Setting Up Go Development with NeoVim

NeoVim Advantages

• Wide selection of color schemes
• Lightweight and fast

Kickstart nvim Configuration

• Single config file at ~/.config/nvim/init.lua

Installing Go Support

1. Install Go syntax highlighting

nvim
:TSInstall go
:TSInstallInfo  # Verify install
:q!

2. Install Go language server

go install golang.org/x/tools/gopls@latest

3. Configure PATH

export PATH=$PATH:$(go env GOPATH)/bin
source ~/.zshrc  # Reload shell

4. Set up LSP in NeoVim

nvim ~/.config/nvim/init.lua
:MasonInstall gopls

Initial impressions are positive, though I recognize no technology is perfect - I’ll need to explore Go’s limitations further.

AWS Glue & Jupyter Notebooks

On my todo list:

• Explore AWS Glue integration
• API deployment from Jupyter Notebooks

Chatbot Enhancement Analysis

Current RAG Pipeline

1. Chunking: Content from ../src/content/**/*.mdx gets broken into chunks
2. Vectorstore: Chunks embedded (OpenAI) and stored in FAISS
3. Retrieval: Finds top 5 similar chunks for each question
4. Generation: Chunks sent to GPT-4o-mini with “TC Heiner” persona

Problem Identified

For prompt “Tell me about your favorite thing you built”:

• ❌ Gets: Generic Astro/Wells Fargo content
• ✅ Should get: ManaBurn, web crawler details

Root Cause

Similarity search matches generic phrases better than project-specific content.

Potential Solutions

1. Content Improvement:

   • Add “favorite project” language to ManaBurn docs

2. Search Parameters:

   • Increase k from 5 to 8-10

3. MMR Retrieval:

retriever = vectorstore.as_retriever(
    search_type="mmr",
    search_kwargs={
        "k": 8,
        "lambda_mult": 0.7,
        "fetch_k": 20
    }
)

4. Custom Scoring:

def custom_retrieval(question, vectorstore):
    docs = vectorstore.similarity_search(question, k=15)
    
    # Apply score boosts based on project recency
    for doc in docs:
        years_ago = get_project_age(doc)
        if years_ago <= 1: doc.score_boost = 0.3
        elif years_ago <= 2: doc.score_boost = 0.2
        elif years_ago <= 3: doc.score_boost = 0.1
        elif years_ago <= 5: doc.score_boost = 0.05
        
    return sorted(docs, key=lambda x: x.score + getattr(x, 'score_boost', 0))[:5]