LangChain Docker Starter Kits

Configuration

Use an .env file for the OPENAI_API_KEY (see env.example)

• Optionally Neo4j credentials for Aura or remote database (TODO) from env file

• configure initial page loads?

• configure Q&A prompt via env-variable?

App 1 - KG + Embedding Construction from Wikipedia

• dataset: countries of the world
  • pre-seeded database
• load/add (additional) wikipedia pages
  • provide examples history / politics / EU / …
• create embeddings
• create knowledge graph
• UI: Neo4j workspace (centrally hosted or docker image)

Endpoints

• /add_page?title=pagename
• /clear

App 2 - Q&A Chatbot - Python Dev

ask questions about the dataset form App 1 use RAG for answering vector index and/or cypher-gen CypherQAChain? Streamlit App Python ? store - questions + answers + embedding + upvotes ? (conversational memory)

Endpoints

• `/answer?title=

For control plane / extension - or env-variables

• /prompt POST / GET
• /configure (temperature, top-k, ...)

App 3 - Report Generator

• Generate Reports based on user question
  • eg.. charts of the GDP of European countries
  • or PDF of the history of African Independence
  • or 5 slides presentation about Mexico
• could use additional containers for the rendering (e.g. markdown to pdf or chart-generator?)