YAML Configuration File

Customize DAIV for your repository using a .daiv.yml YAML configuration file in the default branch and repository root directory.

This file lets you control features, code formatting, and more.

Example Configuration

Below is a complete example of a .daiv.yml YAML configuration file.

You can copy and modify this template for your repository.

YAML
# Repository settings
default_branch: main
context_file_name: "AGENTS.md"

# File access
extend_exclude_patterns:
  - "**/tests/**"
  - "**/*.test.ts"
  - "coverage/**"

omit_content_patterns:
  - "*.min.js"
  - "*.svg"
  - "*.sql"
  - "*.lock"

# Features
issue_addressing:
  enabled: true

code_review:
  enabled: true

quick_actions:
  enabled: true

# Sandbox
sandbox:
  base_image: "python:3.12-alpine"
  format_code:
    - "ruff check --fix"
    - "ruff format"

# Model configuration
models:
  plan_and_execute:
    planning_model: "openrouter:anthropic/claude-sonnet-4.5"
    planning_fallback_model: "openrouter:openai/gpt-5.2"
    planning_thinking_level: "medium"
    execution_model: "openrouter:anthropic/claude-sonnet-4.5"
    execution_fallback_model: "openrouter:openai/gpt-5.2"
    execution_thinking_level: null
    code_review_model: "openrouter:openai/gpt-5.1-codex-mini"

Configure Repository Settings

Repository settings control the default branch and context file configuration.

Option	Type	Default	Description
`default_branch`	`str \\| null`	Repository default branch	The branch DAIV uses by default to load the `.daiv.yml` YAML configuration file.
`context_file_name`	`str \\| null`	`"AGENTS.md"`	File name to load the repository context in the format of https://agents.md/.

Tip

The context file helps agents understand your repository structure and conventions. You can use the AGENTS.md format to define your repository context.
If not specified, DAIV will look for an "AGENTS.md" file by default.

Enable or Disable Features

Control which DAIV features are active in your repository.

Configure the following feature sections in your .daiv.yml YAML configuration file:

Issue Addressing

Option	Type	Default	Description
`enabled`	`bool`	`true`	Enable the issue addressor feature.

Enable automated issue resolution with plan approval:

YAML
# Minimal configuration for automated issue resolution
issue_addressing:
  enabled: true

quick_actions:
  enabled: true  # Required for /approve-plan command

Note

Plan approval is handled via the /approve-plan quick action. See Quick Actions for details.

Code Review

Option	Type	Default	Description
`enabled`	`bool`	`true`	Enable the code review addressor feature.

Quick Actions

Option	Type	Default	Description
`enabled`	`bool`	`true`	Enable quick actions feature.

Tip

Disable features you do not need to reduce noise and speed up processing.

Branch Naming and Commit Message Conventions

DAIV automatically generates branch names and commit messages for pull requests based on repository conventions. You can define these conventions in your AGENTS.md context file.

Example AGENTS.md conventions:

Markdown
# Repo conventions

## Branch naming
Use: pr/<issue-id>-<kebab-summary>
- <issue-id> must be lowercase (e.g., abc-123)

## Commit messages
Use: (<ISSUE-ID>) <type>: <summary>
- <ISSUE-ID> must preserve original casing from the ticket (e.g., ABC-123)
- <type> ∈ feat|fix|chore|docs|refactor|test

If no conventions are defined in AGENTS.md, DAIV uses sensible defaults: - Branch naming: <type>/<short-kebab-summary> where type ∈ {feat, fix, chore, docs, refactor, test} - Commit messages: Conventional Commits style <type>: <short summary>

Tip

The PR describer agent reads your AGENTS.md file to understand your repository's conventions
If multiple conventions exist, the agent chooses the one that best matches the change type
Keep conventions clear and simple to ensure consistent results

Control File Access

Control which files DAIV can see and read.

Warning

Files excluded from being seen and/or read will not be available to DAIV's AI agents.

Option	Type	Default	Description
`extend_exclude_patterns`	`list[str]`	`[]`	Add patterns to exclude more files from being seen.
`exclude_patterns`	`tuple[str]`	`[".pyc", ".log", ".zip", ".coverage", "/.git/", "/.mypy_cache/", "/.tox/", "/vendor/", "/venv/", "/.venv/", "/.env/", "/node_modules/", "/dist/", "/__pycache__/", "/data/", "/.idea/", "/.pytest_cache/", "/.ruff_cache/"]`	Override the default exclude patterns.
`omit_content_patterns`	`tuple[str]`	`["package-lock.json", "pnpm-lock.yaml", ".lock", ".svg", "*.sql"]`	Files that DAIV can see exist but won't read their content.

Tip

Exclude sensitive files and build artifacts.
Prefer using extend_exclude_patterns to add more patterns.
Use omit_content_patterns for large files that shouldn't be read but need to be seen.

Set Up Sandbox

To take advantage of the sandbox to execute commands, you must have a daiv-sandbox instance running (see the daiv-sandbox repository for more information), and the base_image option must be set to enable sandbox functionality.

Under your .daiv.yml YAML configuration file's sandbox: section, configure the following keys:

Option	Type	Default	Description
`base_image`	`str \\| null`	`null`	Docker image for the sandbox. Use distro images only.
`format_code`	`list[str] \\| null`	`null`	List of commands to format code before committing.

Here's how it works:

Before committing code generated by DAIV, DAIV will call daiv-sandbox to:

Create a container from the base_image.
Execute each command in the format_code list sequentially to format the code before committing.

Example configuration:

YAML
sandbox:
  base_image: "python:3.12-alpine"
  format_code:
    - "pip install ruff"
    - "ruff check --fix"
    - "ruff format"

Warning

If any of the commands fail, DAIV will commit the code as is to be manually fixed by the user, if needed.

Tip

Use specific image versions for reproducibility.
Include dependency installation commands in the format_code list if needed.
The sandbox is only enabled when base_image is specified and daiv-sandbox is running.

Configure Model Settings

Override the default model configurations for agents on a per-repository basis. This allows you to use smaller models for simple projects or larger models for complex ones.

Configuration priority (highest to lowest): 1. Issue labels (e.g., daiv-max) - highest priority 2. .daiv.yml models section - per-repository overrides 3. Environment variables - global defaults (lowest priority)

Plan and Execute Agent

Configure models for the plan and execute agent used in issue addressing.

Option	Type	Default	Description
`planning_model`	`str \\| null`	`null`	Model name for planning tasks. Overrides `PLAN_AND_EXECUTE_PLANNING_MODEL_NAME` environment variable.
`planning_fallback_model`	`str \\| null`	`null`	Fallback model name for planning tasks. Overrides `PLAN_AND_EXECUTE_PLANNING_FALLBACK_MODEL_NAME` environment variable.
`planning_thinking_level`	`"minimal" \\| "low" \\| "medium" \\| "high" \\| null`	`null`	Thinking level for planning tasks. Overrides `PLAN_AND_EXECUTE_PLANNING_THINKING_LEVEL` environment variable.
`execution_model`	`str \\| null`	`null`	Model name for execution tasks. Overrides `PLAN_AND_EXECUTE_EXECUTION_MODEL_NAME` environment variable.
`execution_fallback_model`	`str \\| null`	`null`	Fallback model name for execution tasks. Overrides `PLAN_AND_EXECUTE_EXECUTION_FALLBACK_MODEL_NAME` environment variable.
`execution_thinking_level`	`"minimal" \\| "low" \\| "medium" \\| "high" \\| null`	`null`	Thinking level for execution tasks. Overrides `PLAN_AND_EXECUTE_EXECUTION_THINKING_LEVEL` environment variable.
`code_review_model`	`str \\| null`	`null`	Model name for code review tasks. Overrides `PLAN_AND_EXECUTE_CODE_REVIEW_MODEL_NAME` environment variable.
`code_review_thinking_level`	`"minimal" \\| "low" \\| "medium" \\| "high" \\| null`	`null`	Thinking level for code review tasks. Overrides `PLAN_AND_EXECUTE_CODE_REVIEW_THINKING_LEVEL` environment variable.

Example configuration:

YAML
models:
  plan_and_execute:
    planning_model: "openrouter:anthropic/claude-haiku-4.5"  # Use smaller model for simple projects
    execution_model: "openrouter:anthropic/claude-haiku-4.5"
    planning_thinking_level: "low"  # Reduce thinking for faster responses

Review Addressor Agent

Configure models for the review addressor agent used in code review.

Option	Type	Default	Description
`review_comment_model`	`str \\| null`	`null`	Model name for routing review comments. Overrides `REVIEW_ADDRESSOR_REVIEW_COMMENT_MODEL_NAME` environment variable.
`reply_model`	`str \\| null`	`null`	Model name for replying to review comments. Overrides `REVIEW_ADDRESSOR_REPLY_MODEL_NAME` environment variable.
`reply_temperature`	`float \\| null`	`null`	Temperature for the reply model. Overrides `REVIEW_ADDRESSOR_REPLY_TEMPERATURE` environment variable.

Example configuration:

YAML
models:
  review_addressor:
    review_comment_model: "openrouter:openai/gpt-4.1-mini"
    reply_model: "openrouter:anthropic/claude-haiku-4.5"
    reply_temperature: 0.2

Note

The review addressor agent calls the plan and execute agent to address the review comments. Therefore, the review addressor agent uses the same model configuration defined for the plan and execute agent to address the review comments.

Codebase Chat Agent

Configure models for the codebase chat agent.

Option	Type	Default	Description
`model`	`str \\| null`	`null`	Model name for codebase chat. Overrides `CODEBASE_CHAT_MODEL_NAME` environment variable.
`temperature`	`float \\| null`	`null`	Temperature for codebase chat. Overrides `CODEBASE_CHAT_TEMPERATURE` environment variable.

PR Describer Agent

Configure models for the PR describer agent.

Option	Type	Default	Description
`model`	`str \\| null`	`null`	Model name for PR description. Overrides `PR_DESCRIBER_MODEL_NAME` environment variable.

Example configuration:

YAML
models:
  pr_describer:
    model: "openrouter:openai/gpt-4.1-mini"

Note

The PR describer agent automatically reads your AGENTS.md context file to understand branch naming and commit message conventions. See Branch Naming and Commit Message Conventions for details.