dboai

AI-powered Git CLI tools: auto-generate Jira subtasks & code review from staged changes.

Install

Via npm (recommended)

npm install -g dboai

Local development

git clone https://github.com/dindin-dbo/dboai.git
cd dboai
npm install
npm link

Setup

dboai setup

Saved to ~/.dboai/config.json. Supports:

• Anthropic (Claude)
• OpenAI (GPT-4o, o1, o3)
• Google Gemini
• DeepSeek
• Kimi (Moonshot)
• OpenRouter (all models)
• Groq
• Mistral
• IYH (custom OpenAI-compatible endpoint)

Update only AI provider

dboai setup --provider

Update only Jira config

dboai setup --jira

Asks for Jira URL, email, API token, default project key, and your accountId (for auto-assigning created subtasks to yourself).

Getting your Jira API Token

1. Go to https://id.atlassian.com/manage-profile/security/api-tokens
2. Click Create API token
3. Give it a name (e.g. dboai) and click Create
4. Copy the token — you won't be able to see it again

Getting your Jira Account ID

curl -u "[email protected]:your-api-token" \
  "https://yourcompany.atlassian.net/rest/api/3/myself" | python3 -m json.tool | grep accountId

Recommended: Install glow for better review output

dboai review output is markdown. Install glow to render it nicely in the terminal with colored headings, formatted tables, and syntax-highlighted code blocks:

brew install glow

If glow is not installed, output falls back to plain text with a reminder to install it.

Usage

Generate Jira Subtasks

git add .
dboai subtask EX-1234

• Reads staged diff
• Generates subtasks with SP 1/2/3 (breaks down anything bigger)
• Shows preview for confirmation
• Creates subtasks in Jira, auto-assigned if jira_account_id is set

Code Review

Code review rule plans are managed as standalone markdown files at ~/.dboai/rules/. Each file is self-contained — it defines its own checklist, sections, and required output format. dboai review does not impose any output structure of its own; it simply hands your diff and the rule file to the AI and tells it to follow whatever format the rule file specifies.

This means a rule file can ask for plain PASS/FAIL/SKIP findings, severity-tagged inline comments, a ready-to-run gh pr review bash script — whatever you define.

Managing rule plans

dboai rules list                  # list all rule plans
dboai rules add react-native      # create a new one, opens in $EDITOR (or nano)
dboai rules edit react-native     # edit an existing one
dboai rules show react-native     # print contents to stdout
dboai rules delete react-native   # delete it

Rule plans are global (~/.dboai/rules/<name>.md), shared across every project — no per-repo setup needed. Add as many as you want: react, react-native, golang, python, vue, etc.

Running a review

git add .
dboai review                  # uses the first available rule plan (default)
dboai review react-native     # uses a specific named rule plan
dboai review golang

Output follows the format defined in the rule plan file, rendered with glow if installed. If the rule plan outputs a gh pr review bash script, you'll get back a ready-to-run script.

Config Reference

~/.dboai/config.json:

{
  "ai_provider": "anthropic",
  "ai_model": "claude-sonnet-4-6",
  "ai_api_key": "sk-...",
  "jira_url": "https://yourcompany.atlassian.net",
  "jira_email": "[email protected]",
  "jira_api_token": "...",
  "jira_default_project": "PROJ",
  "jira_account_id": "..."
}

Note: Jira "Sub-task Point" field ID and issuetype name may differ per Jira instance. Check yours via:

curl -u "email:token" \
  "https://yourcompany.atlassian.net/rest/api/3/issue/createmeta?projectKeys=PROJ&issuetypeNames=Sub-task&expand=projects.issuetypes.fields"

Then adjust the customfield_XXXXX and issuetype.name in src/utils/jira.js accordingly.

Troubleshooting

400 error when creating subtask: Jira rejected a field. Check the printed error — it surfaces the errors object from Jira's response (e.g. "story_points": "Field cannot be set...").

429 error on review/subtask generation: AI provider rate limit. Try a different/lighter model via dboai setup --provider.

404 error: Wrong model name for the selected provider. Re-run dboai setup --provider and pick "Enter manually" to type the exact model ID.

dboai review says no rule plans found: Run dboai rules add <name> first to create one.