gac 1.0.1__py3-none-any.whl → 1.1.0__py3-none-any.whl

gac/__init__.py

@@ -2,15 +2,14 @@
 
 from gac.__version__ import __version__
 from gac.ai import generate_commit_message
-from gac.ai_providers import (
-    anthropic_generate,
-    cerebras_generate,
-    groq_generate,
-    ollama_generate,
-    openai_generate,
-)
 from gac.git import get_staged_files, push_changes
 from gac.prompt import build_prompt, clean_commit_message
+from gac.providers.anthropic import generate as anthropic_generate
+from gac.providers.cerebras import generate as cerebras_generate
+from gac.providers.groq import generate as groq_generate
+from gac.providers.ollama import generate as ollama_generate
+from gac.providers.openai import generate as openai_generate
+from gac.providers.openrouter import generate as openrouter_generate
 
 __all__ = [
     "__version__",
@@ -24,4 +23,5 @@ __all__ = [
     "groq_generate",
     "ollama_generate",
     "openai_generate",
+    "openrouter_generate",
 ]

gac/__version__.py

@@ -1,3 +1,3 @@
 """Version information for gac package."""
 
-__version__ = "1.0.1"
+__version__ = "1.1.0"

gac/ai.py

@@ -5,121 +5,19 @@ It consolidates all AI-related functionality including token counting and commit
 """
 
 import logging
-import os
-from functools import lru_cache
-from typing import Any
 
-import httpx
-import tiktoken
-
-from gac.ai_providers import (
-    anthropic_generate,
-    cerebras_generate,
-    groq_generate,
-    ollama_generate,
-    openai_generate,
-)
-from gac.constants import EnvDefaults, Utility
+from gac.constants import EnvDefaults
 from gac.errors import AIError
+from gac.providers.anthropic import generate as anthropic_generate
+from gac.providers.cerebras import generate as cerebras_generate
+from gac.providers.groq import generate as groq_generate
+from gac.providers.ollama import generate as ollama_generate
+from gac.providers.openai import generate as openai_generate
+from gac.providers.openrouter import generate as openrouter_generate
 
 logger = logging.getLogger(__name__)
 
 
-def count_tokens(content: str | list[dict[str, str]] | dict[str, Any], model: str) -> int:
-    """Count tokens in content using the model's tokenizer."""
-    text = extract_text_content(content)
-    if not text:
-        return 0
-
-    if model.startswith("anthropic"):
-        anthropic_tokens = anthropic_count_tokens(text, model)
-        if anthropic_tokens is not None:
-            return anthropic_tokens
-        return len(text) // 4
-
-    try:
-        encoding = get_encoding(model)
-        return len(encoding.encode(text))
-    except Exception as e:
-        logger.error(f"Error counting tokens: {e}")
-        return len(text) // 4
-
-
-def anthropic_count_tokens(text: str, model: str) -> int | None:
-    """Call Anthropic's token count endpoint and return the token usage.
-
-    Returns the token count when successful, otherwise ``None`` so callers can
-    fall back to a heuristic estimate.
-    """
-    api_key = os.getenv("ANTHROPIC_API_KEY")
-    if not api_key:
-        logger.debug("ANTHROPIC_API_KEY not set; using heuristic token estimation for Anthropic model")
-        return None
-
-    model_name = model.split(":", 1)[1] if ":" in model else "claude-3-5-haiku-latest"
-    headers = {
-        "Content-Type": "application/json",
-        "x-api-key": api_key,
-        "anthropic-version": "2023-06-01",
-    }
-    payload = {
-        "model": model_name,
-        "messages": [
-            {
-                "role": "user",
-                "content": [
-                    {
-                        "type": "text",
-                        "text": text,
-                    }
-                ],
-            }
-        ],
-    }
-
-    try:
-        response = httpx.post(
-            "https://api.anthropic.com/v1/messages/count_tokens",
-            headers=headers,
-            json=payload,
-            timeout=30.0,
-        )
-        response.raise_for_status()
-        data = response.json()
-
-        if "input_tokens" in data:
-            return data["input_tokens"]
-        if "usage" in data and "input_tokens" in data["usage"]:
-            return data["usage"]["input_tokens"]
-
-        logger.warning("Unexpected response format from Anthropic token count API: %s", data)
-    except Exception as exc:
-        logger.warning("Failed to retrieve Anthropic token count via HTTP: %s", exc)
-
-    return None
-
-
-def extract_text_content(content: str | list[dict[str, str]] | dict[str, Any]) -> str:
-    """Extract text content from various input formats."""
-    if isinstance(content, str):
-        return content
-    elif isinstance(content, list):
-        return "\n".join(msg["content"] for msg in content if isinstance(msg, dict) and "content" in msg)
-    elif isinstance(content, dict) and "content" in content:
-        return content["content"]
-    return ""
-
-
-@lru_cache(maxsize=1)
-def get_encoding(model: str) -> tiktoken.Encoding:
-    """Get the appropriate encoding for a given model."""
-    model_name = model.split(":")[-1] if ":" in model else model
-    try:
-        return tiktoken.encoding_for_model(model_name)
-    except KeyError:
-        return tiktoken.get_encoding(Utility.DEFAULT_ENCODING)
-
-
 def generate_commit_message(
     model: str,
     prompt: str | tuple[str, str],
@@ -176,5 +74,7 @@ def generate_commit_message(
         return cerebras_generate(model_name, prompt, temperature, max_tokens, max_retries, quiet)
     elif provider == "ollama":
         return ollama_generate(model_name, prompt, temperature, max_tokens, max_retries, quiet)
+    elif provider == "openrouter":
+        return openrouter_generate(model_name, prompt, temperature, max_tokens, max_retries, quiet)
     else:
         raise AIError.model_error(f"Unsupported provider: {provider}")

gac/ai_utils.py

@@ -0,0 +1,134 @@
+"""Utilities for AI provider integration for gac.
+
+This module provides utility functions that support the AI provider implementations.
+"""
+
+import logging
+import os
+from functools import lru_cache
+from typing import Any
+
+import httpx
+import tiktoken
+
+from gac.constants import Utility
+
+logger = logging.getLogger(__name__)
+
+
+def count_tokens(content: str | list[dict[str, str]] | dict[str, Any], model: str) -> int:
+    """Count tokens in content using the model's tokenizer."""
+    text = extract_text_content(content)
+    if not text:
+        return 0
+
+    if model.startswith("anthropic"):
+        anthropic_tokens = anthropic_count_tokens(text, model)
+        if anthropic_tokens is not None:
+            return anthropic_tokens
+        return len(text) // 4
+
+    try:
+        encoding = get_encoding(model)
+        return len(encoding.encode(text))
+    except Exception as e:
+        logger.error(f"Error counting tokens: {e}")
+        return len(text) // 4
+
+
+def anthropic_count_tokens(text: str, model: str) -> int | None:
+    """Call Anthropic's token count endpoint and return the token usage.
+
+    Returns the token count when successful, otherwise ``None`` so callers can
+    fall back to a heuristic estimate.
+    """
+    api_key = os.getenv("ANTHROPIC_API_KEY")
+    if not api_key:
+        logger.debug("ANTHROPIC_API_KEY not set; using heuristic token estimation for Anthropic model")
+        return None
+
+    model_name = model.split(":", 1)[1] if ":" in model else "claude-3-5-haiku-latest"
+    headers = {
+        "Content-Type": "application/json",
+        "x-api-key": api_key,
+        "anthropic-version": "2023-06-01",
+    }
+    payload = {
+        "model": model_name,
+        "messages": [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "text",
+                        "text": text,
+                    }
+                ],
+            }
+        ],
+    }
+
+    try:
+        response = httpx.post(
+            "https://api.anthropic.com/v1/messages/count_tokens",
+            headers=headers,
+            json=payload,
+            timeout=30.0,
+        )
+        response.raise_for_status()
+        data = response.json()
+
+        if "input_tokens" in data:
+            return data["input_tokens"]
+        if "usage" in data and "input_tokens" in data["usage"]:
+            return data["usage"]["input_tokens"]
+
+        logger.warning("Unexpected response format from Anthropic token count API: %s", data)
+    except Exception as exc:
+        logger.warning("Failed to retrieve Anthropic token count via HTTP: %s", exc)
+
+    return None
+
+
+def extract_text_content(content: str | list[dict[str, str]] | dict[str, Any]) -> str:
+    """Extract text content from various input formats."""
+    if isinstance(content, str):
+        return content
+    elif isinstance(content, list):
+        return "\n".join(msg["content"] for msg in content if isinstance(msg, dict) and "content" in msg)
+    elif isinstance(content, dict) and "content" in content:
+        return content["content"]
+    return ""
+
+
+@lru_cache(maxsize=1)
+def get_encoding(model: str) -> tiktoken.Encoding:
+    """Get the appropriate encoding for a given model."""
+    model_name = model.split(":")[-1] if ":" in model else model
+    try:
+        return tiktoken.encoding_for_model(model_name)
+    except KeyError:
+        return tiktoken.get_encoding(Utility.DEFAULT_ENCODING)
+
+
+def _classify_error(error_str: str) -> str:
+    """Classify error types based on error message content."""
+    error_str = error_str.lower()
+
+    if (
+        "api key" in error_str
+        or "unauthorized" in error_str
+        or "authentication" in error_str
+        or "invalid api key" in error_str
+    ):
+        return "authentication"
+    elif "timeout" in error_str or "timed out" in error_str or "request timeout" in error_str:
+        return "timeout"
+    elif "rate limit" in error_str or "too many requests" in error_str or "rate limit exceeded" in error_str:
+        return "rate_limit"
+    elif "connect" in error_str or "network" in error_str or "network connection failed" in error_str:
+        return "connection"
+    elif "model" in error_str or "not found" in error_str or "model not found" in error_str:
+        return "model"
+    else:
+        return "unknown"

gac/init_cli.py

@@ -24,6 +24,7 @@ def init() -> None:
         ("Cerebras", "qwen-3-coder-480b"),
         ("Groq", "meta-llama/llama-4-maverick-17b-128e-instruct"),
         ("Ollama", "gemma3"),
+        ("OpenRouter", "openrouter/auto"),
         ("OpenAI", "gpt-4.1-mini"),
     ]
     provider_names = [p[0] for p in providers]

gac/main.py

@@ -10,7 +10,8 @@ import click
 from rich.console import Console
 from rich.panel import Panel
 
-from gac.ai import count_tokens, generate_commit_message
+from gac.ai import generate_commit_message
+from gac.ai_utils import count_tokens
 from gac.config import load_config
 from gac.constants import EnvDefaults, Utility
 from gac.errors import AIError, GitError, handle_error

gac/preprocess.py

@@ -10,7 +10,7 @@ import logging
 import os
 import re
 
-from gac.ai import count_tokens
+from gac.ai_utils import count_tokens
 from gac.constants import (
     CodePatternImportance,
     FilePatterns,

gac/providers/__init__.py

@@ -0,0 +1 @@
+"""AI providers submodule for gac."""

gac/providers/anthropic.py

@@ -0,0 +1,141 @@
+"""Anthropic API provider for gac."""
+
+import logging
+import os
+import time
+
+import httpx
+from halo import Halo
+
+from gac.ai_utils import _classify_error
+from gac.constants import EnvDefaults
+from gac.errors import AIError
+
+logger = logging.getLogger(__name__)
+
+
+def generate(
+    model: str,
+    prompt: str | tuple[str, str],
+    temperature: float = EnvDefaults.TEMPERATURE,
+    max_tokens: int = EnvDefaults.MAX_OUTPUT_TOKENS,
+    max_retries: int = EnvDefaults.MAX_RETRIES,
+    quiet: bool = False,
+) -> str:
+    """Generate commit message using Anthropic API with retry logic.
+
+    Args:
+        model: The model name (e.g., 'claude-3-5-haiku-latest', 'claude-3-opus-latest')
+        prompt: Either a string prompt or tuple of (system_prompt, user_prompt)
+        temperature: Controls randomness (0.0-1.0)
+        max_tokens: Maximum tokens in the response
+        max_retries: Number of retry attempts if generation fails
+        quiet: If True, suppress progress indicators
+
+    Returns:
+        A formatted commit message string
+
+    Raises:
+        AIError: If generation fails after max_retries attempts
+    """
+    api_key = os.getenv("ANTHROPIC_API_KEY")
+    if not api_key:
+        raise AIError.model_error("ANTHROPIC_API_KEY environment variable not set")
+
+    # Handle both old (string) and new (tuple) prompt formats
+    if isinstance(prompt, tuple):
+        system_prompt, user_prompt = prompt
+        messages = [{"role": "user", "content": user_prompt}]
+        payload = {
+            "model": model,
+            "messages": messages,
+            "system": system_prompt,
+            "temperature": temperature,
+            "max_tokens": max_tokens,
+        }
+    else:
+        # Backward compatibility: treat string as user prompt
+        messages = [{"role": "user", "content": prompt}]
+        payload = {
+            "model": model,
+            "messages": messages,
+            "temperature": temperature,
+            "max_tokens": max_tokens,
+        }
+
+    headers = {
+        "Content-Type": "application/json",
+        "x-api-key": api_key,
+        "anthropic-version": "2023-06-01",
+    }
+
+    return _make_request_with_retry(
+        url="https://api.anthropic.com/v1/messages",
+        headers=headers,
+        payload=payload,
+        provider_name=f"Anthropic {model}",
+        max_retries=max_retries,
+        quiet=quiet,
+        response_parser=lambda r: r["content"][0]["text"],
+    )
+
+
+def _make_request_with_retry(
+    url: str,
+    headers: dict,
+    payload: dict,
+    provider_name: str,
+    max_retries: int,
+    quiet: bool,
+    response_parser: callable,
+) -> str:
+    """Make HTTP request with retry logic and common error handling."""
+    if quiet:
+        spinner = None
+    else:
+        spinner = Halo(text=f"Generating commit message with {provider_name}...", spinner="dots")
+        spinner.start()
+
+    last_error = None
+    retry_count = 0
+
+    while retry_count < max_retries:
+        try:
+            logger.debug(f"Trying with {provider_name} (attempt {retry_count + 1}/{max_retries})")
+
+            with httpx.Client(timeout=30.0) as client:
+                response = client.post(url, headers=headers, json=payload)
+                response.raise_for_status()
+
+                response_data = response.json()
+                message = response_parser(response_data)
+
+                if spinner:
+                    spinner.succeed(f"Generated commit message with {provider_name}")
+
+                return message
+
+        except Exception as e:
+            last_error = e
+            retry_count += 1
+
+            if retry_count == max_retries:
+                logger.warning(f"Error generating commit message: {e}. Giving up.")
+                break
+
+            wait_time = 2**retry_count
+            logger.warning(f"Error generating commit message: {e}. Retrying in {wait_time}s...")
+            if spinner:
+                for i in range(wait_time, 0, -1):
+                    spinner.text = f"Retry {retry_count}/{max_retries} in {i}s..."
+                    time.sleep(1)
+            else:
+                time.sleep(wait_time)
+
+    if spinner:
+        spinner.fail(f"Failed to generate commit message with {provider_name}")
+
+    error_type = _classify_error(str(last_error))
+    raise AIError(
+        f"Failed to generate commit message after {max_retries} attempts: {last_error}", error_type=error_type
+    )

gac/providers/cerebras.py

@@ -0,0 +1,134 @@
+"""Cerebras API provider for gac."""
+
+import logging
+import os
+import time
+
+import httpx
+from halo import Halo
+
+from gac.ai_utils import _classify_error
+from gac.constants import EnvDefaults
+from gac.errors import AIError
+
+logger = logging.getLogger(__name__)
+
+
+def generate(
+    model: str,
+    prompt: str | tuple[str, str],
+    temperature: float = EnvDefaults.TEMPERATURE,
+    max_tokens: int = EnvDefaults.MAX_OUTPUT_TOKENS,
+    max_retries: int = EnvDefaults.MAX_RETRIES,
+    quiet: bool = False,
+) -> str:
+    """Generate commit message using Cerebras API with retry logic.
+
+    Args:
+        model: The model name (e.g., 'llama3.1-8b', 'llama3.1-70b')
+        prompt: Either a string prompt or tuple of (system_prompt, user_prompt)
+        temperature: Controls randomness (0.0-1.0)
+        max_tokens: Maximum tokens in the response
+        max_retries: Number of retry attempts if generation fails
+        quiet: If True, suppress progress indicators
+
+    Returns:
+        A formatted commit message string
+
+    Raises:
+        AIError: If generation fails after max_retries attempts
+    """
+    api_key = os.getenv("CEREBRAS_API_KEY")
+    if not api_key:
+        raise AIError.model_error("CEREBRAS_API_KEY environment variable not set")
+
+    # Handle both old (string) and new (tuple) prompt formats
+    if isinstance(prompt, tuple):
+        system_prompt, user_prompt = prompt
+        messages = [{"role": "system", "content": system_prompt}, {"role": "user", "content": user_prompt}]
+    else:
+        # Backward compatibility: treat string as user prompt
+        messages = [{"role": "user", "content": prompt}]
+
+    payload = {
+        "model": model,
+        "messages": messages,
+        "temperature": temperature,
+        "max_tokens": max_tokens,
+    }
+
+    headers = {
+        "Content-Type": "application/json",
+        "Authorization": f"Bearer {api_key}",
+    }
+
+    return _make_request_with_retry(
+        url="https://api.cerebras.ai/v1/chat/completions",
+        headers=headers,
+        payload=payload,
+        provider_name=f"Cerebras {model}",
+        max_retries=max_retries,
+        quiet=quiet,
+        response_parser=lambda r: r["choices"][0]["message"]["content"],
+    )
+
+
+def _make_request_with_retry(
+    url: str,
+    headers: dict,
+    payload: dict,
+    provider_name: str,
+    max_retries: int,
+    quiet: bool,
+    response_parser: callable,
+) -> str:
+    """Make HTTP request with retry logic and common error handling."""
+    if quiet:
+        spinner = None
+    else:
+        spinner = Halo(text=f"Generating commit message with {provider_name}...", spinner="dots")
+        spinner.start()
+
+    last_error = None
+    retry_count = 0
+
+    while retry_count < max_retries:
+        try:
+            logger.debug(f"Trying with {provider_name} (attempt {retry_count + 1}/{max_retries})")
+
+            with httpx.Client(timeout=30.0) as client:
+                response = client.post(url, headers=headers, json=payload)
+                response.raise_for_status()
+
+                response_data = response.json()
+                message = response_parser(response_data)
+
+                if spinner:
+                    spinner.succeed(f"Generated commit message with {provider_name}")
+
+                return message
+
+        except Exception as e:
+            last_error = e
+            retry_count += 1
+
+            if retry_count == max_retries:
+                logger.warning(f"Error generating commit message: {e}. Giving up.")
+                break
+
+            wait_time = 2**retry_count
+            logger.warning(f"Error generating commit message: {e}. Retrying in {wait_time}s...")
+            if spinner:
+                for i in range(wait_time, 0, -1):
+                    spinner.text = f"Retry {retry_count}/{max_retries} in {i}s..."
+                    time.sleep(1)
+            else:
+                time.sleep(wait_time)
+
+    if spinner:
+        spinner.fail(f"Failed to generate commit message with {provider_name}")
+
+    error_type = _classify_error(str(last_error))
+    raise AIError(
+        f"Failed to generate commit message after {max_retries} attempts: {last_error}", error_type=error_type
+    )