Skip to content

usage.md

Latest commit

 

History

History
91 lines (65 loc) · 3 KB

usage.md

File metadata and controls

91 lines (65 loc) · 3 KB

Usage

@Mention

Mention the bot in any MR or issue comment to start a conversation:

@codeward What does this function do?
@codeward Can you explain the changes in this MR?

The bot maintains conversation context within the same discussion thread.

Slash Commands

Command Description
@codeward /review Request a code review on the current MR
@codeward /debug Diagnose the latest pipeline failure
@codeward /security Run security triage on the latest scan results
@codeward /explain Explain the current diff or referenced code
@codeward /summarize Summarize the MR changes
@codeward /help Show available commands

Multi-Agent Flows

Flows chain multiple agents into automated workflows:

Command Description
@codeward /implement Analyze issue, generate code, self-review, open MR
@codeward /full-review Code review + security triage + combined summary
@codeward /fix-pipeline Diagnose pipeline failure and post report

Flows execute step-by-step via the Redis task queue. Track progress with GET /flows/{flow_id}.

See flows.md for YAML schema and custom flow development.

Automatic Triggers

Some agents run automatically without being mentioned:

Agent Trigger Condition
Code Reviewer MR opened or updated Auto-review enabled, not draft, not WIP
Pipeline Debugger Pipeline failed Auto-debug enabled
Security Triage Security scan job completes Auto-triage enabled

Automatic triggers can be controlled globally in codeward.yml and per-project in .codeward.yml.

Incremental Reviews

When a developer pushes new commits to an MR that was already reviewed:

  1. Codeward detects the previous review via stored ReviewState
  2. Only the inter-diff (changes since last review) is sent to the LLM
  3. Previous bot discussions on modified files are auto-resolved with a note
  4. New inline comments are posted only on new/changed lines
  5. The summary note indicates it's an incremental review

If the same SHA was already reviewed, Codeward skips the review entirely.

Project Configuration (.codeward.yml)

Place a .codeward.yml file in the root of your GitLab project to customize Codeward behavior per-project:

codeward:
  disabled_agents:
    - security-triage   # Disable specific agents

  review_rules:
    - "Follow PEP 8 style guidelines"
    - "All public functions must have docstrings"
    - "No hardcoded secrets or API keys"

  skip_paths:
    - "*.lock"
    - "vendor/**"
    - "generated/**"
    - "*.min.js"

Available Options

Option Type Description
disabled_agents list Agent names to disable for this project
enabled_agents list If set, only these agents are allowed
review_rules list Additional review guidelines for the code reviewer
skip_paths list Glob patterns for files to exclude from review