More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Custom Data RAG Chatbot Builder

    8

    Build a full-stack AI chatbot trained on your own documents across any industry — legal, healthcare, e-commerce, HR, finance, real estate, insurance, education, cybersecurity, government, and more.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Custom Data RAG Chatbot Builder

    Custom Data RAG Chatbot Builder

    Example session with this skill installed

    Build me a full chatbot that reads my legal contract PDFs, lets clients ask questions about them, shows them which document and page the answer is from, and makes sure no client can access another client's files.

    • Read your context and instructions
    • Compiled the custom data rag
    • Generated the document

    ✅ Phase 1 Complete: Document Ingestion Pipeline
    — PDF parser configured for recursive 1000-character chunking with 200-character overlap
    — Metadata schema: { documentId, ownerId, pageNumber, filePath, category }

    ✅ Phase 2 Complete: PGVector Database Schema
    — Table: DocumentChunk (id, ownerId, content, embedding Vector(1536))
    — HNSW index on embedding column for sub-ms retrieval

    ✅ Phase 3 Complete: Hybrid Retrieval API Route (POST /api/chat)
    — Dense cosine-similarity search with ownerId tenant filter (≥ 0.78 threshold)
    — Sparse BM25 keyword fallback for exact legal clause references
    — Re-ranked top 4 chunks assembled into bounded system prompt

    ✅ Phase 4 Complete: Chat Widget Component
    — Floating launcher with slide-in animation
    — Streaming word-by-word response delivery (SSE)
    — Citation badge: [contract_nda_v2.pdf — Page 7]

    ✅ Phase 5 Complete: Security Audit Passed
    — Tenant isolation: all queries filtered by session.userId = ownerId ✓
    — Prompt injection guard: user input sanitized and sandboxed ✓
    — No hallucination fallback message implemented ✓

    contract_nda_v2.pdf

    PDF · document

    Generated

    Example file from a real run - the skill writes it into your workspace.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Stop Shipping Chatbots That Hallucinate. Start Shipping AI That Actually Knows Your Business. Most AI chatbots are generic. They answer questions based on training data that stopped in 2023, they fabricate information when they don't know the answer, and they expose sensitive documents to any user who asks the right question. The Universal Custom-Data RAG Chatbot Builder skill is different. This skill programs your AI developer assistant (Claude Code, Cursor, Windsurf) to architect and build a complete, production-ready chatbot platform that reads, indexes, and securely retrieves answers exclusively from your files — PDFs, product catalogs, legal contracts, medical guides, financial reports, internal wikis — anything. What Gets Built, End to End A document ingestion engine that parses your files into intelligent chunks using an overlapping recursive strategy that preserves semantic context, generates 1536-dimensional vector embeddings, and stores them in a Postgres database optimized with HNSW indexes capable of sub-millisecond retrieval across over 1,000,000 records. A hybrid retrieval system that runs two simultaneous search algorithms — Dense Semantic Search (understands what the user means) and Sparse Keyword Search (catches exact technical terms, product codes, legal clause IDs) — then merges both result sets through a Re-Ranking layer to surface only the highest-confidence answers. A streaming chat UI widget with a floating launcher button, animated typing bubbles, real-time text streaming word-by-word, and interactive citation badges that show users exactly which document and page number each answer was pulled from — so users can verify facts themselves. Anti-hallucination prompt constraints baked into the system-level instructions that force the model to respond only from retrieved context. If the answer is not in your documents, the chatbot says so — it never fabricates. Zero-trust tenant isolation written into every database query, making it architecturally impossible for one user's chatbot session to retrieve documents belonging to another user. Works Out of the Box in Any Niche ⚖️ Legal | Contracts, case files, compliance documents with section-level citations 🛒 E-commerce | Product catalogs, pricing tables, inventory CSV files 🏥 Healthcare | Clinical guidelines, patient FAQs with mandatory disclaimer footers 📊 Finance | Balance sheets, financial reports, tabular data with header-aware parsing 🏢 SaaS & Internal Tools | Employee handbooks, help center articles, API documentation What You Get in the Package Full Next.js App Router API route with Zod-validated payloads PGVector or Pinecone database schema with HNSW indexing configuration PDF, CSV, Markdown, and HTML ingestion scripts with overlap chunking Production Tailwind CSS React chat widget with streaming and citations Prompt injection defense layer and tenant metadata security filters .env.example with all required environment variable keys

    External AI API Required: You will need to supply your own API keys for an embedding provider (OpenAI, Cohere, or a self-hosted Ollama model) and a chat completion provider (OpenAI GPT-4, Anthropic Claude). The skill scaffolds the full integration layer; you only need to paste your credentials.

    New Codebases Only: Optimized for building the chatbot platform from scratch. Integrating into a large, pre-existing legacy codebase requires careful manual context mapping before execution.

    No GUI Cloud Console Setup: All infrastructure setup is handled through config files, environment variables, and CLI scripts. Clicking through GCP, AWS, or Supabase dashboards is outside scope.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 5 months ago

    Permissions

    Network Access
    Terminal / Shell
    Read Files
    Write Files
    Browser
    Environment Variables
    Listed5 months ago

    What's inside

    Frequently Asked Questions

    This skill provides your AI developer assistant with the specialized architecture and code templates to build a production-ready Retrieval-Augmented Generation (RAG) system that answers questions based exclusively on your uploaded documents without hallucinating.

    The builder is optimized for top-tier AI coding agents like Claude Code, Cursor, and Windsurf, enabling them to generate a complete full-stack environment including the frontend widget, vector database schema, and ingestion pipeline.

    You receive a complete Next.js application framework featuring a document ingestion engine, a hybrid search system (semantic and keyword), a streaming React chat UI with citations, and pre-configured security layers for multi-tenant isolation.

    The system is designed to handle a wide range of formats including PDFs, CSVs, Markdown, and HTML, using a recursive chunking strategy to ensure that even complex tables and legal clauses are parsed with their semantic context intact.

    Yes, the code includes a citation-based UI that displays exactly which document and page number the information was retrieved from, alongside strict system prompts that prevent the AI from fabricating answers not found in your data.

    Setup is streamlined via a provided .env.example file; once you add your API keys for your preferred LLM and vector provider (like PGVector or Pinecone), the agent can initialize the entire database and API structure automatically.

    New in LLM Ops