Website AI Chatbot
Budget: $1,500 – $3,000 USD
I’m looking for a freelancer to build an MVP of an AI-powered chatbot for a product-based automotive website (aftermarket parts, 10k+ SKUs).
This is a proof-of-concept, not a production system. The chatbot will be hosted as a standalone web app (e.g., Streamlit or simple web UI) on a temporary domain.
Goal
Demonstrate how an AI assistant could act as a front-door experience for website visitors by:
• Answering product-related questions
• Helping users find the correct parts for their vehicle
• Explaining differences between product lines
Core Requirements
1. Chat Interface
• Clean conversational UI
• Hosted and accessible via URL
2. LLM Integration
• Use a modern API (Claude, OpenAI, etc.)
3. Fitment Handling (Important)
• Extract Year/Make/Model from user input
• Route users to correct product listings using existing website structure (URL patterns or logic)
• Return links + clear explanations
4. Knowledge Retrieval (Lightweight RAG)
• Ingest/scrape selected website content (product/category pages)
• Store as structured or semi-structured data
• Use embeddings + vector search to retrieve relevant info
• Use retrieved context in LLM responses
Important Constraints
• Do NOT build a chatbot that relies only on raw prompting
• Do NOT scrape the website live on every query
• Retrieval must be based on preprocessed data (RAG approach)
Nice-to-Have (if feasible in MVP)
• Basic intent classification (fitment vs comparison vs general question)
• Modular structure so new data can be added later
Deliverables
• Working hosted demo (URL)
• Source code (GitHub)
• Brief setup instructions
Notes
The site already has a Year/Make/Model filtering pipeline, so the chatbot should leverage this rather than attempting to infer fitment purely via LLM.
Focus on speed, clarity, and demonstrating value—not production robustness.
This is a proof-of-concept, not a production system. The chatbot will be hosted as a standalone web app (e.g., Streamlit or simple web UI) on a temporary domain.
Goal
Demonstrate how an AI assistant could act as a front-door experience for website visitors by:
• Answering product-related questions
• Helping users find the correct parts for their vehicle
• Explaining differences between product lines
Core Requirements
1. Chat Interface
• Clean conversational UI
• Hosted and accessible via URL
2. LLM Integration
• Use a modern API (Claude, OpenAI, etc.)
3. Fitment Handling (Important)
• Extract Year/Make/Model from user input
• Route users to correct product listings using existing website structure (URL patterns or logic)
• Return links + clear explanations
4. Knowledge Retrieval (Lightweight RAG)
• Ingest/scrape selected website content (product/category pages)
• Store as structured or semi-structured data
• Use embeddings + vector search to retrieve relevant info
• Use retrieved context in LLM responses
Important Constraints
• Do NOT build a chatbot that relies only on raw prompting
• Do NOT scrape the website live on every query
• Retrieval must be based on preprocessed data (RAG approach)
Nice-to-Have (if feasible in MVP)
• Basic intent classification (fitment vs comparison vs general question)
• Modular structure so new data can be added later
Deliverables
• Working hosted demo (URL)
• Source code (GitHub)
• Brief setup instructions
Notes
The site already has a Year/Make/Model filtering pipeline, so the chatbot should leverage this rather than attempting to infer fitment purely via LLM.
Focus on speed, clarity, and demonstrating value—not production robustness.