asky-cli 0.1.6__py3-none-any.whl

asky/config.py

@@ -0,0 +1,270 @@
+"""Configuration constants and static declarations for asky."""
+
+import os
+import tomllib
+from pathlib import Path
+import copy
+from typing import Dict, Any
+
+
+import shutil
+from importlib import resources
+
+
+def _get_config_dir() -> Path:
+    """Return the configuration directory path."""
+    return Path.home() / ".config" / "asky"
+
+
+def _hydrate_models(config: Dict[str, Any]) -> Dict[str, Any]:
+    """Hydrate model definitions with API details."""
+    api_defs = config.get("api", {})
+    models = config.get("models", {})
+
+    for alias, model_data in models.items():
+        model_data["alias"] = alias
+        api_ref = model_data.get("api")
+        if api_ref and api_ref in api_defs:
+            api_config = api_defs[api_ref]
+
+            if "url" in api_config and "base_url" not in model_data:
+                model_data["base_url"] = api_config["url"]
+
+            if "api_key" in api_config and "api_key" not in model_data:
+                model_data["api_key"] = api_config["api_key"]
+
+            if "api_key_env" in api_config and "api_key_env" not in model_data:
+                model_data["api_key_env"] = api_config["api_key_env"]
+
+    return config
+
+
+def load_config() -> Dict[str, Any]:
+    """Load configuration from TOML file, falling back to defaults."""
+    config_dir = _get_config_dir()
+    config_path = config_dir / "config.toml"
+
+    # Ensure config directory exists
+    config_dir.mkdir(parents=True, exist_ok=True)
+
+    bundled_config_path_traversable = resources.files("asky").joinpath("config.toml")
+
+    # Read default config from package
+    try:
+        with bundled_config_path_traversable.open("rb") as f:
+            default_config = tomllib.load(f)
+    except Exception as e:
+        print(f"Error loading bundled config: {e}")
+        # Build a minimal fallback if bundled config fails
+        default_config = {
+            "general": {
+                "query_summary_max_chars": 40,
+                "continue_query_threshold": 160,
+                "answer_summary_max_chars": 200,
+                "searxng_url": "http://localhost:8888",
+                "max_turns": 20,
+                "default_model": "gf",
+                "summarization_model": "lfm",
+                "request_timeout": 60,
+                "default_context_size": 4096,
+            },
+            "api": {},
+            "models": {
+                "gf": {
+                    "id": "gemini-flash-latest",
+                    "api": "gemini",
+                    "max_chars": 1000000,
+                    "context_size": 1000000,
+                }
+            },
+            "prompts": {
+                "system_prefix": "",
+                "force_search": "",
+                "system_suffix": "",
+                "deep_research": "",
+                "deep_dive": "",
+            },
+            "user_prompts": {},
+            "tool": {},
+        }
+
+    # Initialize from defaults
+    final_config = copy.deepcopy(default_config)
+
+    # Copy bundled config to user config dir if it doesn't exist
+    if not config_path.exists():
+        try:
+            with resources.as_file(bundled_config_path_traversable) as source_path:
+                shutil.copy(source_path, config_path)
+            print(f"Created default configuration at {config_path}")
+        except Exception as e:
+            print(f"Warning: Failed to create default config at {config_path}: {e}")
+
+    # Load user config if it exists and merge
+    if config_path.exists():
+        try:
+            with open(config_path, "rb") as f:
+                user_config = tomllib.load(f)
+
+            # Recursive merge with default config
+            def merge(base, update):
+                for k, v in update.items():
+                    if k in base and isinstance(base[k], dict) and isinstance(v, dict):
+                        merge(base[k], v)
+                    else:
+                        base[k] = v
+
+            merge(final_config, user_config)
+
+        except Exception as e:
+            print(f"Warning: Failed to load config from {config_path}: {e}")
+
+    return _hydrate_models(final_config)
+
+
+# --- Initialize Configuration ---
+_CONFIG = load_config()
+
+# --- Expose Constants ---
+
+# General
+_gen = _CONFIG["general"]
+QUERY_SUMMARY_MAX_CHARS = _gen["query_summary_max_chars"]
+CONTINUE_QUERY_THRESHOLD = _gen.get("continue_query_threshold", 160)
+ANSWER_SUMMARY_MAX_CHARS = _gen["answer_summary_max_chars"]
+ANSWER_SUMMARY_MAX_CHARS = _gen["answer_summary_max_chars"]
+SEARXNG_URL = _gen["searxng_url"]
+MAX_TURNS = _gen["max_turns"]
+DEFAULT_MODEL = _gen["default_model"]
+SUMMARIZATION_MODEL = _gen["summarization_model"]
+SEARCH_PROVIDER = _gen.get("search_provider", "searxng")
+SERPER_API_URL = _gen.get("serper_api_url", "https://google.serper.dev/search")
+SERPER_API_KEY_ENV = _gen.get("serper_api_key_env", "SERPER_API_KEY")
+USER_AGENT = _gen.get(
+    "user_agent",
+    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36",
+)
+LLM_USER_AGENT = _gen.get("llm_user_agent", USER_AGENT)
+REQUEST_TIMEOUT = _gen.get("request_timeout", 60)
+DEFAULT_CONTEXT_SIZE = _gen.get("default_context_size", 4096)
+
+# Database
+# DB Path logic:
+# 1. Env Var (name defined in config, e.g. SEARXNG_HISTORY_DB_PATH)
+# 2. Configured 'db_path' in [general]
+# 3. Default: ~/.config/asky/history.db
+
+_db_env_var_name = _gen.get("db_path_env_var", "SEARXNG_HISTORY_DB_PATH")
+_env_path = os.environ.get(_db_env_var_name)
+
+if _env_path:
+    DB_PATH = Path(_env_path)
+elif "db_path" in _gen and _gen["db_path"]:
+    DB_PATH = Path(_gen["db_path"]).expanduser()
+else:
+    DB_PATH = _get_config_dir() / "history.db"
+
+# Models
+MODELS = _CONFIG["models"]
+
+# Prompts
+_prompts = _CONFIG["prompts"]
+SYSTEM_PROMPT = _prompts["system_prefix"]
+FORCE_SEARCH_PROMPT = _prompts["force_search"]
+SYSTEM_PROMPT_SUFFIX = _prompts["system_suffix"]
+DEEP_RESEARCH_PROMPT_TEMPLATE = _prompts["deep_research"]
+DEEP_DIVE_PROMPT_TEMPLATE = _prompts["deep_dive"]
+SUMMARIZE_QUERY_PROMPT_TEMPLATE = _prompts.get(
+    "summarize_query",
+    "Summarize the following query into a single short sentence (max {QUERY_SUMMARY_MAX_CHARS} chars).",
+)
+SUMMARIZE_ANSWER_PROMPT_TEMPLATE = _prompts.get(
+    "summarize_answer",
+    "Summarize the following answer into a short paragraph (max {ANSWER_SUMMARY_MAX_CHARS} chars).",
+)
+USER_PROMPTS = _CONFIG.get("user_prompts", {})
+
+# --- Custom Tools ---
+# These are loaded from [tool.NAME] sections in config.toml
+CUSTOM_TOOLS = _CONFIG.get("tool", {})
+
+
+# --- Tool Definitions ---
+# Tools are code-coupled schemas, keeping them here as constants.
+TOOLS = [
+    {
+        "type": "function",
+        "function": {
+            "name": "web_search",
+            "description": "Search the web and return top results.",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "q": {"type": "string"},
+                    "count": {"type": "integer", "default": 5},
+                },
+                "required": ["q"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_url_content",
+            "description": "Fetch the content of one or more URLs and return their text content (HTML stripped).",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "urls": {
+                        "type": "array",
+                        "items": {"type": "string"},
+                        "description": "List of URLs to fetch content from.",
+                    },
+                    "url": {
+                        "type": "string",
+                        "description": "Single URL (deprecated, use 'urls' instead).",
+                    },
+                    "summarize": {
+                        "type": "boolean",
+                        "description": "If true, summarize the content of the page using an LLM.",
+                    },
+                },
+                "required": [],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_url_details",
+            "description": "Fetch content and extract links from a URL. Use this in deep dive mode.",
+            "parameters": {
+                "type": "object",
+                "properties": {"url": {"type": "string"}},
+                "required": ["url"],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_date_time",
+            "description": "Return the current date and time.",
+            "parameters": {"type": "object", "properties": {}},
+        },
+    },
+]
+
+# Append custom tools from config.toml
+for tool_name, tool_data in CUSTOM_TOOLS.items():
+    tool_entry = {
+        "type": "function",
+        "function": {
+            "name": tool_name,
+            "description": tool_data.get("description", f"Custom tool: {tool_name}"),
+            "parameters": tool_data.get(
+                "parameters", {"type": "object", "properties": {}}
+            ),
+        },
+    }
+    TOOLS.append(tool_entry)

asky/config.toml

@@ -0,0 +1,226 @@
+# asky Configuration File
+# This file defines general settings, API endpoints, user shortcuts, and model configurations.
+
+# --- General Settings ---
+[general]
+# Name of the environment variable that stores the path to the SQLite history database.
+# Default if not set: SEARXNG_HISTORY_DB_PATH
+db_path_env_var = "SEARXNG_HISTORY_DB_PATH"
+
+# Maximum length of the query and answer summaries shown in 'asky -H'.
+query_summary_max_chars = 40
+answer_summary_max_chars = 200
+
+# Threshold for using full query vs summary in --continue-chat mode.
+# If query length is below this, it's used as is.
+continue_query_threshold = 160
+
+# URL of your SearXNG instance.
+searxng_url = "http://localhost:8888"
+
+# Search provider to be used: "searxng" or "serper"
+search_provider = "searxng"
+
+# URL of the Serper API.
+serper_api_url = "https://google.serper.dev/search"
+
+# Name of the environment variable that stores the Serper API key.
+serper_api_key_env = "SERPER_API_KEY"
+
+# Maximum number of turns (tool calls) allowed in a single conversation loop.
+max_turns = 20
+
+# Timeout for API requests in seconds
+request_timeout = 60
+
+# Default context size (if not specified in the model configuration)
+default_context_size = 4096
+
+# Default model used when no model is specified via CLI (-m).
+default_model = "gf"
+
+# Model used specifically for internal text summarization tasks.
+summarization_model = "lfm"
+
+# User-Agent string to be used for search and content retrieval requests.
+user_agent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36"
+
+# Specific User-Agent string for LLM API requests.
+# Some providers only accept specific user agents for certain subscriptions types.
+llm_user_agent = "asky/1.0.0"
+
+# --- API Definitions ---
+# Define reusable API endpoints and authentication details here.
+# For each section [api.NAME]:
+#   url: The base URL for the chat completions endpoint.
+#   api_key_env: (Recommended) Name of the environment variable containing the API key.
+#   api_key: (Optional) The API key directly. Takes precedence over api_key_env if both are missing.
+
+[api.gemini]
+url = "https://generativelanguage.googleapis.com/v1beta/chat/completions"
+api_key_env = "GOOGLE_API_KEY"
+
+
+[api.anthropic]
+url = "https://api.anthropic.com/v1/messages"
+api_key_env = "ANTHROPIC_API_KEY"
+
+
+[api.openai]
+url = "https://api.openai.com/v1/chat/completions"
+api_key_env = "OPENAI_API_KEY"
+
+
+[api.openrouter]
+url = "https://openrouter.ai/api/v1/chat/completions"
+api_key_env = "OPENROUTER_API_KEY"
+
+
+[api.lmstudio]
+url = "http://localhost:1234/v1/chat/completions"
+api_key = "lm-studio"
+
+
+[api.zai]
+url = "https://api.z.ai/api/paas/v4/chat/completions"
+api_key_env = "ZAI_API_KEY"
+
+
+# --- User Prompts (Shortcuts) ---
+# Define shortcuts that can be used via '/key' in the CLI.
+# Example: 'ask /gn' will be expanded to the prompt below.
+[user_prompts]
+gn = "Give me latest news from The Guardian, use https://www.theguardian.com/europe"
+wh = "how is weather in "
+ex = "Explain this: /cp"
+
+# --- User defined tools ---
+# These tools allow you to expose CLI commands to the LLM.
+# Use {parameter_name} in the command to inject arguments, or they will be appended.
+[tool.list_dir]
+command = "ls {flags} {path}"
+description = "List the contents of a directory."
+parameter_type = "object"
+[tool.list_dir.parameters]
+type = "object"
+required = ["path"]
+
+[tool.list_dir.parameters.properties.path]
+type = "string"
+default = "."
+
+[tool.list_dir.parameters.properties.flags]
+type = "string"
+default = "-la"
+
+
+# [tool.grep_search]
+# command = "grep -r --exclude-dir={.venv,node_modules} {pattern} {path}"
+# description = "Search for a pattern in files recursively."
+
+# [tool.grep_search.parameters]
+# type = "object"
+# required = ["pattern"]
+
+# [tool.grep_search.parameters.properties.pattern]
+# type = "string"
+# description = "The regex pattern to search for."
+
+# [tool.grep_search.parameters.properties.path]
+# type = "string"
+# description = "The directory path to search in."
+# default = "."
+
+
+
+# --- Internal Prompt Templates ---
+# Templates used to construct system prompts for different modes.
+# Placeholders like {MAX_TURNS} and {n} are filled at runtime.
+[prompts]
+# Global system prompt prefix.
+system_prefix = """You are a helpful assistant with web searc and URL retrieval capabilities. Use get_date_time for current date/time if needed (e.g., for 'today' or 'recently'). """
+
+# Prompt appended when force_search is enabled (e.g., --force-search).
+force_search = """Unless you are asked to use a specific URL, always use web_search, never try to answer without using web_search. """
+
+# Global system prompt suffix.
+system_suffix = """Then use get_url_content for details of the search results. You can pass a list of URLs to get_url_content to fetch multiple pages efficiently at once. Use tools, don't say you can't.You have {MAX_TURNS} turns to complete your task, if you reach the limit, process will be terminated.You should finish your task before reaching %100 of your token limit."""
+
+# Prompt for summarizing the user query.
+summarize_query = "Summarize the following query into a single short sentence (max {QUERY_SUMMARY_MAX_CHARS} chars)."
+
+# Prompt for summarizing the final answer.
+summarize_answer = """Summarize the following answer into a short paragraph (max {ANSWER_SUMMARY_MAX_CHARS} chars).
+ Be sure to include all numerical values and dates """
+
+# Template for DEEP RESEARCH mode (-d). {n} is the number of searches.
+deep_research = """
+You are in DEEP RESEARCH mode. You MUST perform at least {n} distinct web searches, or make {n} get_url_content calls to gather comprehensive information before providing a final answer.If you need to get links from a URL, use get_url_details. If you just need to get content from a URL, use get_url_content."""
+
+# Template for DEEP DIVE mode (-dd).
+deep_dive = """
+You are in DEEP DIVE mode. Follow these instructions:
+1. Use 'get_url_details' for the INITIAL page to retrieve content and links.
+2. Follow up to 25 relevant links within the same domain to gather comprehensive information.
+3. IMPORTANT: Use 'get_url_details' ONLY for the first page. Use 'get_url_content' for all subsequent links.
+4. Do not rely on your internal knowledge; base your answer strictly on the retrieved content.5. Do not use web_search in deep dive mode."""
+
+# --- Model Definitions ---
+# Each section [models.NAME] configures a specific model:
+#   id: The exact model ID used by the API provider.
+#   api: Reference to a name defined in the [api] section.
+#   max_chars: Character limit for content fetched via get_url_content. More than this will be truncated.
+#   context_size: Total context window size (tokens/chars approximation) for trimming history.
+
+# Note: max_chars is the context_size values of the following models are arbitrarily set for my own use.
+# Check model provider's documentation for the actual context size of the models.
+# Experiment with max_chars to find the optimal value for your use case (depending on your needs, and model/hardware capabilities)
+
+[models.gf]
+id = "gemini-flash-latest"
+api = "gemini"
+max_chars = 10000
+context_size = 1000000
+
+[models.glmair]
+id = "glm-4.5-air"
+api = "zai"
+max_chars = 4000
+context_size = 100000
+
+[models.glmflash]
+id = "glm-4.7-flash"
+api = "zai"
+max_chars = 4000
+context_size = 100000
+
+[models.q34t]
+id = "qwen/qwen3-4b-thinking-2507"
+api = "lmstudio"
+max_chars = 4000
+context_size = 32000
+
+[models.q34]
+id = "qwen/qwen3-4b-2507"
+api = "lmstudio"
+max_chars = 4000
+context_size = 32000
+
+[models.lfm]
+id = "liquid/lfm2.5-1.2b"
+api = "lmstudio"
+max_chars = 100000
+context_size = 32000
+
+[models.q8]
+id = "qwen/qwen3-8b"
+api = "lmstudio"
+max_chars = 4000
+context_size = 32000
+
+[models.q30]
+id = "qwen/qwen3-30b-a3b-2507"
+api = "lmstudio"
+max_chars = 3000
+context_size = 32000
+

asky/html.py

@@ -0,0 +1,62 @@
+"""HTML parsing utilities."""
+
+import re
+from html.parser import HTMLParser
+from typing import Any, Dict, List, Optional
+
+
+class HTMLStripper(HTMLParser):
+    """Parse HTML and extract text content and links."""
+
+    def __init__(self) -> None:
+        super().__init__()
+        self.reset()
+        self.strict = False
+        self.convert_charrefs = True
+        self.text: List[str] = []
+        self.links: List[Dict[str, str]] = []
+        self.ignore = False
+        self.current_href: Optional[str] = None
+
+    def handle_starttag(self, tag: str, attrs: List[Any]) -> None:
+        if tag in ("script", "style"):
+            self.ignore = True
+        elif tag == "a":
+            for k, v in attrs:
+                if k == "href":
+                    self.current_href = v
+                    break
+
+    def handle_endtag(self, tag: str) -> None:
+        if tag in ("script", "style"):
+            self.ignore = False
+        elif tag == "a":
+            self.current_href = None
+
+    def handle_data(self, data: str) -> None:
+        if not self.ignore:
+            text = data.strip()
+            if text:
+                self.text.append(data)
+                if self.current_href:
+                    self.links.append({"text": text, "href": self.current_href})
+
+    def get_data(self) -> str:
+        return "".join(self.text).strip()
+
+    def get_links(self) -> List[Dict[str, str]]:
+        return self.links
+
+
+def strip_tags(html: str) -> str:
+    """Strip HTML tags from text and return plain text content."""
+    s = HTMLStripper()
+    s.feed(html)
+    return s.get_data()
+
+
+def strip_think_tags(text: str) -> str:
+    """Remove <think>...</think> blocks from LLM output."""
+    if not text:
+        return ""
+    return re.sub(r"<think>.*?</think>", "", text, flags=re.DOTALL).strip()