Project Description
We are ready to add a Retrieval-Augmented Generation chatbot to our WordPress site so visitors can request fresh blog-post content on the fly. The goal is simple: when a user asks a question, the bot should generate a clear summary and then point them to the most relevant article link on our site (or an authoritative external document source if nothing in our library matches).
We are open to whichever stack you prefer—OpenAI / ChatGPT, DeepAI, Perplexity, Anthropic, or a comparable LLM—provided you can prove hands-on experience building RAG workflows. Cost-effectiveness matters, so factor in token usage, caching, or lightweight vector stores such as FAISS, Pinecone, or similar to keep the running costs low.
Key points you’ll handle
• Connect the LLM to a vector index of our existing blog archive (about 1 000 posts)
• Build or adapt a WordPress plugin/widget that captures the user prompt, calls the model, retrieves context, and returns the summary + link output in a friendly chat UI
• Train or fine-tune only if it lowers cost; otherwise rely on clever retrieval and prompt engineering
• Ensure GDPR-compliant logging and give me a simple dashboard to monitor usage
• Test for hallucinations and add a fallback “no relevant article found” response
Acceptance criteria
1. User types a query in the chat window and receives, within 5 s, a summary no longer than 150 words plus at least one correctly matched article link.
2. All articles suggested exist and open in a new tab without 404 errors.
3. Monthly cost projection fits within an agreed threshold before go-live.
4. Code, prompts, and environment variables are documented well enough that I can redeploy on staging in under 30 minutes.
If you already delivered a live RAG chatbot, send a quick demo link or short screen-grab in your proposal. We are ready to start as soon as I find the right fit.