Development of an AI-Powered Customer Support
Budget: $25 – $50 USD
I need a production-ready AI chatbot that can read our internal PDFs, FAQs, and policy documents, then serve website visitors with accurate, context-aware answers. The bot must return natural-language text and, whenever relevant, include direct links to the exact page or section in those source files so users can verify the information themselves.
Scope
• Query coverage: general information, technical support, and account issues drawn from the same knowledge base—no hand-written intents.
• Response formats: text plus document links only; no automated account actions for now.
• Single deployment target: our public website (JavaScript widget or REST endpoint we can embed).
Core expectations
– Document ingestion pipeline that handles new/updated PDFs automatically.
– Semantic search or vector retrieval to ensure answers stay strictly within our content.
– Clear fallback logic for edge-cases and a confidence score so we know when to escalate to a human agent.
– Admin dashboard or simple CLI for me to retrain, add documents, and monitor usage.
Deliverables
1. Source code (Python/Node preferred) with README and setup script.
2. Deployed web widget/API integrated in our staging site.
3. Short video or live walkthrough showing ingestion, querying, and escalation flow.
4. Post-launch support for initial tuning (two weeks).
I value clean, documented code and transparent model choices—feel free to suggest OpenAI, LangChain, or a custom transformer if it best fits the job. Let’s create a chatbot that finally puts our scattered documentation to work.
Scope
• Query coverage: general information, technical support, and account issues drawn from the same knowledge base—no hand-written intents.
• Response formats: text plus document links only; no automated account actions for now.
• Single deployment target: our public website (JavaScript widget or REST endpoint we can embed).
Core expectations
– Document ingestion pipeline that handles new/updated PDFs automatically.
– Semantic search or vector retrieval to ensure answers stay strictly within our content.
– Clear fallback logic for edge-cases and a confidence score so we know when to escalate to a human agent.
– Admin dashboard or simple CLI for me to retrain, add documents, and monitor usage.
Deliverables
1. Source code (Python/Node preferred) with README and setup script.
2. Deployed web widget/API integrated in our staging site.
3. Short video or live walkthrough showing ingestion, querying, and escalation flow.
4. Post-launch support for initial tuning (two weeks).
I value clean, documented code and transparent model choices—feel free to suggest OpenAI, LangChain, or a custom transformer if it best fits the job. Let’s create a chatbot that finally puts our scattered documentation to work.