configs

📘 AI Review Configuration

AI Review supports multiple configuration formats and sources. All of them are automatically detected at runtime.

---

📂 Supported formats

• YAML (recommended): .ai-review.yaml
• JSON: .ai-review.json
• ENV: .env

👉 You can combine formats: values are loaded in order of priority.

---

📑 Load priority

1. YAML (.ai-review.yaml or path from AI_REVIEW_CONFIG_FILE_YAML)
2. JSON (.ai-review.json or path from AI_REVIEW_CONFIG_FILE_JSON)
3. ENV (.env or path from AI_REVIEW_CONFIG_FILE_ENV)
4. Environment variables (LLM__PROVIDER=OPENAI, etc.)
5. Initialization arguments (if used as a library)

---

⚙️ Override file paths

You can override default config locations using environment variables:

• AI_REVIEW_CONFIG_FILE_YAML — path to .yaml config
• AI_REVIEW_CONFIG_FILE_JSON — path to .json config
• AI_REVIEW_CONFIG_FILE_ENV — path to .env

By default, configs are loaded from the project root.

---

📘 Examples

• .ai-review.yaml — main YAML config with comments
• .ai-review.json — JSON config example
• .env.example — ENV config example

---

🧠 OpenAI reasoning

OpenAI reasoning options can be configured for models that use the Responses API. The object is optional and is omitted from requests when it is not configured. It is not sent to models that use Chat Completions.

YAML

llm:
  provider: OPENAI
  meta:
    model: gpt-5.6-sol
    reasoning:
      effort: medium
      summary: concise
      context: all_turns
      mode: standard

JSON

{
  "llm": {
    "provider": "OPENAI",
    "meta": {
      "model": "gpt-5.6-sol",
      "reasoning": {
        "effort": "medium",
        "summary": "concise",
        "context": "all_turns",
        "mode": "standard"
      }
    }
  }
}

Environment variables

LLM__PROVIDER=OPENAI
LLM__META__MODEL=gpt-5.6-sol
LLM__META__REASONING__EFFORT=medium
LLM__META__REASONING__SUMMARY=concise
LLM__META__REASONING__CONTEXT=all_turns
LLM__META__REASONING__MODE=standard

Options

Option	Supported values	Description
effort	none, minimal, low, medium, high, xhigh, max	Controls how many reasoning tokens the model uses. Not every model supports every value.
summary	auto, concise, detailed	Requests a summary of the model reasoning.
context	auto, current_turn, all_turns	Controls which reasoning items are provided to the model on later turns.
mode	standard, pro	Selects the reasoning execution mode. Model support may vary.
generate_summary	auto, concise, detailed	Deprecated OpenAI option. Use summary instead.

See the OpenAI Responses API reference for model-specific support and current behavior.

---

🔍 Tips

• Use YAML for most projects — it’s human-friendly and supports comments.
• JSON is convenient for automation (e.g., CI/CD pipelines).
• ENV is useful for local development and quick overrides.