Docker Agent Ask Docs

Put documents in ./docs, then ask questions about them.

Docker Agent Ask Docs is a minimal, configuration-based RAG project that uses Docker Agent's native RAG toolset to answer questions from documents stored in a local folder.

Project structure

docker-agent-ask-docs/
├── README.md
├── .gitignore
├── docker-agent.yaml
├── docs/
│   └── example.md
└── rag/
    └── .gitkeep

How it works

1. Put Markdown, text, or other supported documents in docs/.

2. Start Docker Agent:

   docker agent run docker-agent.yaml

3. Ask a question about the documents:

   What does this document explain?
   

Docker Agent uses the configured RAG tool to index and search the files in docs/, then answers using the relevant retrieved content. The chunked embeddings database is generated locally at rag/chunked_embeddings.db; it is ignored by Git.

Core principle

Answer only from the provided documents.

The assistant is instructed not to use general knowledge or invent information. When the documents do not contain enough information, it says that the answer was not found in the provided documents. Retrieved sources are included as links to local files at the end of the response.

What is included

• Docker Agent configuration
• Local document folder
• Docker Agent's built-in RAG toolset
• Chunked embeddings strategy
• Local embeddings database
• Source links in responses
• Instructions to avoid unsupported answers

What is intentionally excluded

• Python application code
• Custom API server
• Web UI
• LangChain or LlamaIndex
• Separate vector database service
• Custom ingestion pipeline
• Multi-agent orchestration
• Internet search

Add your own documents

Copy documents into docs/ and restart Docker Agent:

cp my-document.md ./docs/
docker agent run docker-agent.yaml

The first run creates the local embeddings database. Do not commit that generated database or credentials; both are covered by .gitignore.