicting Data" flags to alert the human if docs contradict the source code.
• Goal: Show its Recency Awareness and Self-Validation against the environment.

---

3. Strict Rules & Constraints

1. Metadata Requirement: Every piece of context provided to the LLM must have an associated Source ID and Confidence Score.
2. Context Window Protection: Do not allow more than 10 pieces of context per prompt. You must implement a ranker.
3. Cross-Source Verification: If the agent finds conflicting data (e.g., a README vs internal comments), it must ask the Commander for a resolution.
4. Aesthetic Provenance: Format the AI's output to clearly show which part of the answer came from which source.

---

4. Evaluation Rubric

---

5. Submission Requirements

To pass this hackathon, you must provide:

1. The Environment Spec: Your data architecture diagram (Mermaid).
2. The Proof of Data: A log showing the agent's context retriever in action.
3. The Result: The final AI response to a complex query.
4. Reflection: How did Filtering & Metadata improve the answer compared to "Normal Chat"?

[!IMPORTANT] Good luck, Commander. Architect the environment, and the truth will follow.