Applied AI / Local application

Ask the documents.

An answer with something behind it.

WHAT IT DOES

A local document Q&A pipeline combining semantic retrieval with an Ollama language model.

WORK IN THIS REPOSITORY

A local document retrieval and question-answering application.

THE INTERESTING DECISION

Semantic search selects document context before the model produces an answer, helping connect responses to the knowledge base.

START WITH THE EVIDENCEFAISS

Vector search over document embeddings.

Inspect the source ↗

BUILT WITH

LangChainFAISSOllamaStreamlit

PROJECT TYPE

Local application

START HEREREADME & setup

01 / THE IDEA

An answer with something behind it.

The problem

A general language model does not automatically know the contents of a specific handbook. Useful answers need relevant document context and a straightforward interface.

The approach

The pipeline ingests PDFs, creates HuggingFace sentence embeddings, retrieves context through FAISS, and gives that context to a local Ollama model through LangChain. Streamlit provides the interface.

02 / THE SYSTEM AT A GLANCE

Follow the flow.

01PDF documents
02Embeddings
03FAISS retrieval
04Local model
05Answer

A simplified architecture overview. The repository documents the implementation details.

03 / THE ENGINEERING DECISIONS

Why it's built this way.

Retrieve before generating

Semantic search selects document context before the model produces an answer, helping connect responses to the knowledge base.

Keep inference local

Ollama provides local inference. This avoids a paid hosted model dependency but requires a machine capable of running the model.

Start with a concrete collection

The initial use case is the FAST-NUCES FYP handbook; the document pipeline can be adapted to other collections.

04 / WHAT TO LOOK AT

Evidence over adjectives.

RetrievalFAISS

Vector search over document embeddings.

InferenceLocal

Ollama, not a public hosted model endpoint.

InterfaceStreamlit

Interactive document question answering.

Capabilities and result snapshots follow the project's current README, reviewed October 10, 2026. Read the source

05 / THE HONEST PART

What I'd improve next.

Current boundaries

Local Ollama and the document index must be configured to run the application.
Retrieved context does not guarantee factual answers; a dedicated evaluation would strengthen the project.

The next iteration

Add a repeatable evaluation set for retrieval relevance and answer grounding.

Open my Software & AI CV

KEEP EXPLORING

Before checkout ships

The most useful test can be a failure.