Claude Code: Build Enterprise RAG and Knowledge Graph Apps

Design an end-to-end system architecture for a knowledge app, detailing login/signup with JWT, document upload to a knowledge graph and vector store, and retrieval with embedding, cosine search, and re-ranking.

Design a production-ready knowledge graph by defining ontology from competency questions, ingesting data into Neo4j, and using LLMs and DSL to generate Cypher queries.

Set up the code base for the decision intelligence project by creating a virtual environment and a modular backend structure (api, services, repository, utils, schema, db) with src and tests.

Create the cloud.md file by manual setup or using the slash init command, and provide a well-structured readme so cloud accurately infers project name, dependencies, and architecture.

Explore spec driven development and pig driven development, detailing per feature spec files containing problem statements, functional requirements, api controls, constraints, edge cases, and acceptance criteria.

Learn how Claude code skills turn a general-purpose model into specialist agents using reusable file-based skills, and how sub-agents provide context isolation, modularity, and parallelism for tasks like codebase exploration.

Explore cloud code driven implementation of a large IT service consulting revenue intelligence platform, from product planning and user stories to design, spec files, testing, review, and e2e validation.

Create the signup page user interface using cloud code, organizing with dot plot folders for rules, skills, and specs; define auth skill.md and ui design rules to guide authentication interfaces.

Implement a signup page UI with Claude Code by following the spec and UI design docs, wiring post /api/signup, configuring frontend routing, and planning feature-based spec files for scalable tests.

Create the login page by generating its spec and test plans under specs, implement the UI per design guidelines, and validate with unit, integration, and security tests.

Implement the post /api/signup user signup API within a cloud code workflow, wiring the auth flow to MongoDB, Redis, and JWT, with username validation, password hashing, and test specs.

Test the authentication workflow by running the frontend on port 8080 and the backend on port 8000, and store the JWT as a cookie.

Run signup flow unit and end-to-end tests, fix failures until all tests pass, and generate a review report covering bugs, security issues, architecture violations, error handling, logging, and monitoring.

Implement a MongoDB-backed login API that authenticates users, returns a JWT token, stores it in browser session storage, and redirects to the upload page; ensure token naming consistency with signup.

Design a document upload page with a python ui, including base64 encoding, and an etl pipeline that processes PDFs and CSV/XLSX by chunking and embedding into a vector database.

Analyze the upload page UI from mockups or screenshots, generate unit test cases with cloud code, and plan selective e2e tests once the API is ready.

implement a post /api/upload endpoint that accepts a list of base64-encoded files, converts each to binary, determines file type, and routes pdf content extraction via pymu pdf and csv/xlsx parsing.

Build the pdf chunking pipeline in the etl workflow, chunking extracted pdfs into approximately 500-token parents and approximately 100-token children, storing parents in MongoDB and embeddings in a vector database.

Build the csv chunking pipeline by converting each csv row to key-value text, grouping every 10 rows, generating OpenAI embeddings, and storing in-memory chroma DB via the chunking service.

Define domain and competency questions; design the ontology. Ingest data with llms to produce structured json, validate and resolve entities, and write to Neo4j, then retrieve with Cypher queries.

Install and start Neo4j desktop, create instance, and seed a knowledge graph ontology. Design the ontology in seed files using competency questions to generate ontology_seed.py with the Neo4j Python driver.

Build a knowledge graph ingestion workflow for pdf and csv/xls data, applying ontology context to extract and validate entities and relationships, and persist to neo4j via a fastapi app.

Debug and resolve knowledge graph issues by enabling csv and pdf workflows, uploading data, validating records in Neo4j, and applying cloud code fixes for token errors.

Update and run end-to-end tests for the upload endpoint, validate rag and knowledge graph pipelines with CSV and PDF data, and fix encoding and database routing issues.

Design a modern, responsive chatbot UI using Python 9's UI, guided by the specs file and cloud code, accessible at /chatbot and using HTTP POST (no WebSocket) to /chat.

The lecture designs a hybrid retrieval workflow that fetches context from the knowledge graph and vector database using an LLM router with KG and rack sub-agents, then synthesizes the answer.

Explore building a multi-agent workflow in LangGraph with a planner, vector, and KG agents, using tools to retrieve data from Chroma DB and a knowledge graph.

Create skeleton spec files for the retrieval workflow, kg pipeline, and rack pipeline, guided by a planner agent that routes queries to sub agents and the synthesis agent.

This lecture shows creating a spec file and unit tests for a retrieval workflow, adding a chat API endpoint, and clarifying out-of-scope rag and knowledge graph implementations.

learn to implement a knowledge graph retrieval and query pipeline, from llm semantic parsing to cipher query execution in neo4j, including graph search, dsl, ontology validation, and end-to-end testing.

Implement a vector rag retrieval agent that embeds queries with OpenAI, performs semantic search over chroma db child chunks, deduplicates parent chunks, and returns a combined context.