bhai-cli 0.1.0__py3-none-any.whl

bhai_cli/__init__.py

@@ -0,0 +1,9 @@
+"""
+BHAI-CLI: The Dual-Brain AI Coding Agent with Punjabi Swagger.
+
+Lean heritage. Zero-state resilience. Universal LLM support via litellm.
+Indic-native voice interface via Sarvam AI (Saaras v3 STT / Bulbul v3 TTS).
+"""
+
+__version__ = "0.1.0"
+__app_name__ = "bhai-cli"

bhai_cli/audio_engine.py

@@ -0,0 +1,269 @@
+"""
+Audio Engine for BHAI-CLI.
+
+Handles:
+  - Live microphone recording via sounddevice (primary)
+  - File-based audio fallback (configurable)
+  - STT via Sarvam Saaras v3 (translate mode for code-mixed Punjabi/Hindi → English)
+  - TTS via Sarvam Bulbul v3 (English → Indian-accented speech)
+
+Edge Case: If Sarvam fails, gracefully returns error flags so the orchestrator
+can downgrade to text-only mode without crashing.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import base64
+import io
+import tempfile
+import wave
+from pathlib import Path
+from typing import Optional
+
+import httpx
+from rich.console import Console
+
+console = Console()
+
+# ---------------------------------------------------------------------------
+# Sarvam API Constants
+# ---------------------------------------------------------------------------
+SARVAM_BASE = "https://api.sarvam.ai"
+STT_ENDPOINT = f"{SARVAM_BASE}/speech-to-text"
+TTS_ENDPOINT = f"{SARVAM_BASE}/text-to-speech"
+
+SAMPLE_RATE = 16000  # Optimal for Saaras v3
+CHANNELS = 1
+DTYPE = "int16"
+
+
+# ---------------------------------------------------------------------------
+# Audio Recording
+# ---------------------------------------------------------------------------
+
+class AudioRecorder:
+    """Thread-safe live microphone recorder using sounddevice."""
+
+    def __init__(self, sample_rate: int = SAMPLE_RATE, channels: int = CHANNELS):
+        self.sample_rate = sample_rate
+        self.channels = channels
+        self._frames: list = []
+        self._recording = False
+        self._stream = None
+
+    def _check_sounddevice(self):
+        """Check if sounddevice is available (optional dependency)."""
+        try:
+            import sounddevice  # noqa: F401
+            return True
+        except ImportError:
+            console.print(
+                "[bold red]⚠ Audio support not installed.[/] "
+                "Run: [cyan]pip install bhai-cli\\[audio][/]"
+            )
+            return False
+
+    def start(self) -> bool:
+        """Start recording from microphone. Returns False if unavailable."""
+        if not self._check_sounddevice():
+            return False
+
+        import sounddevice as sd
+        import numpy as np
+
+        self._frames = []
+        self._recording = True
+
+        def callback(indata, frames, time_info, status):
+            if self._recording:
+                self._frames.append(indata.copy())
+
+        try:
+            self._stream = sd.InputStream(
+                samplerate=self.sample_rate,
+                channels=self.channels,
+                dtype=DTYPE,
+                callback=callback,
+            )
+            self._stream.start()
+            return True
+        except Exception as e:
+            console.print(f"[bold red]⚠ Microphone error:[/] {e}")
+            self._recording = False
+            return False
+
+    def stop(self) -> Optional[bytes]:
+        """Stop recording and return WAV bytes, or None on failure."""
+        self._recording = False
+        if self._stream:
+            self._stream.stop()
+            self._stream.close()
+            self._stream = None
+
+        if not self._frames:
+            return None
+
+        import numpy as np
+
+        audio_data = np.concatenate(self._frames, axis=0)
+
+        # Convert to WAV bytes
+        buf = io.BytesIO()
+        with wave.open(buf, "wb") as wf:
+            wf.setnchannels(self.channels)
+            wf.setsampwidth(2)  # 16-bit = 2 bytes
+            wf.setframerate(self.sample_rate)
+            wf.writeframes(audio_data.tobytes())
+
+        return buf.getvalue()
+
+
+def load_audio_file(filepath: str) -> Optional[bytes]:
+    """Load audio from a file path as bytes. Returns None on failure."""
+    path = Path(filepath)
+    if not path.exists():
+        console.print(f"[bold red]⚠ Audio file not found:[/] {filepath}")
+        return None
+    try:
+        return path.read_bytes()
+    except Exception as e:
+        console.print(f"[bold red]⚠ Failed to read audio file:[/] {e}")
+        return None
+
+
+# ---------------------------------------------------------------------------
+# Sarvam STT (Saaras v3)
+# ---------------------------------------------------------------------------
+
+async def speech_to_text(
+    audio_bytes: bytes,
+    api_key: str,
+    model: str = "saaras:v3",
+    mode: str = "translate",
+) -> tuple[bool, str]:
+    """
+    Transcribe/translate audio using Sarvam Saaras v3.
+
+    Args:
+        audio_bytes: WAV audio data.
+        api_key: Sarvam API subscription key.
+        model: STT model identifier.
+        mode: transcribe | translate | verbatim | translit | codemix
+
+    Returns:
+        (success: bool, text_or_error: str)
+    """
+    if not api_key:
+        return False, "[STT_ERROR] No Sarvam API key configured. Run: bhai-code setup"
+
+    try:
+        async with httpx.AsyncClient(timeout=30.0) as client:
+            response = await client.post(
+                STT_ENDPOINT,
+                headers={"api-subscription-key": api_key},
+                files={"file": ("audio.wav", audio_bytes, "audio/wav")},
+                data={"model": model, "mode": mode},
+            )
+
+        if response.status_code == 200:
+            data = response.json()
+            transcript = data.get("transcript", "")
+            if transcript:
+                return True, transcript
+            return False, "[STT_ERROR] Empty transcript returned."
+        else:
+            return False, f"[STT_ERROR] Sarvam API returned {response.status_code}: {response.text}"
+
+    except httpx.TimeoutException:
+        return False, "[STT_ERROR] Sarvam STT request timed out."
+    except Exception as e:
+        return False, f"[STT_ERROR] {type(e).__name__}: {e}"
+
+
+# ---------------------------------------------------------------------------
+# Sarvam TTS (Bulbul v3)
+# ---------------------------------------------------------------------------
+
+async def text_to_speech(
+    text: str,
+    api_key: str,
+    model: str = "bulbul:v3",
+    speaker: str = "shubh",
+    language: str = "en-IN",
+) -> tuple[bool, Optional[bytes]]:
+    """
+    Convert text to speech using Sarvam Bulbul v3.
+
+    Args:
+        text: Text to speak (max 2500 chars).
+        api_key: Sarvam API subscription key.
+        model: TTS model identifier.
+        speaker: Voice speaker name.
+        language: Target language code.
+
+    Returns:
+        (success: bool, audio_bytes_or_none)
+    """
+    if not api_key:
+        return False, None
+
+    # Truncate to API limit
+    text = text[:2500]
+
+    try:
+        payload = {
+            "text": text,
+            "target_language_code": language,
+            "speaker": speaker,
+            "model": model,
+            "speech_sample_rate": 24000,
+            "pace": 1.0,
+            "temperature": 0.6,
+        }
+
+        async with httpx.AsyncClient(timeout=30.0) as client:
+            response = await client.post(
+                TTS_ENDPOINT,
+                headers={
+                    "api-subscription-key": api_key,
+                    "Content-Type": "application/json",
+                },
+                json=payload,
+            )
+
+        if response.status_code == 200:
+            data = response.json()
+            audios = data.get("audios", [])
+            if audios:
+                audio_bytes = base64.b64decode(audios[0])
+                return True, audio_bytes
+            return False, None
+        else:
+            console.print(
+                f"[dim]TTS warning: Sarvam returned {response.status_code}[/]"
+            )
+            return False, None
+
+    except Exception as e:
+        console.print(f"[dim]TTS fallback: {type(e).__name__}: {e}[/]")
+        return False, None
+
+
+async def play_audio(audio_bytes: bytes) -> None:
+    """Play audio bytes through the default output device."""
+    try:
+        import sounddevice as sd
+        import soundfile as sf
+
+        buf = io.BytesIO(audio_bytes)
+        data, samplerate = sf.read(buf)
+        sd.play(data, samplerate)
+        sd.wait()
+    except ImportError:
+        # Save to temp file and play with system player as fallback
+        with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as f:
+            f.write(audio_bytes)
+            console.print(f"[dim]Audio saved to: {f.name}[/]")
+    except Exception as e:
+        console.print(f"[dim]Audio playback failed: {e}[/]")

bhai_cli/cli.py

@@ -0,0 +1,305 @@
+"""
+BHAI-CLI — Typer CLI Application.
+
+Commands:
+  bhai-code setup   — Interactive configuration wizard
+  bhai-code text    — Send a text prompt to BHAI
+  bhai-code listen  — Voice input via Sarvam STT → BHAI pipeline
+"""
+
+from __future__ import annotations
+
+import asyncio
+from typing import Optional
+
+import typer
+import pyfiglet
+from rich.console import Console
+from rich.panel import Panel
+from rich.prompt import Prompt
+from rich.table import Table
+from rich.text import Text
+
+from bhai_cli import __version__
+from bhai_cli.config_manager import (
+    load_config,
+    save_config,
+    CONFIG_FILE,
+)
+
+console = Console()
+
+# ---------------------------------------------------------------------------
+# CLI Setup
+# ---------------------------------------------------------------------------
+
+# "BHAI" in 6 Indian scripts
+BHAI_LANGS = [
+    ("भाई", "Hindi"),
+    ("ভাই", "Bengali"),
+    ("ભાઈ", "Gujarati"),
+    ("ਭਾਈ", "Punjabi"),
+    ("భాయ్", "Telugu"),
+    ("ಭಾಯ್", "Kannada"),
+]
+
+
+def print_banner() -> None:
+    """Print the BHAI-CLI ASCII banner with multilingual flair."""
+    ascii_banner = pyfiglet.figlet_format("BHAI - CLI", font="slant")
+    banner_text = Text(ascii_banner, style="bold bright_cyan")
+    console.print(banner_text)
+
+    # Show "BHAI" in multiple Indian scripts
+    lang_line = " · ".join(f"[bold]{word}[/][dim]({lang})[/]" for word, lang in BHAI_LANGS)
+    console.print(f"  {lang_line}\n")
+
+    console.print(
+        "[bold yellow]🇮🇳 The Dual-Brain AI Coding Agent — 22 Indian Languages.[/]\n"
+        "[dim]Powered by litellm & Sarvam AI | v" + __version__ + "[/]\n"
+    )
+
+app = typer.Typer(
+    name="bhai-code",
+    help="BHAI — Your 10x Senior SDE from India. 22 Languages. One CLI.",
+    add_completion=False,
+    rich_markup_mode="rich",
+    no_args_is_help=True,
+)
+
+def _version_callback(value: bool) -> None:
+    if value:
+        print_banner()
+        raise typer.Exit()
+
+@app.callback(invoke_without_command=True)
+def main(
+    ctx: typer.Context,
+    version: bool = typer.Option(
+        False, "--version", "-v",
+        help="Show version and exit.",
+        callback=_version_callback,
+        is_eager=True,
+    ),
+) -> None:
+    """[bold cyan]BHAI-CLI[/] — 10x Senior SDE Agent from India. 22 Languages."""
+    # Only print banner if no subcommand was invoked (which means help will be shown automatically via no_args_is_help)
+    if ctx.invoked_subcommand is None:
+        print_banner()
+
+# ---------------------------------------------------------------------------
+# bhai-code setup
+# ---------------------------------------------------------------------------
+
+@app.command()
+def setup() -> None:
+    """[bold yellow]⚙️  setup[/] — Interactive wizard for VibeLLM, CoderAgent, and Sarvam AI."""
+    print_banner()
+    
+    console.print(
+        Panel(
+            "[bold white]Let's get your dual-brain engine running, veere.[/]\n"
+            "[dim]We need to configure your Persona (Brain A) and your Execution Engine (Brain B).[/]",
+            title="[bold yellow]Setup Wizard[/]",
+            border_style="yellow",
+            padding=(1, 2),
+        )
+    )
+
+    cfg = load_config()
+
+    console.print("\n[bold cyan]━━━ Brain A: VibeLLM (Persona & Intent) ━━━[/]")
+    console.print(
+        "[dim]Handles the BHAI persona and parses your intent.\n"
+        "Format: provider/model-name (e.g., sarvam/sarvam-105b, groq/llama-3.3-70b-versatile)[/]\n"
+    )
+    cfg.vibe.model = Prompt.ask("VibeLLM model", default=cfg.vibe.model)
+    cfg.vibe.api_key = Prompt.ask("VibeLLM API key", default=cfg.vibe.api_key or "", password=True)
+    cfg.vibe.api_base = Prompt.ask("VibeLLM API base URL [dim](optional)[/]", default=cfg.vibe.api_base or "")
+
+    console.print("\n[bold cyan]━━━ Brain B: CoderAgent (Tool Execution) ━━━[/]")
+    console.print(
+        "[dim]Executes coding tasks using MCP-style tools. Must support tool-calling.\n"
+        "Recommended: groq/llama-3.3-70b-versatile, openai/gpt-4o, anthropic/claude-sonnet-4-20250514[/]\n"
+    )
+    cfg.coder.model = Prompt.ask("CoderAgent model", default=cfg.coder.model)
+    cfg.coder.api_key = Prompt.ask("CoderAgent API key", default=cfg.coder.api_key or "", password=True)
+    cfg.coder.api_base = Prompt.ask("CoderAgent API base URL [dim](optional)[/]", default=cfg.coder.api_base or "")
+
+    console.print("\n[bold cyan]━━━ Sarvam AI (Voice Interface) ━━━[/]")
+    console.print("[dim]Optional: Enables voice input (STT) and output (TTS).[/]\n")
+    cfg.sarvam.api_key = Prompt.ask("Sarvam API key [dim](leave empty to skip)[/]", default=cfg.sarvam.api_key or "", password=True)
+
+    path = save_config(cfg)
+
+    console.print()
+    console.print(
+        Panel(
+            f"[bold green]✅ All set, boss! Config saved to:[/] {path}\n\n"
+            f"🧠 [bold]VibeLLM:[/]    {cfg.vibe.model}\n"
+            f"🛠️  [bold]CoderAgent:[/] {cfg.coder.model}\n"
+            f"🎙️  [bold]Sarvam:[/]     {'Active ✓' if cfg.sarvam.api_key else 'Skipped (Text-only)'}",
+            title="[bold green]Setup Complete[/]",
+            border_style="green",
+            padding=(1, 2),
+        )
+    )
+
+# ---------------------------------------------------------------------------
+# bhai-code text
+# ---------------------------------------------------------------------------
+
+@app.command()
+def text(
+    prompt: str = typer.Argument(..., help="Your message to BHAI."),
+    vibe_model: Optional[str] = typer.Option(None, "--vibe-model", "-vm", help="Override VibeLLM model."),
+    coder_model: Optional[str] = typer.Option(None, "--coder-model", "-cm", help="Override CoderAgent model."),
+    interactive: bool = typer.Option(False, "--interactive", "-i", help="Enter interactive chat mode."),
+) -> None:
+    """[bold magenta]💬 text[/] — Send a text prompt to BHAI."""
+    print_banner()
+    from bhai_cli.orchestrator import BhaiOrchestrator
+
+    cfg = load_config(vibe_model_override=vibe_model, coder_model_override=coder_model)
+    orchestrator = BhaiOrchestrator(cfg)
+
+    asyncio.run(orchestrator.process(prompt))
+
+    if interactive:
+        console.print("\n[bold cyan]Interactive mode — type 'exit' or 'quit' to leave.[/]\n")
+        while True:
+            try:
+                user_input = Prompt.ask("[bold green]You[/]")
+                if user_input.strip().lower() in ("exit", "quit", "bye", "chal hatja", "nikal"):
+                    console.print("\n[bold yellow]BHAI:[/] Accha bhai, phir milenge! 🤙")
+                    break
+                asyncio.run(orchestrator.process(user_input))
+            except (KeyboardInterrupt, EOFError):
+                console.print("\n[bold yellow]BHAI:[/] Theek hai boss, baad mein baat karte hain! 👋")
+                break
+
+# ---------------------------------------------------------------------------
+# bhai-code listen
+# ---------------------------------------------------------------------------
+
+@app.command()
+def listen(
+    vibe_model: Optional[str] = typer.Option(None, "--vibe-model", "-vm", help="Override VibeLLM model."),
+    coder_model: Optional[str] = typer.Option(None, "--coder-model", "-cm", help="Override CoderAgent model."),
+    file: Optional[str] = typer.Option(None, "--file", "-f", help="Path to an audio file instead of live mic."),
+) -> None:
+    """[bold red]🎙️  listen[/] — Voice input — speak to BHAI in any Indian language."""
+    print_banner()
+    from bhai_cli.orchestrator import BhaiOrchestrator
+    from bhai_cli.audio_engine import (
+        AudioRecorder, load_audio_file, speech_to_text, text_to_speech, play_audio,
+    )
+
+    cfg = load_config(vibe_model_override=vibe_model, coder_model_override=coder_model)
+
+    if not cfg.sarvam.api_key:
+        console.print(
+            Panel(
+                "[bold red]⚠ Sarvam API key not configured.[/]\n"
+                "Run [cyan]bhai-code setup[/] to add your Sarvam key,\n"
+                "or use [cyan]bhai-code text[/] for text-only mode.",
+                title="[bold red]Error[/]", border_style="red"
+            )
+        )
+        raise typer.Exit(1)
+
+    orchestrator = BhaiOrchestrator(cfg)
+    audio_bytes = None
+    audio_file = file or cfg.audio_fallback_file
+
+    if audio_file:
+        console.print(f"[dim]Loading audio from: {audio_file}[/]")
+        audio_bytes = load_audio_file(audio_file)
+        if audio_bytes is None:
+            raise typer.Exit(1)
+    else:
+        recorder = AudioRecorder()
+        console.print(
+            Panel(
+                "[bold white]Press ENTER to start recording, then ENTER again to stop.[/]",
+                title="[bold magenta]🎙️  Live Recording[/]",
+                border_style="magenta", padding=(1, 2),
+            )
+        )
+        input() 
+
+        if not recorder.start():
+            console.print("[bold red]Failed to start recording. Falling back to text mode.[/]")
+            raise typer.Exit(1)
+
+        console.print("[bold red blink]🔴 Recording... Press ENTER to stop.[/]")
+        input()
+
+        audio_bytes = recorder.stop()
+        if audio_bytes is None:
+            console.print("[bold red]No audio captured.[/]")
+            raise typer.Exit(1)
+        console.print("[bold green]✅ Audio captured successfully.[/]")
+
+    with console.status("[bold cyan]Transcribing with Sarvam Saaras v3... 🎧[/]"):
+        success, transcript = asyncio.run(
+            speech_to_text(audio_bytes, api_key=cfg.sarvam.api_key, model=cfg.sarvam.stt_model, mode=cfg.sarvam.stt_mode)
+        )
+
+    if not success:
+        console.print(f"[bold red]STT failed:[/] {transcript}")
+        console.print("[dim]Falling back to text mode. Use bhai-code text instead.[/]")
+        raise typer.Exit(1)
+
+    console.print(f"[bold green]📝 You said:[/] {transcript}\n")
+    result = asyncio.run(orchestrator.process(transcript))
+
+    async def _speak(text: str) -> None:
+        ok, audio = await text_to_speech(
+            text, api_key=cfg.sarvam.api_key, model=cfg.sarvam.tts_model,
+            speaker=cfg.sarvam.tts_speaker, language=cfg.sarvam.tts_language,
+        )
+        if ok and audio:
+            with console.status("[bold magenta]Speaking... 🔊[/]"):
+                await play_audio(audio)
+
+    try:
+        asyncio.run(_speak(result[:2500]))
+    except Exception:
+        pass
+
+# ---------------------------------------------------------------------------
+# bhai-code config
+# ---------------------------------------------------------------------------
+
+@app.command()
+def config() -> None:
+    """[bold cyan]📋 config[/] — Show current configuration."""
+    print_banner()
+    cfg = load_config()
+
+    table = Table(title="BHAI Configuration Engine", show_header=True, header_style="bold cyan", border_style="dim", expand=True)
+    table.add_column("Key", style="bold white")
+    table.add_column("Value", style="cyan")
+
+    table.add_row("Config File", str(CONFIG_FILE))
+    table.add_section()
+    table.add_row("[yellow]VibeLLM Model[/]", cfg.vibe.model)
+    table.add_row("VibeLLM API Key", "••••" + cfg.vibe.api_key[-4:] if len(cfg.vibe.api_key) > 4 else "[red](not set)[/]")
+    table.add_row("VibeLLM API Base", cfg.vibe.api_base or "[dim](default)[/]")
+    table.add_section()
+    table.add_row("[yellow]CoderAgent Model[/]", cfg.coder.model)
+    table.add_row("CoderAgent API Key", "••••" + cfg.coder.api_key[-4:] if len(cfg.coder.api_key) > 4 else "[red](not set)[/]")
+    table.add_row("CoderAgent API Base", cfg.coder.api_base or "[dim](default)[/]")
+    table.add_section()
+    table.add_row("[yellow]Sarvam API Key[/]", "••••" + cfg.sarvam.api_key[-4:] if len(cfg.sarvam.api_key) > 4 else "[red](not set)[/]")
+    table.add_row("STT Model", cfg.sarvam.stt_model)
+    table.add_row("TTS Model", cfg.sarvam.tts_model)
+    table.add_section()
+    table.add_row("Command Timeout", f"{cfg.command_timeout}s")
+
+    console.print(table)
+    console.print()
+
+if __name__ == "__main__":
+    app()