Customer support · Production

ApertureDB builds a RAG chatbot on its own documentation to surface and close content gaps

The problem

ApertureDB's documentation was scattered across multiple tools and repositories with no semantic search, making it hard for users unfamiliar with the product's specific terminology to discover how to accomplish tasks.

First attempt

GitHub Pages documentation went out of sync with the codebase, and separate Python SDK documentation in a second repository made lookup inconvenient with nothing tying the two sources together.

Workflow diagram · grounded in source
1
Documentation crawl ingestion
integration
“we built another RAG chain that used a crawl of our marketing website and product documentation”
2
User submits natural-language query
trigger
“We asked our chatbot questions like "What command do I use for adding embeddings?"”
3
Semantic retrieval via ApertureDB
ai_action
“This LangChain-based implementation uses ApertureDB under the covers as the vectorstore / retriever for high-performance look up of documents that are semantically similar to the user's query”
4
LLM answer and reference generation
ai_action
“the answer gave a good description of how to do it, and also gave a reference to our AddDescriptor command”
5
Gap identification and doc update
feedback_loop
“Now we can look at the questions that resulted in insufficient or incorrect responses and introduce helpful and accurate information where it belongs”
Reported outcome

The RAG chatbot handles 10–50 queries per week on its alpha release, helps users discover documentation semantically without knowing ApertureDB-specific terms, and surfaces documentation gaps that the team then fills.

Reported metrics
Weekly chatbot queries (alpha)10-50 queries a week
Documentation discoverability vs keyword searchimmensely more powerful than mere keyword search
Reported stack
ApertureDBLangChainRAGLLMDocusaurusCohere
Source
https://mlops.community/blog/can-a-rag-chatbot-really-improve-content
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The RAG chatbot handles 10–50 queries per week on its alpha release, helps users discover documentation semantically without knowing ApertureDB-specific terms, and surfaces documentation gaps that the team then fills.

What tools did this team use?

ApertureDB, LangChain, RAG, LLM, Docusaurus, Cohere.

What results were reported?

Weekly chatbot queries (alpha): 10-50 queries a week; Documentation discoverability vs keyword search: immensely more powerful than mere keyword search (source-reported, not independently verified).

What failed first in this deployment?

GitHub Pages documentation went out of sync with the codebase, and separate Python SDK documentation in a second repository made lookup inconvenient with nothing tying the two sources together.

How is this customer support AI workflow structured?

Documentation crawl ingestion → User submits natural-language query → Semantic retrieval via ApertureDB → LLM answer and reference generation → Gap identification and doc update.

Related customer support cases