Build from documents
Load reviewed Markdown and text files, preserve source IDs, and inspect a lexical retrieval baseline.
AI development · EDITION 1.1.1
Build and Evaluate Your First RAG Assistant
Build an inspectable Python document assistant, study real local-model results and failures, and evaluate retrieval, citations and missing evidence.
By Sanjaay Gowda T S · For python developers building their first document assistant.
You need: Python/CLI basics · Python 3.10+Evidence: no API key · Optional Ollama + local model
Runtime and hardware details
Basic Python and command-line familiarity. Python 3.10 or later. The companion code uses the standard library.
Evidence mode runs without an API key or model. Generative mode requires Ollama, a separately installed local model, and suitable hardware. Model weights and subscriptions are not included.
Full setup and scope ↓$2.99USD · one-time purchase
Prepared with AI assistance. Independent human technical and assistive-technology review remains pending. Automated checks and agent review are not certification.
PDF + EPUB + companion project. Any applicable tax is shown at checkout. Single-reader ebook terms; companion code is MIT licensed.
LOOK INSIDE
See the pages.
Follow the evidence.
Actual pages from the 10-page free sample. Explore the introduction and Chapter 1 before choosing the complete edition.
START HERE Start with a clear scope
See the prerequisites, learning goals and setup before you begin.
Open sample
FOLLOW THE EVIDENCE Inspect evidence and source IDs
See a recorded retrieval result and trace the path from documents to a supported answer.
Open sample
TRY THE PROJECT Run your first question
Follow the command-line exercise using the fictional shop’s return policy.
Open sample
Read free sample 10 pages · PDF · No checkout required
WHAT YOU WILL LEARN
Understand each part.
Inspect every boundary.
Answer with evidence
Connect a local Ollama model and inspect recorded successes, citation failures and unsupported claims.
Evaluate the limits
Compare measured retrieval settings, trace claims to source text, and check your work against exercise solutions.
INSIDE THE BOOK
9 focused chapters.
- Trace the first question
- Prepare documents that can be cited
- Build a transparent retriever
- Make answers inspectable
- Connect a local language model
- Evaluate the whole question set
- Debug failures in the right order
- Extend deliberately
- Reference cards and next steps
Your download includes
- 38-page tagged PDF with linked contents
- Reflowable EPUB for ebook readers
- Companion Python source, fixtures and 17 behavioural tests
- Four fictional policy documents and 12 labelled evaluation cases
- Recorded retrieval reports, actual model outputs and claim-review notes
- 6 explanatory figures and a worked exercise answer guide
Before you begin
Basic Python and command-line familiarity. Python 3.10 or later. The companion code uses the standard library.
Evidence mode runs without an API key or model. Generative mode requires Ollama, a separately installed local model, and suitable hardware. Model weights and subscriptions are not included.
CLEAR EXPECTATIONS
Before you buy.
Is this a complete hosted chatbot?
This is a hands-on ebook and local Python CLI project. It teaches lexical retrieval, a local-model connection, citation checks and evaluation. It does not include hosting, authentication, a vector database, PDF ingestion or production deployment.
What was tested?
The 17 companion tests passed. A recorded local Ollama run covers 12 synthetic cases, one threshold experiment and one repeated question: 9 real model calls and 5 retrieval bypasses. The book includes raw outputs, a rejected invented citation and a contradictory answer with a valid citation. These are teaching observations, not production accuracy claims. Independent human technical and assistive-technology review remains pending.
How do I get the files?
Complete the secure Lemon Squeezy checkout to access the PDF, EPUB and companion ZIP. Keep your receipt and download access private. This is a one-time purchase of edition 1.1.1; ongoing updates and consulting are not included.
Can I share the book or the companion code?
The full ebook is licensed to one reader. The labelled free sample may be shared unchanged. The companion project is MIT licensed and may be adapted or shared with its licence notice. See the exact ebook terms.
Find the evidence. Trace the answer.
One book. A complete learning project. Evidence you can inspect.
AI-assisted · Independent human review pending
$2.99USDBuy ebook