API Service Related Configuration
While api-server.js prioritizes reading configurations from command-line arguments, it also supports loading default values from a config.json file (if present). Below are key configuration items related to the API service and their descriptions:
| Parameter Name | Description | Default Value |
|---|---|---|
REQUIRED_API_KEY | The key required for clients to access the API. | "123456" |
SERVER_PORT | The port on which the API service listens. | 3000 |
HOST | The address on which the API service listens. | "localhost" |
MODEL_PROVIDER | The AI model provider. Optional values include: - gemini-cli-oauth: Uses Google Gemini CLI authentication. - openai-custom: Uses OpenAI compatible API. - claude-custom: Uses Claude compatible API.- kiro-api: Uses Kiro/CodeWhisperer to access Claude. | "gemini-cli-oauth" |
OPENAI_API_KEY | OpenAI API key when MODEL_PROVIDER is openai-custom. | null |
OPENAI_BASE_URL | OpenAI API base URL when MODEL_PROVIDER is openai-custom. | null |
CLAUDE_API_KEY | Claude API key when MODEL_PROVIDER is claude-custom. | null |
CLAUDE_BASE_URL | Claude API base URL when MODEL_PROVIDER is claude-custom. | null |
GEMINI_OAUTH_CREDS_BASE64 | Base64 string of Gemini OAuth credentials when MODEL_PROVIDER is gemini-cli-oauth. | null |
GEMINI_OAUTH_CREDS_FILE_PATH | Path to the Gemini OAuth credentials JSON file when MODEL_PROVIDER is gemini-cli-oauth. | null |
KIRO_OAUTH_CREDS_BASE64 | Base64 string of Kiro OAuth credentials when MODEL_PROVIDER is kiro-api. | null |
KIRO_OAUTH_CREDS_FILE_PATH | Path to the Kiro OAuth credentials JSON file when MODEL_PROVIDER is kiro-api. | null |
PROJECT_ID | Google Cloud Project ID when MODEL_PROVIDER is gemini-cli-oauth. | null |
SYSTEM_PROMPT_FILE_PATH | Path to the file containing system prompts. | "input_system_prompt.txt" |
SYSTEM_PROMPT_MODE | System prompt processing mode: overwrite or append. | "overwrite" |
PROMPT_LOG_BASE_NAME | Base name for prompt log files. | "prompt_log" |
PROMPT_LOG_MODE | Prompt log mode: console, file, or none. | "none" |
REQUEST_MAX_RETRIES | Maximum number of retries for API requests upon failure. | 3 |
REQUEST_BASE_DELAY | Base delay time (in milliseconds) for API request retries, which increases exponentially with each retry. | 1000 |
CRON_NEAR_MINUTES | The number of minutes before token expiration to trigger a refresh. | 5 |
CRON_REFRESH_TOKEN | Whether to enable automatic token refresh. | true |
Example config.json (can be placed in the project root directory):
json
{
"REQUIRED_API_KEY": "my_super_secret_key",
"SERVER_PORT": 8005,
"HOST": "127.0.0.1",
"MODEL_PROVIDER": "gemini-cli-oauth",
"GEMINI_OAUTH_CREDS_FILE_PATH": "/path/to/credentials.json",
"PROJECT_ID": "your-gcp-project-id",
"SYSTEM_PROMPT_FILE_PATH": "./custom_system_prompt.txt",
"SYSTEM_PROMPT_MODE": "append",
"PROMPT_LOG_MODE": "file",
"PROMPT_LOG_BASE_NAME": "aiclient_2_api_logs",
"REQUEST_MAX_RETRIES": 5,
"REQUEST_BASE_DELAY": 2000,
"CRON_NEAR_MINUTES": 5,
"CRON_REFRESH_TOKEN": true
}By appropriately configuring config.json, you can finely control the behavior of AIClient-2-API to better suit your needs and hardware environment.