As a software engineer, I need two types of tools for my daily tasks: one for working inside a file and the other outside.
This AI tool assists you with all the tasks beyond file editing on your system complementing Continue, Cline and the like.
git clone https://github.com/qiangli/ai.git
cd ai
make build
make install
ai [OPTIONS] AGENT [message...]
ai @ask "What is the capital of France?"
git diff origin main|ai @ask generate commit message for git
ai / What tools could I use to search for a pattern in files
ai --help
No API calls will be made in dry-run
mode.
ai --dry-run --dry-run-content "fake data" ...
Default system prompts can be replaced for testing and evaluation.
ai --role "system" --role-content "custom prompt" ...
//https://github.com/jfcg/sorty/issues/6
// go test -c -o bin/test ./internal/db
{
"name": "TestGetByVector",
"type": "go",
"request": "launch",
"mode": "exec",
"program": "./bin/test",
"args": ["-test.run", "^TestGetByVector$"],
},
ai --verbose ...
$ ai
AI Command Line Tool
Usage:
ai [OPTIONS] [AGENT] message...
Examples:
ai what is fish
ai / what is fish
ai @ask what is fish
Use ai help [info|agents|commands|tools] for more details.
$ ai help
AI Command Line Tool
Usage:
ai [OPTIONS] AGENT message...
There are multiple ways to interact with the AI tool.
+ Command line input:
ai AGENT what is fish?
+ Read from standard input:
ai AGENT -
Ctrl+D to send, Ctrl+C to cancel.
+ Here document:
ai AGENT <<eof
what is the weather today?
eof
+ Piping input:
echo "What is Unix?" | ai AGENT
git diff origin main | ai AGENT [message...]
curl -sL https://en.wikipedia.org/wiki/Artificial_intelligence | head -100 | ai AGENT [message...]
+ File redirection:
ai AGENT [message...] < file.txt
+ Reading from system clipboard:
ai AGENT [message...] =
Use system copy (Ctrl+C on Unix) to send selected contents.
Ctrl+C to cancel.
+ Composing with text editor:
export AI_EDITOR=nano # default: vi
ai AGENT
Agent:
/[command] [message...] Get help with system command and shell scripting tasks
@[agent/command] [message...] Engage specialist agents for assistance with complex tasks
Miscellaneous:
setup Setup the AI configuration
Options:
--api-key string LLM API key
--base-url string LLM Base URL (default "https://api.openai.com/v1/")
--config string config file (default "/Users/qiang.li/.ai/config.yaml")
--doc-template string Document template file
--editor string Specify editor to use (default "vi")
--file string Read input from files. May be given multiple times to add multiple file content
--format string Output format, must be either raw or markdown. (default "markdown")
-h, --help help for ai
--image-model string Image LLM model (default "dall-e-3")
--l1-model string Level1 basic LLM model (default "gpt-4o-mini")
--l2-model string Level2 standard LLM model (default "gpt-4o")
--l3-model string Level3 advanced LLM model (default "o1-mini")
--max-time int Max number of seconds for timeout (default 3600)
--max-turns int Max number of turns (default 32)
--mcp-server-url string MCP server URL (default "http://localhost:58080/sse")
--model string LLM model (default "gpt-4o")
-o, --output string Save final response to a file.
--pb-read Read input from the clipboard. Alternatively, append '=' to the command
--pb-write Copy output to the clipboard. Alternatively, append '=+' to the command
--quiet Operate quietly
--verbose Show debugging information
-w, --workspace string Workspace directory
Environment variables:
AI_API_KEY, AI_BASE_URL, AI_CONFIG, AI_DOC_TEMPLATE, AI_DRY_RUN, AI_DRY_RUN_CONTENT, AI_EDITOR, AI_FILE, AI_FORMAT, AI_HELP, AI_IMAGE_API_KEY, AI_IMAGE_BASE_URL, AI_IMAGE_MODEL, AI_IMAGE_VIEWER, AI_INTERACTIVE, AI_L1_API_KEY, AI_L1_BASE_URL, AI_L1_MODEL, AI_L2_API_KEY, AI_L2_BASE_URL, AI_L2_MODEL, AI_L3_API_KEY, AI_L3_BASE_URL, AI_L3_MODEL, AI_LOG, AI_MAX_TIME, AI_MAX_TURNS, AI_MCP_SERVER_URL, AI_MESSAGE, AI_MODEL, AI_NO_META_PROMPT, AI_OUTPUT, AI_PB_READ, AI_PB_WRITE, AI_QUIET, AI_ROLE, AI_ROLE_PROMPT, AI_SQL_DB_HOST, AI_SQL_DB_NAME, AI_SQL_DB_PASSWORD, AI_SQL_DB_PORT, AI_SQL_DB_USERNAME, AI_TRACE, AI_VERBOSE, AI_WORKSPACE
$ ai help agents
Available agents:
aider: Integrate LLMs for collaborative coding, refactoring, bug fixing, and test development.
ask: Deliver concise, reliable answers on a wide range of topics.
doc: Create a polished document by integrating draft materials into the provided template.
draw: Generate images based on user input, providing visual representations of text-based descriptions.
eval: Evaluate and test tools.
git: Automate git commit message creation for clarity and consistency in version control
git/conventional: Generate concise git commit messages based on the provided diffs using the Conventional Commits specification
git/short: Generate concise, one-line git commit messages based on the provided diffs.
gptr: Deliver live, realtime, accurate, relevant insights from diverse online sources.
launch: Dispatch to the most appropriate agent based on the user's input.
meta-prompt: Generates a system prompt based on the user's input.
oh: Engineering assistant promoting incremental development and detailed refactoring support.
pr: Enhance PR management with automated summaries, reviews, suggestions, and changelog updates.
pr/changelog: Update the CHANGELOG.md file with the PR changes
pr/describe: Generate PR description - title, type, summary, code walkthrough and labels
pr/improve: Provide code suggestions for improving the PR
pr/review: Give feedback about the PR, possible issues, security concerns, review effort and more
script: Assist with scripting, command execution, and troubleshooting shell tasks.
sql: Streamline SQL query generation, helping users derive insights without SQL expertise.
workspace: Determines the user's workspace based on user's input.
/ is shorthand for @script
Not sure which agent to use? Simply enter your message and AI will choose the most appropriate one for you:
ai "message..."