# GitHub cleanup & three public repos

This environment cannot create or archive GitHub repositories for you (`gh` is read-only here). Do these steps from your laptop once.

## A. Archive or hide low-signal repos

**Archive** (Settings → Archive) anything that’s a tutorial clone, empty scaffold, or decade-old demo. Candidates from the current public list:

- `discourse` (upstream clone)
- `psitsmike_example_2`, `meteor-slack`, `meteor-chat-room`, `source`, `PubNub-`, `pub-messenger`
- `cs50`, `negdemo`, `negitdemo`, `business-idea`
- `machine_learning_class_data`, `mushroom-csv-data`, `shark-tank-csv-data`
- `hackernews-clone`, `Django-Server`, `bollywoodflask`, `hotdogfibrosis` (unless you still want them visible as humor)
- `langchain-ask-the-doc`, `knowledge_gpt` if they’re thin forks without your README/evals
- `oligarchy-game`, `streamerstakes_ios` if unfinished and unexplained
- `shlokzartsharma.github.io` if superseded by shloksharma.com

**Keep visible / improve:**

- `driversvoice` (already strong README — pin it)
- `newfisite` (private site — fine)
- New showcase repos below
- Optionally polish `sonicstream` with a real README or archive it after writing the “failed experiment” story

**Pin (exactly 6):** FlyingMinds-related demo(s), `driversvoice`, the three new showcases, IvyOdds if public.

## B. Publish the three showcases in this repo

Folders under `showcase/` are ready to become standalone public repos:

| Folder | Suggested GitHub name | Signal |
|---|---|---|
| `showcase/llm-eval-harness` | `llm-eval-harness` | Eval-driven LLM grading / agreement metrics |
| `showcase/structured-extract` | `structured-extract` | Zod-validated LLM extraction pipeline |
| `showcase/bkt-adaptive` | `bkt-adaptive` | Bayesian Knowledge Tracing practice kit |

### Publish each

```bash
cd showcase/llm-eval-harness
git init
git add .
git commit -m "Initial commit: LLM eval harness"
gh repo create llm-eval-harness --public --source=. --remote=origin --push
```

Repeat for `structured-extract` and `bkt-adaptive`.

Then pin them and link from https://shloksharma.com/work.html.

## C. Profile README (optional)

Add `shlokzartsharma/shlokzartsharma` profile README pointing to shloksharma.com, FlyingMinds case study, and the three pins — three links max in the hero blurb.