Tacit CLI

Tacit helps operators turn incident prompts and existing operational context into evidence-grounded investigations from the command line.

The tacit command can:

• configure local Grafana, SignalFx, and LLM settings
• run environment health checks
• run an investigation workflow from a natural-language prompt
• start the local API and browser UI
• run the demo stack
• ingest existing dashboards, alerts, runbooks, and incidents as reusable operational context
• export an anonymized assessment bundle for sharing

Install

uvx --from tacit-ai tacit --help

Or install into your environment:

pip install tacit-ai
tacit --help

Quick Start

Configure Tacit:

tacit setup

Check your local environment:

tacit doctor

Run a test investigation:

tacit test --prompt "checkout-service p95 latency is high"

Run the local API and UI:

tacit serve --host 127.0.0.1 --port 8000 --no-slack

Then open http://127.0.0.1:8000.

Common Commands

tacit setup
tacit doctor
tacit test --prompt "5xx errors on checkout-service"
tacit serve --no-slack
tacit demo
tacit learn dashboard <dashboard_uid>
tacit learn approve <dashboard_uid>
tacit history list
tacit export-report --anonymous --validate

Configuration

Tacit reads settings from environment variables, .env, and optional YAML config. For local development, start from the repository's .env.example or tacit.yaml.example.

Useful settings include:

• GRAFANA_URL
• GRAFANA_API_KEY
• LLM_PROVIDER
• LLM_MODEL
• LLM_API_KEY
• SIGNALFX_API_TOKEN
• API_AUTH_ENABLED
• API_AUTH_KEY

More Documentation

The full repository README covers architecture, development workflows, Docker compose stacks, evaluation notes, and project roadmap:

https://github.com/aditki/tacit