GPT News Recommendation Chatbot
Budget: ₹400 – ₹750 INR
A bilingual (Hindi + English) chatbot powered by the OpenAI ChatGPT API needs to sit on our website and guide visitors toward the most relevant news articles in our database. The core flow is simple: a visitor asks for information, the bot converses naturally, then surfaces article links that match the topic or sentiment the visitor expressed.
Technical framework
• Codebase may be written in either Node.js or Python—whichever you are most comfortable maintaining.
• The OpenAI key must be pulled from an environment variable; no hard-coding.
• All calls should stream responses to keep the conversation fluid.
• Include basic session memory so the bot can reference earlier messages during a single chat.
Content logic
News-article metadata is already exposed through a REST endpoint. Your task is to fetch, rank, and recommend up to five articles per query. OpenAI can be used for semantic matching or summarisation, but the final message must contain direct titles and URLs.
Deliverables
1. Complete source code with clear folder structure.
2. One-page README covering setup, environment variables, and deployment to a standard Linux host.
3. An embeddable frontend widget (plain JS or React) that loads the chat window on any page of our site.
4. Brief demo video or GIF proving Hindi and English conversations work smoothly.
Acceptance criteria
• Responses are contextually accurate and switch languages when the user does.
• Recommendations arrive in under three seconds on an average broadband connection.
• No API keys or secrets committed to the repository.
Once merged and verified on a staging URL, the project will be considered complete.
Technical framework
• Codebase may be written in either Node.js or Python—whichever you are most comfortable maintaining.
• The OpenAI key must be pulled from an environment variable; no hard-coding.
• All calls should stream responses to keep the conversation fluid.
• Include basic session memory so the bot can reference earlier messages during a single chat.
Content logic
News-article metadata is already exposed through a REST endpoint. Your task is to fetch, rank, and recommend up to five articles per query. OpenAI can be used for semantic matching or summarisation, but the final message must contain direct titles and URLs.
Deliverables
1. Complete source code with clear folder structure.
2. One-page README covering setup, environment variables, and deployment to a standard Linux host.
3. An embeddable frontend widget (plain JS or React) that loads the chat window on any page of our site.
4. Brief demo video or GIF proving Hindi and English conversations work smoothly.
Acceptance criteria
• Responses are contextually accurate and switch languages when the user does.
• Recommendations arrive in under three seconds on an average broadband connection.
• No API keys or secrets committed to the repository.
Once merged and verified on a staging URL, the project will be considered complete.