Full-Cycle Ai-Driven Personal Assistant for Website

We are seeking a skilled freelancer or team to develop and deploy a comprehensive AI-driven personal assistant for our website. This full-cycle solution will involve conversation design, NLP integration, web widget development, backend integration, knowledge base management, analytics, and deployment.

The AI assistant should be capable of greeting visitors, answering their questions, and performing simple tasks such as booking calls, collecting feedback, and guiding users to relevant pages or documents.

Key requirements and expectations for the proposal include:

1.  Natural Language Processing (NLP): The dialogue flow must be natural and friendly, powered by modern NLP frameworks such as OpenAI, Dialogflow CX, Rasa, or a comparable solution.
2.  Web Integration: The assistant needs to be implemented as a web-based widget or an embedded chat panel that seamlessly integrates with our existing site stack, which is built on React and Node.js.
3.  Backend Connectivity: Secure webhooks or REST endpoints are required to enable the bot to trigger server-side actions and log user interactions in our database.
4.  Knowledge Base Management: The solution must provide a straightforward method for updating the knowledge base without requiring code redeployment. Acceptable formats include Markdown, Google Sheet, or JSON.
5.  Analytics: The system should include analytics capabilities to track usage, fallback rates, and user sentiment, facilitating iterative improvements to the AI model.
6.  Deployment & Documentation: We prefer deployment instructions for Docker or serverless environments, along with documented source code in Git.

Acceptance Criteria:

*  The assistant must reliably recognize at least 90% of the test intents provided.
*  All text responses must render correctly on both desktop and mobile web browsers.
*  No hard-coded secrets are allowed; environment variables and HTTPS must be used exclusively.
*  A handover protocol to human support (via email form) should be triggered when the AI's confidence in its response is low.

Proposals should outline the intended framework, showcase prior chatbot work, and provide an estimated timeline for an MVP and subsequent refinement rounds.