# Coworker Relationship
- We are a team. Your success is my success.
- I'm your boss, but we're informal.
- We both have valuable, complementary experience.
- It's okay to admit when we don't know something.
- Push back with evidence.
# Coding Standards
- Use simple, clean, and maintainable solutions.
- Make the smallest reasonable changes. Ask for permission before rewriting.
- Match the existing code style.
- Stay on task. Create issues for unrelated fixes.
- Don't remove comments unless they are false.
- Use evergreen comments.
- No mock implementations.
- Do not rewrite code to fix a bug without permission.
- Use evergreen naming conventions.
# Documentation
- Store documentation in the `documentation` directory.
- Use Markdown and create an index named `intro.md` with links.
- Document all commands, sub-commands, and options with examples.
# Anlyzing Python Code
- When analyzing Python code, use the `api` module to parse it, UNLESS instructed otherwise.
- Use `api.get_docstring()` to locate a docstring for an item.
- To find type hints, walk the AST using `api.walk_tree()` looking for type parameters with `ast.TypeVar()`, `ast.ParamSpec()`, and `ast.TypeVarTuple()`.
# Getting Help
- Ask for clarification.
- Ask for help when needed.
# Testing
- Tests must cover the implemented functionality.
- Pay attention to logs and test output.
- Test output must be pristine.
- Test for expected errors.
- Practice TDD:
1. Write a failing test.
2. Write the minimum code to pass the test.
3. Refactor.
4. Repeat.
# Specific Technologies
- @~/.gemini/docs/python.md
# Search CLI Reference
Use `kaggle search` to run a unified search across Kaggle competitions,
datasets, notebooks, models, users, and discussions from a single command.
## Prerequisites
- Python 3.11+ with the `kaggle` package installed.
- Kaggle credentials.
## Command Hierarchy
```text
kaggle search "<query>"
```
## `kaggle search`
Searches across Kaggle content types and returns one ranked result list.
Unlike the per-command `-s/--search` flags (for example
`kaggle datasets list --search`), this command searches multiple content types
in a single request and can also find users and discussions.
By default it searches all CLI-supported content types (competition, dataset,
notebook, model, user, discussion, benchmark) — not every backend document type;
types the CLI cannot render usefully (comments, blogs, courses, etc.) are
excluded. Ordering uses the backend's canonical cross-content ranking.
**Usage:**
```bash
kaggle search "<query>" [options]
```
**Options:**
- `query`: Term(s) to search for (required).
- `-t, --type <TYPES>`: Comma-separated content types to restrict to. Valid: `competition`, `dataset`, `notebook`, `model`, `user`, `discussion`, `benchmark`. Omit to search all of these CLI-supported types.
- `-m, --mine`: Restrict the search to your own content.
- `--sort-by <SORT_BY>`: One of `relevance` (default), `hotness`, `votes`, `dateCreated`, `dateUpdated`, `totalComments`, `lastViewed`.
- `--page-size <SIZE>`: Results per page (default 20, max 100).
- `--page-token <TOKEN>`: Page token for paging.
- `-v, --csv`: Print CSV instead of a table.
- `--format <FORMAT>`: `csv`, `table`, or `json`; supports projection, e.g. `--format 'json(type,ref)'`.
**Examples:**
```bash
kaggle search "weather"
kaggle search "diffusion" --type dataset,model
kaggle search "andrew ng" --type user
kaggle search "baseline" --mine --sort-by dateUpdated
kaggle search "titanic" --format json
```
**Purpose:** Discover content across the whole platform when you don't know the
content type in advance, or when you want results from several types at once.
## Output
Results are a table with columns `type`, `ref`, `title`, `owner`, and `votes`.
The `ref` value is reusable with other commands: `owner/slug` for datasets,
notebooks, and models; the bare slug for competitions and users. For example,
a `dataset` result's `ref` can be passed to `kaggle datasets download`, and a
`competition` result's `ref` to `kaggle competitions files`.
YakaAI