Laur Ivan - Laur's blog

Sign in Subscribe

Laur Ivan

Vibe Coding - Add a field to a Qdrant collection

Vibe Coding - Add a field to a Qdrant collection

I have created a Qdrant collection off a repository of mine and I had the following payload fields: * file_name - the source code file name (path actually) * branch - the git repository branch that I have indexed * repository_name - the git repository name * chunk - the text chunk

Chunk and index text

Chunk and index text

This is part of the Use RAG with Continue.dev series. Once we have the dependencies installed, we can start filling up the vector db. Set up First part, is to create the LanceDB database and embedding model: from lancedb.embeddings import get_registry from lancedb import connect OLLAMA_HOST

Prepare Ollama and LanceDB

Prepare Ollama and LanceDB

This is part of the Use RAG with Continue.dev series. Prerequisites I kept in line with the general prerequisites for the previous steps (python 3, uv for package management). In addition, I will assume you have ollama installed. Check here for different installation methods. Setup Ollama Once Ollama is

Bitbucket - Scan a repository

Bitbucket - Scan a repository

Following the previous post, we need to get the list of files in the repository. This is similar with listing the contents of a directory, without having the possibility to do it in a recursive way. A call to the list files API will return a list containing both files

Bitbucket - Get Repository Information

Bitbucket - Get Repository Information

In the previous post, we wrote the helper functions, which will allow us to get multi-page responses from the Bitbucket server. Now, we can use them to get various repository information. Prerequisites As in the previous post, we need the requests python package. Once we have that, we can initialise

Bitbucket - API Helper Functions

Bitbucket - API Helper Functions

Get all pages Normally, Bitbucket list queries return a JSON. They are structured so all the data is under a values key. You get a bunch of other data, among which a flag isLastPage that tells you if the page you retrieved is the last one. I wrote below a

Use Rag with continue.dev

Use Rag with continue.dev

continue.dev is a plugin system that is available for VS Code and IntelliJ products. It allows using the code base to answer AI questions, as well as custom RAG extensions, where you can index e.g. your proprietary repositories. The purpose of this exercise is to create an ollama-based

Updated the Ansible Unifi role

Updated the Ansible Unifi role

After moving my Unifi controller to a new VM, I kept updating its version. Recently, Ubiquiti release a new version (9.0.108 at the moment of writing), which has a bunch of changes. The ones I am concerned with are: 1. The name is changed from unifi-controller to unifi-network-application

RAG Sample - 5: Bring it all Together

RAG Sample - 5: Bring it all Together

We have now all data available in the right places: 1. Select a Vector DB 2. Build a RAG System 3. Generate Schema-compliant Embeddings 4. Build embeddings and insert into DB I have indexed a research paper and can ask the system a question. The process is somewhat simple: 1.

Movie review - The departed

Movie review - The departed

I've seen this one as well. It's way better than the Bond movie described in the previous post. Given that is 5am, I'm just going to give it a 9/10. Lots of good actors, Nicholson is perfect for that crazed gangster role. The

Magellan

A long time ago, I worked on developing Magellan, an outlook-like client for Linux/KDE. I worked on front-end and some backend stuff (until the back-end moved to CORBA that is). I also did the art (icons, gui...) for it. Here are some screenshots: The email interface supported POP3 and

Text shadows in KDE

Text shadows in KDE

Long time ago, I wrote the original code for having text shadows on KDE's desktop. It was using a blur effect 😄 Here are some images from those days:

About the E-tray

About the E-tray

E-tray is used by EPSO now as a stop-gap for various competitions (First batch was the cycle for AD in 2015). It previously was part of the AST assessment centre. From EPSO: An e-tray exercise is a computer-based simulation of a real work situation and replicates an email inbox which

A Strategy for the E-tray

A Strategy for the E-tray

First, some terminology: * CBT - Computer based test * Question - the big scenario with the three options * Option - The three items you need to evaluate (from -- to ++) The E-tray looks like a SJT, but has some major differences: * You need to remember that each option has its own

Interview Skills and Competencies (generic)

Interview Skills and Competencies (generic)

Following skills and competencies are assessed generally in an interview. Interviews in the Assessment Center are usually a subset of these criteria, mapped on competences. Personal Effectiveness * Self-motivated – strong commitment to achieving challenging work goals * Strong energy levels * Resilient and works well under pressure * Perseveres when faced with difficulties * Reacts

Tip: Memorise the Competences

Tip: Memorise the Competences

Memorise the EPSO competencies via acronyms.

RAG Sample - 4. Build Embeddings and Insert to DB

RAG Sample - 4. Build Embeddings and Insert to DB

Building embeddings I followed two sources for embeddings: 1. Building an Agentic RAG locally with Milvus, Ollama and LangGraph and the associated code langgraph-rag-agent-local.ipynb 2. The "Build RAG with Milvus | Milvus Documentation" tutorial This is where I had my first surprise: Building embeddings is not necessarily a

RAG Sample - 3. Generate Schema-compliant Embeddings

RAG Sample - 3. Generate Schema-compliant Embeddings

Until now, we've initialised the collection with a custom schema and built the capability to generate embeddings via Ollama. Now, we'll tie it all together. A sample to process We define a format for the document (based on the paperless-ngx API response): doc = { "id"

RAG Sample - 2. Build the RAG system

RAG Sample - 2. Build the RAG system

Prerequisites Before starting work, I made sure I have the following items in place: 1. Ollama - installed on Windows, just to eliminate headaches with having a NVIDIA card available from docker. I also made sure I have llama3:latest installed via ollama pull llama3 2. Milvus via docker compose.

RAG Sample - 1. Select a Vector DB

RAG Sample - 1. Select a Vector DB

Create the RAG Initially, I thought that RAG is a simple matter: * Set up the vector DB * Create and add the vectors to the DB * Do a query in the DB with the prompt * Get the results from the DB * Feed the results and the prompt into the LLM I&

Get the Text-to-Speech from Articulate Storyline 360

I needed a process to extract the TTS content from an Articulate Storyline file, and make it reproducible. Here's my process and outcome. The Spec I've started to investigate a bit the format, and to my surprise, it was the same approach as MS Office: a

Install Sidero Omni Locally

In the future, I'll have a raspberry pi cluster. I'm toying in advance with installing kubernetes and various apps, to learn more about k8s and see if it's all worth it for homelabbing. I settled on installing Talos linux and I've seen

Install ArgoCD

Well, I'm not about to do manual installs for everything, and I like to automate and keep thing up to date as much as possible. So first thing is to install argocd. I'm following this tutorial and this is the first step I performed after I

Configure Talos linux

This entry assumes you have already available some Talos linux instances (be it VMs, or bare metal). It follows the documented steps (v.17 at the moment of writing). Prerequisites You need to have the IPs of the Talos linux instances, so you can build the cluster(s). Select one

Talos Linux on ESXi

Talos is an immutable OS. In practical terms, it means that the physical disk is formatted and you can't put anything on it anymore, like adding kubernetes storage. You can actually do it if your HDD/SSD supports namespaces, but in ESXi I haven't found a