OrgChat Check

A lightweight CLI to check production-readiness of an organization chatbot project before delivery.

Main command:

orgchat check

Quick Start (no setup required)

cd your-chatbot
npx orgchat check

No clone. No manual Python install. No mandatory orgchat.toml.

If you already have the repo:

pip install -e .
orgchat check

It runs seven important checks:

1. Where is the data and how is it refreshed?
2. Who has access to what information?
3. How do we measure correct answers?
4. What happens if the model doesn't know the answer?
5. How do we observe cost and latency?
6. Who restores the system when it breaks?
7. Who maintains the system after delivery?

Local install

Requires Python 3.11+.

python -m venv .venv
source .venv/bin/activate       # Linux/macOS
# .venv\Scripts\activate       # Windows PowerShell
pip install -e .

Run without cloning (PyPI / uvx)

No local install or clone needed:

uvx orgchat check --config ./orgchat.toml
# or global install
pip install orgchat
orgchat check

Quick start

In the root of the project you want to check:

orgchat init

Fill in orgchat.toml values, then:

orgchat check

For file output:

orgchat check --format markdown --output orgchat-report.md
orgchat check --format json --output orgchat-report.json

To check a different path:

orgchat check --path ./my-chatbot

For CI:

orgchat check --strict

Exit codes:

• 0: All checks pass or only non-strict warnings exist.
• 1: At least one check failed, or warnings exist in --strict mode.
• 2: Input, path, or settings error.

Report logic

orgchat.toml is the main evaluation source. Explicit false values or wrong paths cause FAIL. Unset values get WARN. The CLI also looks for common files/folders (evals, retrieval, monitoring, runbook) as supporting evidence.

Discovered evidence alone does not replace explicit configuration; its purpose is to provide a useful initial report from the existing project and show the path to completion.

Example output

ORGCHAT CHECK
Project: Enterprise Support Chatbot
Overall: WARN | Score: 64.5/100

Checks:
  [PASS] 100.0/100  Data and freshness
  [WARN]  55.0/100  Quality evaluation
  [FAIL]  40.0/100  Access and security

Next actions:
  1. Access and security: filter retrieval with the same user's permissions.

Project structure

src/orgchat/
  checker.py    # seven-check logic
  config.py     # TOML and path reading
  report.py     # terminal/json/markdown output
  cli.py        # orgchat check and orgchat init commands
tests/
examples/

This release reports operational readiness status; it is not a replacement for security testing, model evaluation, or full organizational audit.