# whathaveidone

> A Rust terminal UI (TUI) that scans Git repositories below the current directory, lists recent commits grouped by time and repo, and turns them into a ready-to-read daily-standup summary using AI. Works with Google Gemini out of the box, or any OpenAI-compatible API (OpenRouter, the Vercel AI Gateway, OpenAI, a local server). The binary is `whathaveidone`, with a short alias `whid`.

This file follows the llms.txt convention (https://llmstxt.org). It is a concise, machine-readable description of the project and how to use it.

## What it is

- Name: whathaveidone (alias: whid)
- Type: command-line / terminal UI tool, written in Rust (Ratatui + Crossterm)
- Purpose: summarize your recent Git commits for daily standups with AI
- License: MIT
- Repository: https://github.com/heroheman/whathaveidone
- Package: https://crates.io/crates/whathaveidone
- Documentation: https://heroheman.github.io/whathaveidone/docs.html
- It shells out to the `git` binary (it does not use a git library). `find_git_repos` recurses one level unless the start dir is itself a repo.

## Install

```sh
cargo install whathaveidone
```

This installs two binaries: `whathaveidone` and the short alias `whid`. Rust is required to build.

## Quick usage

```sh
whathaveidone                 # interactive TUI; today's commits across all repos below ./
whathaveidone week --lang german
```

On first launch an interactive setup wizard configures provider, model, API key, and language. Inside the TUI press `a` to generate an AI overview, `?` for help, `q` to quit.

## Programmatic / non-interactive use (recommended for agents and scripts)

Two flags skip the TUI entirely and print to stdout, so the output is pipeable. The setup wizard and API-key prompt are skipped in this mode; a missing key or provider error is written to stderr with a non-zero exit code, so piped output stays clean.

```sh
# Print raw commits (grouped per repo with a "## reponame" header) and exit
whathaveidone --list
whathaveidone -l week

# Generate an AI summary, print it to stdout, and exit
whathaveidone --generate
whathaveidone -g week --lang german > standup.md
whathaveidone -g | pbcopy
```

Both honour the timeframe argument and the `--from` / `--to`, `--lang`, `--provider`, `--model`, and `--base-url` flags. A successful `--generate` is also saved to the overview history (capped by `recent_generations`).

## CLI flags

- `[timeframe]` — positional start window: `today`/`24` (default), `48`, `72`/`yesterday`, `week`, `month`, or a raw number of hours.
- `--lang <language>` — summary language (e.g. `english`, `german`).
- `--model <model>` — model for the active provider.
- `--provider <gemini|custom>` — AI backend (default `gemini`).
- `--base-url <url>` — base URL of a custom OpenAI-compatible endpoint.
- `--from <YYYY-MM-DD>` / `--to <YYYY-MM-DD>` — custom date range (`--to` defaults to today).
- `--prompt <file.txt>` — custom prompt template.
- `--setup` — re-run the setup wizard (provider, model, API key, language).
- `-l`, `--list` — print commits to stdout and exit.
- `-g`, `--generate` — generate an AI summary, print to stdout, and exit.
- `--debug` — show prompt-construction debug info in the summary.

## Providers, models, and API keys

- Default provider is Google Gemini; default model `gemini-3.1-flash-lite`.
- Gemini key resolution: config `gemini_api_key` → `GEMINI_API_KEY` env var → interactive prompt.
- Custom OpenAI-compatible providers: set `--provider custom`, a `--base-url`, and a `--model`. Key comes from `custom_api_key` in config or the `CUSTOM_API_KEY` env var.

```sh
export GEMINI_API_KEY=your-key-here
export CUSTOM_API_KEY=sk-...

# OpenRouter
whathaveidone --provider custom --base-url https://openrouter.ai/api/v1 --model openai/gpt-4o-mini

# OpenAI
whathaveidone --provider custom --base-url https://api.openai.com/v1 --model gpt-4o-mini
```

## Custom prompt placeholders

A `--prompt file.txt` template may use: `{from}`, `{to}`, `{project}` / `{projectname}`, `{interval}`, `{lang}`, `{commits}`.

## Configuration

TOML config at `~/.config/whid/whid.toml` (auto-created on first run). Settings layer: embedded blueprint defaults → user config → optional `whid.toml` in the current directory → CLI flags (highest precedence).

Keys: `provider`, `gemini_model`, `gemini_api_key`, `prompt_for_api_key`, `custom_base_url`, `custom_model`, `custom_api_key`, `openrouter_api_key`, `vercel_api_key`, `openai_api_key`, `custom_prompt_path`, `lang`, `recent_generations` (default 20 — how many recent AI generations to keep in `~/.config/whid/overviews.json`, used by both the TUI and `--generate`).

## TUI: three views

- Commits (`1`) — repository sidebar + commit list. Toggle Timeframe/Selection mode with `s`.
- Overviews (`2`) — master/detail browser of saved AI summaries (persist across restarts).
- Stats (`3`) — full-screen dashboard: totals (commits, active days, avg per active day, busiest day, repos), commits per day / weekday / hour, per-repo and per-author breakdown, conventional-commit type breakdown (feat/fix/chore/…), and ticket-reference breakdown (e.g. ABC-123).

## TUI: key bindings

- Global: `1`/`2`/`3` switch view, `Tab`/`Shift+Tab` move focus, arrows or `h`/`j`/`k`/`l` navigate, `?` help, `q` quit.
- Commits: `Space` detail pane, `s` list mode, `m` mark commit/repo, `[`/`]` timeframe, `u` mine/all authors, `d` detailed view, `a` generate overview.
- Overviews: `Enter`/`c` copy, `r` regenerate, `x`/`Del` delete, `a` generate, `Esc` back.
- Stats: `[`/`]` timeframe, `w` week, `u` mine/all, `d` detailed breakdown.

## Links

- [Home](https://heroheman.github.io/whathaveidone/): project landing page.
- [Documentation](https://heroheman.github.io/whathaveidone/docs.html): full usage docs with examples.
- [GitHub repository](https://github.com/heroheman/whathaveidone): source, issues, README.
- [crates.io package](https://crates.io/crates/whathaveidone): install via `cargo install whathaveidone`.