autopsy-cli 0.2.1

AI-powered incident diagnosis for engineering teams. Pull logs and deploys, get a structured root cause in 30 seconds.

Autopsy CLI

AI-powered incident diagnosis for engineering teams. Pull production error logs and recent deploys, send them to an LLM, and get a structured root cause analysis in the terminal in under a minute. Zero-trust: your data never leaves your environment.

Install

pip install autopsy-cli          # core CLI (no TUI)
pip install "autopsy-cli[tui]"   # + interactive terminal UI (requires textual)

Or from source:

git clone https://github.com/zaappy/autopsy.git && cd autopsy
pip install -e ".[dev]"

Quick Start

autopsy           # Launch interactive TUI (menu, then run Diagnose or Setup)
autopsy init      # Or: interactive config wizard (~/.autopsy/config.yaml)
autopsy diagnose  # Or: run diagnosis directly (CloudWatch + GitHub → AI → panels)

Interactive TUI — Run autopsy with no arguments to open the interactive terminal UI:

• AUTOPSY logo and tagline (AI-powered incident diagnosis • zero-trust)
• Arrow-key menu: Diagnose, Setup configuration, Validate connections, Show config, Diagnosis history (cloud, disabled)
• Shortcuts: d Diagnose, i Init, v Validate, c Config, q Quit, Esc Back
• Choosing Diagnose runs the full pipeline inside the TUI (progress steps, then 4-panel result). Errors are shown inline; press Esc to return to the menu.
• Choosing Setup / Validate / Show config exits the TUI and runs the corresponding CLI command in your terminal.

Three steps: install → init → diagnose (via TUI or direct commands).

Configuration

After autopsy init, edit ~/.autopsy/config.yaml or re-run the wizard. The init wizard stores credentials in ~/.autopsy/.env — no manual env var exports needed.

Section	Purpose
aws	CloudWatch region, log groups, time window (minutes). Uses your AWS CLI credentials.
github	Repo (owner/repo), branch, number of recent commits to analyze. Uses GITHUB_TOKEN.
ai	Provider (anthropic or openai), model, API keys.

Credentials are loaded from ~/.autopsy/.env automatically. If you prefer env vars, export them in your shell — they take precedence over the .env file.

Security: Add ~/.autopsy/.env to .gitignore if you ever copy the config directory. Never commit credentials. If your home directory is backed up or synced (e.g. OneDrive, Time Machine, Google Drive), the .env file may be included — consider excluding ~/.autopsy/ from sync or use env vars instead.

autopsy config show       # Print config (secrets masked)
autopsy config validate   # Check env vars and connectivity

How It Works

┌─────────────┐     ┌──────────────────┐     ┌─────────────┐     ┌──────────────┐
│   Config    │────▶│  Data Collectors │────▶│  AI Engine  │────▶│  Renderers   │
│ ~/.autopsy  │     │  CloudWatch      │     │  (Claude /   │     │  Terminal or │
│ config.yaml │     │  GitHub          │     │   OpenAI)    │     │  JSON        │
└─────────────┘     └──────────────────┘     └─────────────┘     └──────────────┘
                           │                          │
                           ▼                          ▼
                    Logs + recent commits      Structured diagnosis:
                    (deduped, truncated)       root cause, deploy, fix, timeline

1. Collect — CloudWatch Logs Insights (error-level) and GitHub (last N commits + diffs).
2. Reduce — Log dedup and token budget; diff filters (code files only, cap per file).
3. Diagnose — Single prompt with logs + deploys; LLM returns JSON (root cause, correlated deploy, suggested fix, timeline).
4. Render — Rich panels in the terminal or --json for piping.

CLI Reference

Command	Description
autopsy	Interactive TUI — menu with Diagnose, Setup, Validate, Config (requires textual)
autopsy init	Interactive config wizard
autopsy diagnose	Run full diagnosis pipeline (same as TUI “Diagnose”)
autopsy diagnose --json	Output raw JSON
autopsy diagnose --time-window 15	Override log window (minutes)
autopsy diagnose --log-group /aws/lambda/foo	Override log groups (repeatable)
autopsy diagnose --provider openai	Use OpenAI instead of Anthropic
autopsy config show	Print config (secrets masked)
autopsy config validate	Check credentials and connectivity
autopsy version / autopsy --version	CLI version, prompt version, Python version

If textual is not installed, autopsy with no arguments prints help instead of starting the TUI.

Contributing

1. Fork the repo and create a branch.
2. Install dev deps: pip install -e ".[dev]".
3. Run lint and tests: ruff check . && pytest.
4. Open a PR against main.

We follow the layout and conventions in the repo (collectors, AI engine, renderers, no business logic in cli.py).

License

MIT. See LICENSE.