refactor make_scripts, add functionality for saving scripts to files

echo_crafter/config/__init__.py

@@ -1,3 +1,4 @@
 from .config import Config
+from .openai_config import OpenAIConfig
 
-__all__= ['Config']
+__all__= ['Config', 'OpenAIConfig']

echo_crafter/config/computer-commands.yml

@@ -0,0 +1,71 @@
+context:
+  expressions:
+    getScript:
+      - "@promptVerb (me) [a, an] ($programmingLanguage:language) script"
+    answerQuestion:
+      - "@promptVerb (me) [an, the, the following] answer to [this, that, the,
+        the following] question"
+      - Answer (me) [this, the, the following, that] question
+    simplyTranscribe:
+      - Send (this, that, the following, what follows) to (the) keyboard
+      - Write [this, that, the following, what follows]
+      - Transcribe
+    focusWindow:
+      - "[select, go to, focus the] window number
+        $pv.SingleDigitInteger:windowNumber"
+      - "[go to, focus, switch to] the [$pv.SingleDigitOrdinal:windowNumber,
+        $windowName:windowName] window"
+    openWindow:
+      - "[open, launch] (a) new emacs window"
+    setVolume:
+      - $volumeSetting:volumeSetting (the volume)
+      - (Set) (the) [volume, sound] (to) $pv.Percent:percentage
+    cancel:
+      - (no) wait a second
+      - (no) cancel
+      - (no) never mind
+  slots:
+    programmingLanguage:
+      - stump
+      - lisp
+      - e lisp
+      - c plus plus
+      - javascript
+      - emacs
+      - shell
+      - python
+    promptType:
+      - question
+      - script
+    windowName:
+      - internet
+      - web
+      - google search
+      - sound control
+      - sound
+      - chrome
+      - fire fox
+      - kitty
+      - browser
+      - emacs
+      - shell
+    volumeSetting:
+      - unmute
+      - mute
+  macros:
+    promptVerb:
+      - I need
+      - I want
+      - generate
+      - give
+      - write
+    articlesAndDemonstratives:
+      - that
+      - this
+      - the
+      - an
+      - a
+    generate:
+      - give
+      - write
+      - generate

echo_crafter/config/config.py

@@ -38,14 +38,16 @@ class _Config(TypedDict):
     PYTHON_PACKAGES: str
 
     PICOVOICE_API_KEY: str
+    PORCUPINE_KEYWORD_FILE: str
     CHEETAH_MODEL_FILE: str
+    LEOPARD_MODEL_FILE: str
     RHINO_CONTEXT_FILE: str
     FRAME_LENGTH: int
+    BUFFER_SIZE: int
     ENDPOINT_DURATION_SEC: float
 
-    TRANSCRIPT_BEGIN_WAV: str
-    TRANSCRIPT_SUCCESS_WAV: str
-    SOCKET_PATH: str
+    WAKE_WORD_DETECTED_WAV: str
+    INTENT_SUCCESS_WAV: str
 
 
 Config: _Config = {
@@ -54,12 +56,14 @@ class _Config(TypedDict):
     "PYTHON_PACKAGES":        build_path(".venv/lib/python3.11/site-packages/python_packages"),
 
     "PICOVOICE_API_KEY":      str(os.getenv('PICOVOICE_API_KEY')),
-    "CHEETAH_MODEL_FILE":     build_path("data/speech-command-cheetah-v2.pv"),
-    "RHINO_CONTEXT_FILE":     build_path("data/computer-commands_en_linux_v3_0_0.rhn"),
+    "PORCUPINE_KEYWORD_FILE": build_path("data/computer_linux.ppn"),
+    "CHEETAH_MODEL_FILE":     build_path("data/speech-command-cheetah.pv"),
+    "LEOPARD_MODEL_FILE":     build_path("data/speech-command-leopard.pv"),
+    "RHINO_CONTEXT_FILE":     build_path("data/computer-commands_en_linux.rhn"),
     "FRAME_LENGTH":           512,
+    "BUFFER_SIZE":            512 * 2 * 30,  # 30 seconds
     "ENDPOINT_DURATION_SEC":  1.5,
 
-    "TRANSCRIPT_BEGIN_WAV":   build_path("data/transcript_begin.wav"),
-    "TRANSCRIPT_SUCCESS_WAV": build_path("data/transcript_success.wav"),
-    "SOCKET_PATH":            str(Path(os.environ.get('EC_SOCKET_FILE', '/tmp/echo-crafter.sock')))
+    "WAKE_WORD_DETECTED_WAV": build_path("data/transcript_begin.wav"),
+    "INTENT_SUCCESS_WAV":     build_path("data/transcript_success.wav"),
 }

echo_crafter/config/openai_config.py

@@ -0,0 +1,88 @@
+"""@file Configuration for the prompt package."""
+
+from dataclasses import dataclass
+import os
+from subprocess import check_output
+from pathlib import Path
+from typing import Literal, TypedDict
+
+
+def get_openai_api_key() -> str:
+    """Get the OpenAI API key from the password store."""
+    api_key = os.environ.get('OPENAI_API_KEY')
+    if api_key is None:
+        api_key = str(check_output(["pass", "openai/echo-crafter"]), encoding="utf-8").rstrip()
+    if not api_key:
+        raise ValueError("No OpenAI API key found.")
+    return api_key
+
+def get_openai_log_path() -> str:
+    """Get the path to the OpenAI log file."""
+    xdg_data_dir = Path(os.getenv("XDG_DATA_HOME") or Path.home()/".local/share")
+    return str(xdg_data_dir/"openai/new_logs.jsonl")
+
+def get_history_file() -> str:
+    """Get the path to the OpenAI history file."""
+    xdg_data_dir = Path(os.getenv("XDG_DATA_HOME") or Path.home()/".local/share")
+    return str(xdg_data_dir/"openai/history.jsonl")
+
+
+@dataclass
+class Pricing(TypedDict):
+    """Pricing information for a model.
+
+    The pricing information is given in units of USD/(1 million tokens).
+    """
+
+    input: float
+    output: float
+
+
+@dataclass
+class Model(TypedDict):
+    """A model object."""
+
+    name: Literal[
+        "gpt-4-0125-preview",
+        "gpt-3.5-turbo-0125",
+        "text-embedding-3-small",
+        "text-embedding-3-large"
+    ]
+    pricing: Pricing
+
+
+@dataclass
+class _OpenAIConfig(TypedDict):
+    """Configuration for the prompt package."""
+
+    API_KEY: str
+    LOG_FILE: str
+    HISTORY_FILE: str
+    DEFAULT_MODEL: str
+    MODELS: list[Model]
+
+
+OpenAIConfig: _OpenAIConfig = {
+    "API_KEY": get_openai_api_key(),
+    "LOG_FILE": get_openai_log_path(),
+    "HISTORY_FILE": get_history_file(),
+    "MODELS": [
+        {
+            "name": "gpt-4-0125-preview",
+            "pricing": {"input": 10, "output": 30}
+        },
+        {
+            "name": "gpt-3.5-turbo-0125",
+            "pricing": {"input": 0.5, "output": 1.5}
+        },
+        {
+            "name": "text-embedding-3-small",
+            "pricing": {"input": 0.02, "output": 0.02}
+        },
+        {
+            "name": "text-embedding-3-large",
+            "pricing": {"input": 0.13, "output": 0.13}
+        }
+    ],
+    "DEFAULT_MODEL": "gpt-4-0125-preview"
+}

echo_crafter/logger/__init__.py

@@ -7,6 +7,7 @@
 from pathlib import Path
 
 LOG_FILE = Path(os.environ.get('EC_LOG_DIR', '')) / "transcripts.jsonl"
+only_true_once = True
 
 class CustomRecord(logging.LogRecord):
     """Custom LogRecord class with a timestamp attribute."""

echo_crafter/prompts/__init__.py

@@ -0,0 +1,3 @@
+from .openaiAPI import OpenAIAPI
+
+__all__ = ['OpenAIAPI']

echo_crafter/prompts/make_script.py

@@ -0,0 +1,133 @@
+#!/usr/bin/env python3
+
+import argparse
+import sys
+import re
+sys.path.append('os.path.abspath(os.path.join(os.path.dirname(__file__), "..", ".."))')
+from rich.console import Console
+from rich.markdown import Markdown
+from prompt_toolkit import PromptSession, prompt
+from prompt_toolkit.history import FileHistory
+from echo_crafter.config import OpenAIConfig
+from echo_crafter.prompts import OpenAIAPI
+from echo_crafter.prompts.templates import (
+    PYTHON_BASE_PROMPT,
+    SHELL_BASE_PROMPT,
+    ELISP_BASE_PROMPT
+)
+
+
+def extract_sections_from_markdown(markdown_text: str, sections: list[str]) -> dict:
+     """
+     Extracts specific sections from a markdown document.
+
+     Args:
+         - markdown_text (str): The markdown text to be parsed.
+         - sections (list[str]): A list of section names to be extracted.
+
+     Returns:
+         dict: A dictionary containing the extracted content as {'section': content}.
+     """
+     # Regular expressions for extracting the sections
+     section_patterns = [f"## {section}:\n(.*)\n\n## {sections[i+1]}" for i, section in enumerate(sections[:-1])]
+     section_patterns.append(f"## {sections[-1]}:\n(.*)")
+     # Extracting the content
+     section_contents = [re.search(section_pattern, markdown_text, re.DOTALL) for section_pattern in section_patterns]
+
+     # Preparing the result dictionary
+     result = {
+         section: (section_contents[i].group(1).strip() if section_contents[i] is not None else "") for i, section in enumerate(sections)  # type: ignore
+     }
+
+     return result
+
+
+def extract_code_block(content):
+    """Parse the answer content."""
+    match = re.search(r"```(?:\S+)(.*)```", content, re.DOTALL)
+    if match is not None:
+        return match.group(1).strip()
+
+
+def main():
+    parser = argparse.ArgumentParser(description='Process some arguments.')
+
+    parser.add_argument('command',          nargs='?',  help='Optional command')
+    parser.add_argument('--model',          type=str,   help='Model to use.', default=OpenAIConfig['DEFAULT_MODEL'])
+    parser.add_argument('--language',       type=str,   help='Language to use.', default='python')
+    parser.add_argument('--temperature',    type=float, help='Sampling temperature to use [floating point number between 0 and 2]', default=0.2)
+    parser.add_argument('--max_new_tokens', type=int,   help='Specify an upper bound on number of tokens generated per response.')
+
+    args = parser.parse_args()
+
+    match args.language:
+        case 'python':
+            base_prompt = PYTHON_BASE_PROMPT
+            extension = ".py"
+            shebang = "#!/usr/bin/env python3\n\n"
+        case 'shell':
+            shebang = "#!/usr/bin/env zsh\n\n"
+            extension = ".sh"
+            base_prompt = SHELL_BASE_PROMPT
+        case 'elisp':
+            shebang = "#!/usr/bin/env emacs --script\n\n"
+            extension = ".el"
+            base_prompt = ELISP_BASE_PROMPT
+        case _:
+            raise ValueError(f"Unsupported language: {args.language}")
+
+    console = Console()
+    session = PromptSession(history=FileHistory(OpenAIConfig['HISTORY_FILE']))
+    api = OpenAIAPI(base_prompt, model=args.model, max_new_tokens=args.max_new_tokens, temperature=args.temperature)
+
+    command = args.command
+
+    try:
+        while True:
+            if command is not None:
+                console.print(">>> ", command, style="bold cyan")
+            else:
+                console.print("User command (Q/q[uit] to quit)...", style="bold cyan")
+                command = session.prompt(">>> ")
+                if command.lower() == 'q' or command.lower() == 'quit':
+                    console.print("User terminated chat", style="bold red")
+                    break
+
+            with console.status("[bold yellow]Waiting for ChatGPT's answer..."):
+                response = api.create_chat_completion(command)
+                command = None
+
+            sections = extract_sections_from_markdown(response, ["CODE", "FILENAME", "DESCRIPTION"])
+            console.print("Code Section:", Markdown(sections["CODE"]))
+            console.print("Filename Section:", sections["FILENAME"])
+            console.print("Description Section:", sections["DESCRIPTION"])
+
+            _fname = sections["FILENAME"]
+            fname = _fname if _fname and _fname.endswith(extension) else None
+
+            if fname is not None:
+                console.print(f"Save to {sections['FILENAME']}? [Y/n]", style="bold cyan")
+                save = prompt("> ")
+                if save.lower() == 'y':
+                    try:
+                        code = extract_code_block(sections["CODE"])
+                        with open(_fname, 'w') as f:
+                            f.write(f"{shebang}{code}")
+                            console.print(f"Saved to {_fname}", style="bold green")
+                            break
+                    except OSError as e:
+                        console.print(f"Error saving to {_fname}: {e}", style="bold orange")
+                else:
+                    console.print("Let me know how I can edit my previous answer to better reflect your intent.", style="bold green")
+
+    except KeyboardInterrupt:
+        console.print("User terminated chat", style="bold red")
+
+    finally:
+        if api.usage['total_tokens'] > 0:
+            api.log_session()
+
+    return None
+
+if __name__ == '__main__':
+    main()