13 | 
18 | 
23 |
28 | 
33 | 4 | AI in the command line. 5 |
6 | 7 |
8 |
23 |
36 | 
37 |
38 | You just need to write a comment or variable name and the AI will write the corresponding code. 39 |
40 | 41 | 42 | ## What is it? 43 | 44 | This is a ZSH plugin that enables you to use AI powered code completion in the command line. It now supports both OpenAI's Codex and Google's Generative AI (Gemini). OpenAI Codex is the AI that also powers GitHub Copilot, while Gemini is Google's advanced language model. 45 | 46 | ## How do I install it? 47 | 48 | ### Manual Installation 49 | 50 | 1. Install the OpenAI package, the Google package, or boto3. 51 | 52 | ```bash 53 | pip3 install openai 54 | ``` 55 | 56 | or 57 | 58 | ```bash 59 | pip3 install google-generativeai 60 | ``` 61 | 62 | or 63 | 64 | ```bash 65 | pip3 install boto3 66 | ``` 67 | 68 | 2. Download the ZSH plugin. 69 | 70 | ```bash 71 | git clone https://github.com/tom-doerr/zsh_codex.git ~/.oh-my-zsh/custom/plugins/zsh_codex 72 | ``` 73 | 74 | 3. Add the following to your `.zshrc` file. 75 | 76 | Using oh-my-zsh: 77 | 78 | ```bash 79 | plugins=(zsh_codex) 80 | bindkey '^X' create_completion 81 | ``` 82 | 83 | Without oh-my-zsh: 84 | 85 | ```bash 86 | # in your/custom/path you need to have a "plugins" folder and in there you clone the repository as zsh_codex 87 | export ZSH_CUSTOM="your/custom/path" 88 | source "$ZSH_CUSTOM/plugins/zsh_codex/zsh_codex.plugin.zsh" 89 | bindkey '^X' create_completion 90 | ``` 91 | 92 | 4. Create a file called `zsh_codex.ini` in `~/.config`. 93 | Example: 94 | 95 | ```ini 96 | ; Primary service configuration 97 | ; Set 'service' to match one of the defined sections below. 98 | [service] 99 | service = groq_service 100 | 101 | ; Example configuration for a self-hosted Ollama service. 102 | [my_ollama] 103 | api_type = openai 104 | api_key = dummy_key 105 | model = llama3.1 106 | base_url = http://localhost:11434/v1 107 | 108 | ; OpenAI service configuration 109 | ; Provide the 'api_key' and specify a 'model' if needed. 110 | [openai_service] 111 | api_type = openai 112 | api_key =
142 |
143 | ## Troubleshooting
144 |
145 | ### Unhandled ZLE widget 'create_completion'
146 |
147 | ```
148 | zsh-syntax-highlighting: unhandled ZLE widget 'create_completion'
149 | zsh-syntax-highlighting: (This is sometimes caused by doing `bindkey
175 | 
176 |
196 | 
197 |
198 |
199 | 200 | 201 | --- 202 | 203 | [Fish Version](https://github.com/tom-doerr/codex.fish) 204 | 205 | [Traffic Statistics](https://tom-doerr.github.io/github_repo_stats_data/tom-doerr/zsh_codex/latest-report/report.html) 206 | -------------------------------------------------------------------------------- /create_completion.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python3 2 | 3 | import argparse 4 | import sys 5 | 6 | from services.services import ClientFactory 7 | 8 | 9 | def main(): 10 | parser = argparse.ArgumentParser( 11 | description="Generate command completions using AI." 12 | ) 13 | parser.add_argument( 14 | "cursor_position", type=int, help="Cursor position in the input buffer" 15 | ) 16 | args = parser.parse_args() 17 | 18 | client = ClientFactory.create() 19 | 20 | # Read the input prompt from stdin. 21 | buffer = sys.stdin.read() 22 | zsh_prefix = "#!/bin/zsh\n\n" 23 | buffer_prefix = buffer[: args.cursor_position] 24 | buffer_suffix = buffer[args.cursor_position :] 25 | full_command = zsh_prefix + buffer_prefix + buffer_suffix 26 | 27 | completion = client.get_completion(full_command) 28 | 29 | if completion.startswith(zsh_prefix): 30 | completion = completion[len(zsh_prefix) :] 31 | 32 | line_prefix = buffer_prefix.rsplit("\n", 1)[-1] 33 | # Handle all the different ways the command can be returned 34 | for prefix in [buffer_prefix, line_prefix]: 35 | if completion.startswith(prefix): 36 | completion = completion[len(prefix) :] 37 | break 38 | 39 | if buffer_suffix and completion.endswith(buffer_suffix): 40 | completion = completion[: -len(buffer_suffix)] 41 | 42 | completion = completion.strip("\n") 43 | if line_prefix.strip().startswith("#"): 44 | completion = "\n" + completion 45 | 46 | sys.stdout.write(completion) 47 | 48 | 49 | if __name__ == "__main__": 50 | main() 51 | -------------------------------------------------------------------------------- /services/services.py: -------------------------------------------------------------------------------- 1 | import os 2 | import sys 3 | from abc import ABC, abstractmethod 4 | from configparser import ConfigParser 5 | 6 | CONFIG_DIR = os.getenv("XDG_CONFIG_HOME", os.path.expanduser("~/.config")) 7 | CONFIG_PATH = os.path.join(CONFIG_DIR, "zsh_codex.ini") 8 | 9 | 10 | class BaseClient(ABC): 11 | """Base class for all clients""" 12 | 13 | api_type: str = None 14 | system_prompt = "You are a zsh shell expert, please help me complete the following command, you should only output the completed command, no need to include any other explanation. Do not put completed command in a code block." 15 | 16 | @abstractmethod 17 | def get_completion(self, full_command: str) -> str: 18 | pass 19 | 20 | 21 | class OpenAIClient(BaseClient): 22 | """ 23 | config keys: 24 | - api_type="openai" 25 | - api_key (required) 26 | - base_url (optional): defaults to "https://api.openai.com/v1". 27 | - organization (optional): defaults to None 28 | - model (optional): defaults to "gpt-4o-mini" 29 | - temperature (optional): defaults to 1.0. 30 | """ 31 | 32 | api_type = "openai" 33 | default_model = os.getenv("OPENAI_DEFAULT_MODEL", "gpt-4o-mini") 34 | 35 | def __init__(self, config: dict): 36 | try: 37 | from openai import OpenAI 38 | except ImportError: 39 | print( 40 | "OpenAI library is not installed. Please install it using 'pip install openai'" 41 | ) 42 | sys.exit(1) 43 | 44 | self.config = config 45 | self.config["model"] = self.config.get("model", self.default_model) 46 | self.client = OpenAI( 47 | api_key=self.config["api_key"], 48 | base_url=self.config.get("base_url", "https://api.openai.com/v1"), 49 | organization=self.config.get("organization"), 50 | ) 51 | 52 | def get_completion(self, full_command: str) -> str: 53 | response = self.client.chat.completions.create( 54 | model=self.config["model"], 55 | messages=[ 56 | {"role": "system", "content": self.system_prompt}, 57 | {"role": "user", "content": full_command}, 58 | ], 59 | temperature=float(self.config.get("temperature", 1.0)), 60 | ) 61 | return response.choices[0].message.content 62 | 63 | 64 | class GoogleGenAIClient(BaseClient): 65 | """ 66 | config keys: 67 | - api_type="gemeni" 68 | - api_key (required) 69 | - model (optional): defaults to "gemini-1.5-pro-latest" 70 | """ 71 | 72 | api_type = "gemeni" 73 | default_model = os.getenv("GOOGLE_GENAI_DEFAULT_MODEL", "gemini-1.5-pro-latest") 74 | 75 | def __init__(self, config: dict): 76 | try: 77 | import google.generativeai as genai 78 | except ImportError: 79 | print( 80 | "Google Generative AI library is not installed. Please install it using 'pip install google-generativeai'" 81 | ) 82 | sys.exit(1) 83 | 84 | self.config = config 85 | genai.configure(api_key=self.config["api_key"]) 86 | self.config["model"] = config.get("model", self.default_model) 87 | self.model = genai.GenerativeModel(self.config["model"]) 88 | 89 | def get_completion(self, full_command: str) -> str: 90 | chat = self.model.start_chat(history=[]) 91 | prompt = f"{self.system_prompt}\n\n{full_command}" 92 | response = chat.send_message(prompt) 93 | return response.text 94 | 95 | 96 | class GroqClient(BaseClient): 97 | """ 98 | config keys: 99 | - api_type="groq" 100 | - api_key (required) 101 | - model (optional): defaults to "llama-3.2-11b-text-preview" 102 | - temperature (optional): defaults to 1.0. 103 | """ 104 | 105 | api_type = "groq" 106 | default_model = os.getenv("GROQ_DEFAULT_MODEL", "llama-3.2-11b-text-preview") 107 | 108 | def __init__(self, config: dict): 109 | try: 110 | from groq import Groq 111 | except ImportError: 112 | print( 113 | "Groq library is not installed. Please install it using 'pip install groq'" 114 | ) 115 | sys.exit(1) 116 | 117 | self.config = config 118 | self.config["model"] = self.config.get("model", self.default_model) 119 | self.client = Groq( 120 | api_key=self.config["api_key"], 121 | ) 122 | 123 | def get_completion(self, full_command: str) -> str: 124 | response = self.client.chat.completions.create( 125 | model=self.config["model"], 126 | messages=[ 127 | {"role": "system", "content": self.system_prompt}, 128 | {"role": "user", "content": full_command}, 129 | ], 130 | temperature=float(self.config.get("temperature", 1.0)), 131 | ) 132 | return response.choices[0].message.content 133 | 134 | 135 | class MistralClient(BaseClient): 136 | """ 137 | config keys: 138 | - api_type="mistral" 139 | - api_key (required) 140 | - model (optional): defaults to "codestral-latest" 141 | - temperature (optional): defaults to 1.0. 142 | """ 143 | 144 | api_type = "mistral" 145 | default_model = os.getenv("MISTRAL_DEFAULT_MODEL", "codestral-latest") 146 | 147 | def __init__(self, config: dict): 148 | try: 149 | from mistralai import Mistral 150 | except ImportError: 151 | print( 152 | "Mistral library is not installed. Please install it using 'pip install mistralai'" 153 | ) 154 | sys.exit(1) 155 | 156 | self.config = config 157 | self.config["model"] = self.config.get("model", self.default_model) 158 | self.client = Mistral( 159 | api_key=self.config["api_key"], 160 | ) 161 | 162 | def get_completion(self, full_command: str) -> str: 163 | response = self.client.chat.complete( 164 | model=self.config["model"], 165 | messages=[ 166 | {"role": "system", "content": self.system_prompt}, 167 | {"role": "user", "content": full_command}, 168 | ], 169 | temperature=float(self.config.get("temperature", 1.0)), 170 | ) 171 | return response.choices[0].message.content 172 | 173 | class AmazonBedrock(BaseClient): 174 | """ 175 | config keys: 176 | - api_type="bedrock" 177 | - aws_region (optional): defaults to environment variable AWS_REGION 178 | - aws_access_key_id (optional): defaults to environment variable AWS_ACCESS_KEY_ID 179 | - aws_secret_access_key (optional): defaults to environment variable AWS_SECRET_ACCESS_KEY 180 | - aws_session_token (optional): defaults to environment variable AWS_SESSION_TOKEN 181 | - model (optional): defaults to "anthropic.claude-3-5-sonnet-20240620-v1:0" or environment variable BEDROCK_DEFAULT_MODEL 182 | - temperature (optional): defaults to 1.0. 183 | """ 184 | 185 | api_type = "bedrock" 186 | default_model = os.getenv("BEDROCK_DEFAULT_MODEL", "anthropic.claude-3-5-sonnet-20240620-v1:0") 187 | 188 | def __init__(self, config: dict): 189 | try: 190 | import boto3 191 | except ImportError: 192 | print( 193 | "Boto3 library is not installed. Please install it using 'pip install boto3'" 194 | ) 195 | sys.exit(1) 196 | 197 | self.config = config 198 | self.config["model"] = self.config.get("model", self.default_model) 199 | 200 | session_kwargs = {} 201 | if "aws_region" in config: 202 | session_kwargs["region_name"] = config["aws_region"] 203 | if "aws_access_key_id" in config: 204 | session_kwargs["aws_access_key_id"] = config["aws_access_key_id"] 205 | if "aws_secret_access_key" in config: 206 | session_kwargs["aws_secret_access_key"] = config["aws_secret_access_key"] 207 | if "aws_session_token" in config: 208 | session_kwargs["aws_session_token"] = config["aws_session_token"] 209 | 210 | self.client = boto3.client("bedrock-runtime", **session_kwargs) 211 | 212 | def get_completion(self, full_command: str) -> str: 213 | import json 214 | 215 | messages = [ 216 | {"role": "user", "content": full_command} 217 | ] 218 | 219 | # Format request body based on model type 220 | if "claude" in self.config["model"].lower(): 221 | body = { 222 | "anthropic_version": "bedrock-2023-05-31", 223 | "max_tokens": 1000, 224 | "system": self.system_prompt, 225 | "messages": messages, 226 | "temperature": float(self.config.get("temperature", 1.0)) 227 | } 228 | else: 229 | raise ValueError(f"Unsupported model: {self.config['model']}") 230 | 231 | response = self.client.invoke_model( 232 | modelId=self.config["model"], 233 | body=json.dumps(body) 234 | ) 235 | 236 | response_body = json.loads(response['body'].read()) 237 | return response_body["content"][0]["text"] 238 | 239 | 240 | 241 | class ClientFactory: 242 | api_types = [OpenAIClient.api_type, GoogleGenAIClient.api_type, GroqClient.api_type, MistralClient.api_type, AmazonBedrock.api_type] 243 | 244 | @classmethod 245 | def create(cls): 246 | config_parser = ConfigParser() 247 | config_parser.read(CONFIG_PATH) 248 | service = config_parser["service"]["service"] 249 | try: 250 | config = {k: v for k, v in config_parser[service].items()} 251 | except KeyError: 252 | raise KeyError(f"Config for service {service} is not defined") 253 | 254 | api_type = config["api_type"] 255 | match api_type: 256 | case OpenAIClient.api_type: 257 | return OpenAIClient(config) 258 | case GoogleGenAIClient.api_type: 259 | return GoogleGenAIClient(config) 260 | case GroqClient.api_type: 261 | return GroqClient(config) 262 | case MistralClient.api_type: 263 | return MistralClient(config) 264 | case AmazonBedrock.api_type: 265 | return AmazonBedrock(config) 266 | case _: 267 | raise KeyError( 268 | f"Specified API type {api_type} is not one of the supported services {cls.api_types}" 269 | ) 270 | -------------------------------------------------------------------------------- /zsh_codex.plugin.zsh: -------------------------------------------------------------------------------- 1 | #!/bin/zsh 2 | 3 | # This ZSH plugin reads the text from the current buffer 4 | # and uses a Python script to complete the text. 5 | api="openai" 6 | 7 | _ZSH_CODEX_REPO=$(dirname $0) 8 | 9 | create_completion() { 10 | # Get the text typed until now. 11 | local text=$BUFFER 12 | if [[ "$ZSH_CODEX_PREEXECUTE_COMMENT" == "true" ]]; then 13 | text="$(echo -n "echo \"$text\"" | zsh)" 14 | fi 15 | local ZSH_CODEX_PYTHON="${ZSH_CODEX_PYTHON:-python3}" 16 | local completion=$(echo -n "$text" | $ZSH_CODEX_PYTHON $_ZSH_CODEX_REPO/create_completion.py $CURSOR) 17 | local text_before_cursor=${BUFFER:0:$CURSOR} 18 | local text_after_cursor=${BUFFER:$CURSOR} 19 | 20 | # Add completion to the current buffer. 21 | BUFFER="${text_before_cursor}${completion}${text_after_cursor}" 22 | 23 | # Put the cursor at the end of the completion 24 | CURSOR=$((CURSOR + ${#completion})) 25 | } 26 | 27 | # Bind the create_completion function to a key. 28 | zle -N create_completion 29 | # You may want to add a key binding here, e.g.: 30 | # bindkey '^X^E' create_completion 31 | --------------------------------------------------------------------------------