StandupBot

Generate standup updates from recent git activity with a local Ollama model or Groq's free cloud tier.

What is new in 0.2.4

• Fixed _infer_modules: two-component paths like tests/test_auth.py returned the filename instead of the parent directory; now returns tests
• Fixed auto_cleanup_if_needed: was unreachable because its threshold (400) exceeded the _enforce_max_rows ceiling (365); now delegates directly
• Fixed quality scoring timeout: _score_with_ollama had the same Ollama timeout bug fixed in v0.2.3 but was never patched
• Fixed startup script: standup warm-up --install-startup generated an invalid standup standup warm-up command; now outputs standup warm-up
• Groq model validation changed from a hard allowlist to a non-empty string check so users are not blocked by newer models
• _get_provider_slug replaced fragile class-name string matching with explicit isinstance checks
• Modern type annotations enforced throughout (list[X], X | None, built-in generics) after dropping Python 3.9

Quick Start

# Install dependencies
pip install -e .

# Configure StandupBot
standup --setup

# Generate a standup
standup

For Ollama, install and pull a model first:

ollama pull llama3

Providers

Provider	Cost	Privacy	Best for
Ollama	Free	Fully local	Private, offline workflows
Groq	Free tier	Cloud	Fast generation without local model setup

Groq API keys should be supplied through GROQ_API_KEY when possible.

Core Commands

# Standard generation
standup
standup --hours 48
standup --week
standup --provider groq

# Output options
standup --copy
standup --slack
standup --raw
standup --template slack
standup --verbose

# Cache and filtering controls
standup --no-cache
standup --no-filter

# Maintenance
standup doctor
standup usage
standup logs
standup logs --tail 50
standup logs --clear
standup models
standup templates
standup history
standup history --limit 25
standup history --clear
standup history --clear --days 30
standup warm-up
standup warm-up --install-startup
standup warm-up --uninstall-startup
standup --maintenance

Templates

Built-in templates:

• default
• slack
• minimal
• detailed
• jira

You can also define custom templates in config. StandupBot extracts yesterday, today, and blockers from the LLM output, then renders the final format with these variables:

• {yesterday}
• {today}
• {blockers}
• {date}
• {time}
• {commit_count}
• {repos}
• {provider}
• {author_email}

Config

StandupBot reads ~/.standup.json.

{
  "repos": [
    "/path/to/repo1",
    "/path/to/repo2"
  ],
  "author_email": "you@example.com",
  "hours_lookback": 24,
  "tone": "casual",
  "slack_webhook_url": "",
  "provider": {
    "name": "ollama",
    "ollama": {
      "base_url": "http://localhost:11434",
      "model": "llama3"
    },
    "groq": {
      "api_key": "",
      "model": "llama-3.1-8b-instant"
    }
  },
  "rate_limit": {
    "cooldown_minutes": 30,
    "max_calls_per_day": 10,
    "enabled": true
  },
  "quality": {
    "enabled": true,
    "min_score": 0,
    "show_breakdown": false
  },
  "noise_filter_enabled": true,
  "template": "default",
  "custom_templates": {
    "my_format": "Done: {yesterday} | Doing: {today} | Help needed: {blockers}"
  },
  "auto_warm_up": false
}

Caching and History

Every generated standup is stored locally in ~/.standup_history.db.

• Cache keys are based on a SHA256 fingerprint of sorted commit hashes.
• Cache reuse is limited to the same day, tone, and provider.
• The database stores standup text, provider metadata, repo names, lookback hours, and quality score.
• Raw commit messages are not stored in the database.

Quality Scoring

After generation, StandupBot can score the standup from 0-100 and show a colored badge.

• Green: 80+
• Yellow: 60-79
• Red: below 60

When quality.min_score is above zero, StandupBot retries low-quality outputs up to two times with refined guidance.

Warm-Up

Use standup warm-up to pre-load the configured model before your first real run.

• Ollama: sends a minimal warm-up request to keep the selected model ready in memory.
• Groq: runs a lightweight availability ping.
• auto_warm_up can trigger a silent warm-up when the model has not been used recently.

Security

StandupBot treats config, git metadata, templates, provider responses, and local storage as hostile inputs until proven otherwise.

• ~/.standup.json, ~/.standup_usage.json, ~/.standup_history.db, and ~/.standup.log use restricted permissions on Unix/macOS.
• Repo paths go through explicit path-safety checks to block traversal tricks, network paths, and unsafe symlinks.
• Commit messages and LLM responses are length-capped before they reach prompts, storage, or terminal rendering.
• Custom templates only substitute a fixed allowlist of variables and reject Python-style format syntax.
• All history database queries are parameterized, and stored standup text is sanitized before persistence.
• User-facing exception messages are sanitized so file paths, emails, and API-key-shaped values are not echoed back to the terminal.
• standup doctor now checks log health, DB size, schema version, WAL mode, file permissions, and full config validity.

Development

python -m pytest tests/ -q

If your environment restricts Python temp directories or __pycache__ writes, set a writable --basetemp or PYTHONDONTWRITEBYTECODE=1 while testing.

License

MIT