@baimingtao/lbs-agent 1.0.0

package/runtime/core/llm/provider.py

@@ -0,0 +1,120 @@
+"""
+Provider profile 基类。
+
+ProviderProfile 在一个地方声明推理服务的所有信息：鉴权、端点、客户端怪癖、请求时怪癖。
+
+Profile 是声明式的 —— 它们只描述……ribe behavior, not own client construction.
+
+Migrated from LBS Agent providers/base.py.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import urllib.request
+from dataclasses import dataclass, field
+from typing import Any
+from urllib.parse import urlparse
+
+logger = logging.getLogger(__name__)
+
+# Sentinel: some providers (Kimi) want temperature omitted entirely
+OMIT_TEMPERATURE = object()
+
+
+@dataclass
+class ProviderProfile:
+    """Base provider profile — subclass or instantiate with overrides."""
+
+    # -- Identity --
+    name: str
+    api_mode: str = "chat_completions"
+    aliases: tuple = ()
+
+    # -- Human-readable metadata --
+    display_name: str = ""
+    description: str = ""
+    signup_url: str = ""
+
+    # -- Auth & endpoints --
+    env_vars: tuple = ()
+    base_url: str = ""
+    models_url: str = ""
+    auth_type: str = "api_key"  # api_key|oauth_device_code|oauth_external|copilot|aws_sdk
+    supports_health_check: bool = True
+
+    # -- Model catalog --
+    fallback_models: tuple = ()
+    hostname: str = ""
+
+    # -- Client-level quirks --
+    default_headers: dict[str, str] = field(default_factory=dict)
+
+    # -- Request-level quirks --
+    fixed_temperature: Any = None
+    default_max_tokens: int | None = None
+    default_aux_model: str = ""
+
+    # -- Hooks --
+
+    def get_hostname(self) -> str:
+        if self.hostname:
+            return self.hostname
+        if self.base_url:
+            return urlparse(self.base_url).hostname or ""
+        return ""
+
+    def prepare_messages(self, messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        """Provider-specific message preprocessing. Default: pass-through."""
+        return messages
+
+    def build_extra_body(self, **context: Any) -> dict[str, Any]:
+        """Provider-specific extra_body fields."""
+        return {}
+
+    def build_api_kwargs_extras(
+        self,
+        *,
+        reasoning_config: dict | None = None,
+        **context: Any,
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        """把 reasoning 配置拆分到 extra_body 和顶层 kwargs。
+
+        返回 (extra_body_additions, top_level_kwargs)。
+        """
+        return {}, {}
+
+    def get_max_tokens(self, model: str | None) -> int | None:
+        """Default max_tokens cap. Override for per-model limits."""
+        return self.default_max_tokens
+
+    def fetch_models(
+        self,
+        *,
+        api_key: str | None = None,
+        timeout: float = 8.0,
+    ) -> list[str] | None:
+        """Fetch live model list. Returns list of model IDs or None."""
+        url = (self.models_url or "").strip()
+        if not url:
+            if not self.base_url:
+                return None
+            url = self.base_url.rstrip("/") + "/models"
+
+        req = urllib.request.Request(url)
+        if api_key:
+            req.add_header("Authorization", f"Bearer {api_key}")
+        req.add_header("Accept", "application/json")
+        req.add_header("User-Agent", "lbs-agent/0.1")
+        for k, v in self.default_headers.items():
+            req.add_header(k, v)
+
+        try:
+            with urllib.request.urlopen(req, timeout=timeout) as resp:
+                data = json.loads(resp.read().decode())
+            items = data if isinstance(data, list) else data.get("data", [])
+            return [m["id"] for m in items if isinstance(m, dict) and "id" in m]
+        except Exception as exc:
+            logger.debug("fetch_models(%s): %s", self.name, exc)
+            return None

package/runtime/core/llm/redact.py

@@ -0,0 +1,142 @@
+"""
+Redact — 增强版敏感信息脱敏。
+
+对标 Hermes redact.py。覆盖：
+- API key (sk-xxx, sk-ant-xxx 等)
+- Bearer token
+- password=xxx, api_key=xxx 等 key=value 模式
+- URL query params 中的 token/key
+- URL userinfo (user:pass@host)
+- 手机号
+- 邮箱
+- 环境变量赋值 (export KEY=value)
+- JSON 中的敏感字段
+"""
+from __future__ import annotations
+
+import re
+import logging
+
+logger = logging.getLogger(__name__)
+
+# ── 正则模式 ───────────────────────────────────────
+
+# API keys
+_RE_API_KEY = re.compile(r"(sk-[a-zA-Z0-9]{8})[a-zA-Z0-9_-]+", re.I)
+_RE_API_KEY_ANT = re.compile(r"(sk-ant-[a-zA-Z0-9]{8})[a-zA-Z0-9_-]+", re.I)
+
+# Bearer tokens
+_RE_BEARER = re.compile(r"(Bearer\s+)[a-zA-Z0-9._-]{8,}", re.I)
+
+# key=value patterns
+_SENSITIVE_KEYWORDS = ("api_key", "api_secret", "password", "token", "secret", "access_key", "private_key")
+_RE_KEY_VALUE = re.compile(
+    r"(?i)(" + "|".join(_SENSITIVE_KEYWORDS) + r")" + r"[\'\"\"]?\s*[:=]\s*[\'\"\"]?[^\'\"\s,}\]]+",
+    re.I,
+)
+
+# URL query params with token/key
+_RE_URL_QUERY_TOKEN = re.compile(
+    r"([?&](?:token|key|api_key|access_token|secret|password)=)[^&\s]+",
+    re.I,
+)
+
+# URL userinfo (user:pass@host)
+_RE_URL_USERINFO = re.compile(
+    r"(://[^/@\s]+:)[^@\s]+(@)",
+    re.I,
+)
+
+# Phone numbers (Chinese: 1xx-xxxx-xxxx)
+_RE_PHONE = re.compile(r"(1[3-9]\d)\d{4}(\d{4})")
+
+# Email (partial mask)
+_RE_EMAIL = re.compile(r"([a-zA-Z0-9._%+-]{2})[a-zA-Z0-9._%+-]*(@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,})")
+
+# Environment variable assignment
+_RE_ENV_ASSIGN = re.compile(
+    r"(export\s+[" + "|".join(k.upper() for k in _SENSITIVE_KEYWORDS) + r"]+=)" + r'["\']?[^"\'\s]+',
+    re.I,
+)
+
+# JSON sensitive fields
+_RE_JSON_SENSITIVE = re.compile(
+    r'("(?:' + "|".join(_SENSITIVE_KEYWORDS) + r')"\s*:\s*")[^"]+(")',
+    re.I,
+)
+
+
+def redact_sensitive_text(text: str) -> str:
+    """脱敏文本中的敏感信息。
+
+    Args:
+        text: 原始文本
+
+    Returns:
+        脱敏后的文本
+    """
+    if not text:
+        return text
+
+    # API keys
+    text = _RE_API_KEY.sub(r"\1***", text)
+    text = _RE_API_KEY_ANT.sub(r"\1***", text)
+
+    # Bearer tokens
+    text = _RE_BEARER.sub(r"\1***", text)
+
+    # key=value patterns
+    def _replace_kv(m):
+        keyword = m.group(1)
+        return f"{keyword}=***"
+    text = _RE_KEY_VALUE.sub(_replace_kv, text)
+
+    # URL query params
+    text = _RE_URL_QUERY_TOKEN.sub(r"\1***", text)
+
+    # URL userinfo
+    text = _RE_URL_USERINFO.sub(r"\1***\2", text)
+
+    # Phone numbers
+    text = _RE_PHONE.sub(r"\1****\2", text)
+
+    # Email
+    text = _RE_EMAIL.sub(r"\1***\2", text)
+
+    # Environment variable assignment
+    def _replace_env(m):
+        return m.group(1) + "***"
+    text = _RE_ENV_ASSIGN.sub(_replace_env, text)
+
+    # JSON sensitive fields
+    def _replace_json(m):
+        return m.group(1) + "***" + m.group(2)
+    text = _RE_JSON_SENSITIVE.sub(_replace_json, text)
+
+    return text
+
+
+def mask_secret(secret: str, visible: int = 4) -> str:
+    """遮罩密钥，只保留前 visible 位。
+
+    Args:
+        secret: 原始密钥
+        visible: 前可见字符数
+
+    Returns:
+        遮罩后的字符串
+    """
+    if not secret or len(secret) <= visible:
+        return "***"
+    return secret[:visible] + "*" * (len(secret) - visible - 2) + secret[-2:]
+
+
+class RedactingFormatter(logging.Formatter):
+    """日志格式化器 — 自动脱敏日志中的敏感信息。"""
+
+    def __init__(self, fmt=None, datefmt=None, style="%", **kwargs):
+        super().__init__(fmt, datefmt, style, **kwargs)
+
+    def format(self, record: logging.LogRecord) -> str:
+        msg = super().format(record)
+        return redact_sensitive_text(msg)

package/runtime/core/llm/registry/__init__.py

@@ -0,0 +1,50 @@
+"""
+供应商注册表 —— 所有已知 provider profile 的中央目录。
+
+支持三种发现模式：1. 显式 register_provider() 调用 2. 目录扫描（plugins/）3. 运行时注册自……om config
+
+Migrated from LBS Agent providers/__init__.py, simplified.
+"""
+
+from __future__ import annotations
+
+import logging
+from typing import Optional
+
+from runtime.core.llm.provider import ProviderProfile
+
+logger = logging.getLogger(__name__)
+
+_REGISTRY: dict[str, ProviderProfile] = {}
+_ALIASES: dict[str, str] = {}
+
+
+def register_provider(profile: ProviderProfile) -> None:
+    """Register a provider profile by name and aliases."""
+    _REGISTRY[profile.name] = profile
+    for alias in profile.aliases:
+        _ALIASES[alias] = profile.name
+
+
+def get_provider(name: str) -> Optional[ProviderProfile]:
+    """Look up provider by name or alias. Returns None if not found."""
+    canonical = _ALIASES.get(name, name)
+    return _REGISTRY.get(canonical)
+
+
+def list_providers() -> list[ProviderProfile]:
+    """Return all registered provider profiles."""
+    seen: set[int] = set()
+    result: list[ProviderProfile] = []
+    for profile in _REGISTRY.values():
+        pid = id(profile)
+        if pid not in seen:
+            seen.add(pid)
+            result.append(profile)
+    return result
+
+
+def clear_registry() -> None:
+    """Clear all registrations. Mainly for testing."""
+    _REGISTRY.clear()
+    _ALIASES.clear()

package/runtime/core/llm/registry/models.yaml

@@ -0,0 +1,233 @@
+default:
+  provider: seeyon
+  model: gpt-5.5
+  soul: 二次元
+builtin:
+  openrouter:
+    api_key: ${OPENROUTER_API_KEY}
+    base_url: https://openrouter.ai/api/v1
+    api_mode: chat_completions
+  anthropic:
+    api_key: ${ANTHROPIC_API_KEY}
+    base_url: https://api.anthropic.com
+    api_mode: anthropic_messages
+  openai:
+    api_key: ${OPENAI_API_KEY}
+    base_url: https://api.openai.com/v1
+    api_mode: chat_completions
+  deepseek:
+    api_key: ${DEEPSEEK_API_KEY}
+    base_url: https://api.deepseek.com/v1
+    api_mode: chat_completions
+    models:
+    - deepseek-chat
+    - deepseek-reasoner
+  gemini:
+    api_key: ${GEMINI_API_KEY}
+    base_url: https://generativelanguage.googleapis.com/v1beta/openai
+    api_mode: chat_completions
+  xai:
+    api_key: ${XAI_API_KEY}
+    base_url: https://api.x.ai/v1
+    api_mode: chat_completions
+  volcano:
+    api_key: ${VOLCANO_API_KEY}
+    base_url: https://ark.cn-beijing.volces.com/api/v3
+    api_mode: chat_completions
+    models:
+    - doubao-pro-32k
+    - doubao-pro-128k
+    - doubao-lite-32k
+  qwen:
+    api_key: ${QWEN_API_KEY}
+    base_url: https://dashscope.aliyuncs.com/compatible-mode/v1
+    api_mode: chat_completions
+  zai:
+    api_key: ${ZAI_API_KEY}
+    base_url: https://open.bigmodel.cn/api/paas/v4
+    api_mode: chat_completions
+    models:
+    - glm-4-plus
+    - glm-4-air
+    - glm-4-flash
+  kimi:
+    api_key: ${KIMI_API_KEY}
+    base_url: https://api.moonshot.cn/v1
+    api_mode: chat_completions
+custom:
+  seeyon:
+    display_name: 致远云
+    api_key: ${SEETON_API_KEY}
+    base_url: https://api.seeyoncloud.com
+    api_mode: anthropic_messages
+    models:
+    - deepseek-v4-flash
+    - deepseek-v4-pro
+    - glm-4.7
+    - glm-5
+    - glm-5.1
+    - glm-5.2
+    - glm-5v-turbo
+    - gpt-5.3-codex-spark
+    - gpt-5.4
+    - gpt-5.4-mini
+    - gpt-5.5
+    - gpt-5.6-luna
+    - gpt-5.6-sol
+    - gpt-5.6-terra
+    - gpt-image-2
+    - qwen3.7-max
+    capabilities:
+      default:
+        supports_text: true
+        supports_tools: true
+        supports_vision: false
+        supports_json: true
+        context_length: 128000
+      gpt-5.4:
+        supports_text: true
+        supports_tools: true
+        supports_vision: true
+        supports_json: true
+        context_length: 256000
+        preferred_for:
+        - browser
+        - code
+        - audit
+        - security
+        - vision
+        - document
+      gpt-5.4-mini:
+        supports_text: true
+        supports_tools: true
+        supports_vision: true
+        supports_json: true
+        context_length: 256000
+        preferred_for:
+        - verifier
+        - quick-check
+      gpt-5.5:
+        supports_text: true
+        supports_tools: true
+        supports_vision: true
+        supports_json: true
+        context_length: 256000
+        preferred_for:
+        - fallback
+        - complex-reasoning
+      gpt-5.6-terra:
+        supports_text: true
+        supports_tools: true
+        supports_vision: true
+        supports_json: true
+        context_length: 256000
+      gpt-5.6-luna:
+        supports_text: true
+        supports_tools: true
+        supports_vision: true
+        supports_json: true
+        context_length: 256000
+      gpt-5.6-sol:
+        supports_text: true
+        supports_tools: true
+        supports_vision: true
+        supports_json: true
+        context_length: 256000
+      glm-5.2:
+        supports_text: true
+        supports_tools: true
+        supports_vision: false
+        supports_json: true
+        context_length: 256000
+      deepseek-v4-flash:
+        supports_text: true
+        supports_tools: true
+        supports_vision: false
+        supports_json: true
+        context_length: 128000
+      deepseek-v4-pro:
+        supports_text: true
+        supports_tools: true
+        supports_vision: false
+        supports_json: true
+        context_length: 128000
+      qwen3.7-max:
+        supports_text: true
+        supports_tools: true
+        supports_vision: false
+        supports_json: true
+        context_length: 128000
+        preferred_for:
+        - fallback
+        - coding
+    api_keys:
+    - sk-19VlvDegfZc9jeFdceu2f1gCyhsPG7li5asb6RnchWgCMd2i
+    max_retries: 0
+    request_timeout: 0
+  volcano-coding:
+    display_name: 火山 Coding
+    api_key: ${VOLCANO_CODING_API_KEY}
+    base_url: https://ark.cn-beijing.volces.com/api/coding
+    api_mode: anthropic_messages
+    models:
+    - claude-sonnet-4
+    - doubao-pro-32k
+  local-ollama:
+    display_name: 本地 Ollama
+    api_key: ${OLLAMA_API_KEY:-ollama}
+    base_url: http://localhost:11434/v1
+    api_mode: chat_completions
+    models:
+    - qwen2.5:14b
+    - llama3.1:8b
+  minmax:
+    display_name: MiniMax
+    api_key: ${MINIMAX_API_KEY}
+    base_url: https://api.minimaxi.com/anthropic
+    api_mode: anthropic_messages
+    models:
+    - MiniMax-M2
+    - MiniMax-M2.1
+    - MiniMax-M2.1-highspeed
+    - MiniMax-M2.5
+    - MiniMax-M2.5-highspeed
+    - MiniMax-M2.7
+    - MiniMax-M2.7-highspeed
+    - MiniMax-M3
+    capabilities:
+      default:
+        supports_text: true
+        supports_tools: true
+        supports_vision: false
+        supports_json: true
+      MiniMax-M3:
+        supports_text: true
+        supports_tools: true
+        supports_vision: true
+        supports_json: true
+        preferred_for:
+        - general
+        - fallback
+  灵眸:
+    display_name: 灵眸
+    base_url: https://api.lmuai.com
+    api_mode: anthropic_messages
+    api_key: ${LMUAI_API_KEY}
+    models:
+    - MiniMax-M2.5
+    - MiniMax-M2.5-highspeed
+    - MiniMax-M2.7
+    - MiniMax-M2.7-highspeed
+    - MiniMax-M3
+    - deepseek-v4-flash
+    - deepseek-v4-pro
+    - glm-5
+    - glm-5.1
+    - glm-5.2
+    - kimi-k2.5
+    - kimi-k2.6
+    - kimi-k2.7-code
+    - qwen3.6-flash
+    - qwen3.6-plus
+    - qwen3.7-max
+    - qwen3.7-plus