mula
/
dashboard-nanobot
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

完整的部署文件

main
mula.liu 2026-04-13 16:36:58 +08:00
parent 4d7aa42a18
commit e8932bec17
25 changed files with 503 additions and 49 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
.dockerignore
View File

@ -6,7 +6,14 @@ frontend/node_modules
frontend/dist
backend/venv
data
data/*
!data/templates/
!data/templates/**
!data/skills/
!data/skills/**
!data/model/
data/model/*
!data/model/README.md
workspace
**/__pycache__

7
.env.full.example
View File

@ -1,9 +1,8 @@
# Public exposed port (only nginx is exposed)
NGINX_PORT=8080
# Required absolute host paths.
# They must exist or be creatable by the deployment user and writable by docker daemon.
HOST_DATA_ROOT=/opt/dashboard-nanobot/data
# Project data is always mounted from the repository root `./data`.
# Only workspace root still needs an absolute host path.
HOST_BOTS_WORKSPACE_ROOT=/opt/dashboard-nanobot/workspace/bots
# Optional custom image tags
@ -72,7 +71,7 @@ WORKSPACE_DOWNLOAD_EXTENSIONS=.pdf,.doc,.docx,.xls,.xlsx,.xlsm,.ppt,.pptx,.odt,.
# Local speech-to-text (Whisper via whisper.cpp model file)
STT_ENABLED=true
STT_MODEL=ggml-small-q8_0.bin
STT_MODEL_DIR=${HOST_DATA_ROOT}/model
STT_MODEL_DIR=/app/data/model
STT_DEVICE=cpu
STT_MAX_AUDIO_SECONDS=20
STT_DEFAULT_LANGUAGE=zh

9
.env.prod.example
View File

@ -1,9 +1,8 @@
# Public exposed port (only nginx is exposed)
NGINX_PORT=8080
# REQUIRED absolute host paths.
# They must exist and be writable by docker daemon.
HOST_DATA_ROOT=/opt/dashboard-nanobot/data
# Project data is always mounted from the repository root `./data`.
# Only workspace root still needs an absolute host path.
HOST_BOTS_WORKSPACE_ROOT=/opt/dashboard-nanobot/workspace/bots
# Optional custom image tags
@ -25,7 +24,7 @@ NPM_REGISTRY=https://registry.npmmirror.com
# Database (choose one: SQLite / PostgreSQL / MySQL)
# SQLite example:
# DATABASE_URL=sqlite:///${HOST_DATA_ROOT}/nanobot_dashboard.db
# DATABASE_URL=sqlite:////app/data/nanobot_dashboard.db
# PostgreSQL example:
# DATABASE_URL=postgresql+psycopg://user:password@127.0.0.1:5432/nanobot_dashboard
# MySQL example:
@ -65,7 +64,7 @@ WORKSPACE_DOWNLOAD_EXTENSIONS=.pdf,.doc,.docx,.xls,.xlsx,.xlsm,.ppt,.pptx,.odt,.
# Local speech-to-text (Whisper via whisper.cpp model file)
STT_ENABLED=true
STT_MODEL=ggml-small-q8_0.bin
STT_MODEL_DIR=${HOST_DATA_ROOT}/model
STT_MODEL_DIR=/app/data/model
STT_DEVICE=cpu
STT_MAX_AUDIO_SECONDS=20
STT_DEFAULT_LANGUAGE=zh

9
.gitignore vendored
View File

@ -30,7 +30,14 @@ backend/__pycache__/
backend/*.log
# Project runtime data (generated locally)
data/
data/*
!data/templates/
!data/templates/**
!data/skills/
!data/skills/**
!data/model/
data/model/*
!data/model/README.md
workspace/
engines/

22
README.md
View File

@ -13,7 +13,7 @@ Dashboard Nanobot 是面向 `nanobot` 的控制平面项目，提供镜像管理
- `USER.md`
- `TOOLS.md`
- `IDENTITY.md`
- 模板管理：系统级模板改为文件化配置（`backend/templates/agent_md_templates.json` 与 `backend/templates/topic_presets.json`）。
- 模板管理：系统级模板改为文件化配置（`data/templates/agent_md_templates.json` 与 `data/templates/topic_presets.json`）。
- 2D 运维 Dashboard：Bot 列表、启停、命令发送、日志流、遥测。
- UI 全局支持：Light/Dark 切换、中文/English 切换。
@ -57,6 +57,13 @@ graph TD
- 架构设计：`design/architecture.md`
- 数据库设计：`design/database.md`
## 默认资源
- 项目根目录 `data/templates/` 保存默认模板资源，会在初始化时同步到运行时数据目录。
- 项目根目录 `data/skills/` 保存默认 skill 包，会在数据库初始化阶段自动注册到 `skill_market_item`。
- `data/model/` 不包含语音识别模型文件；模型需要用户自行下载放入该目录或 `STT_MODEL_DIR` 指向的目录。
- 如果语音模型缺失，后端启动时会打印明确告警，但不会阻断服务启动。
## 环境变量配置
- 后端：
@ -102,10 +109,10 @@ graph TD
1. 准备部署变量
- 复制 `.env.prod.example` 为 `.env.prod`（位于项目根目录）
- 配置绝对路径：
- `HOST_DATA_ROOT`
- `data/` 会自动映射到宿主机项目根目录下的 `./data`
- 只需要配置绝对路径：
- `HOST_BOTS_WORKSPACE_ROOT`
- 如启用本地语音识别，请将 Whisper `.bin` 模型文件放到 `${HOST_DATA_ROOT}/model/`
- 如启用本地语音识别，请将 Whisper `.bin` 模型文件放到宿主机项目根目录的 `data/model/`
并让 `STT_MODEL` 指向完整文件名，例如 `ggml-small-q8_0.bin`
- 中国网络建议配置加速项：
- `PIP_INDEX_URL`、`PIP_TRUSTED_HOST`
@ -122,6 +129,7 @@ graph TD
- `backend` 不开放宿主机端口，仅在内部网络被 Nginx 访问。
- 上传大小使用单一参数 `UPLOAD_MAX_MB` 控制（后端校验 + Nginx 限制）。
- 必须挂载 `/var/run/docker.sock`，否则后端无法操作 Bot 镜像与容器。
- `data/` 始终绑定到宿主机项目根目录下的 `./data`，其中模板、默认 skills、语音模型和运行数据都落在这里。
- `HOST_BOTS_WORKSPACE_ROOT` 必须是宿主机绝对路径，并且在 `docker-compose.prod.yml` 中以“同路径”挂载到后端容器。
原因：后端通过 Docker API 创建 Bot 容器时，使用的是宿主机可见的 bind 路径。
- 语音识别当前基于 `pywhispercpp==1.3.1` + Whisper `.bin` 模型文件，不使用 `faster-whisper`。
@ -144,13 +152,13 @@ graph TD
1. 准备部署变量
- 复制 `.env.full.example` 为 `.env.full`
- `data/` 会自动映射到宿主机项目根目录下的 `./data`
- 必填修改：
- `HOST_DATA_ROOT`
- `HOST_BOTS_WORKSPACE_ROOT`
- `POSTGRES_SUPERPASSWORD`
- `POSTGRES_APP_PASSWORD`
- `PANEL_ACCESS_PASSWORD`
- 如启用本地语音识别，请将 Whisper `.bin` 模型文件放到 `${HOST_DATA_ROOT}/model/`
- 如启用本地语音识别，请将 Whisper `.bin` 模型文件放到宿主机项目根目录的 `data/model/`
2. 启动完整栈
- `./scripts/deploy-full.sh`
3. 访问
@ -173,5 +181,5 @@ graph TD
### 注意事项
- `deploy-prod.sh` 和 `deploy-full.sh` 使用的是两套 compose 文件，但复用了相同容器名，不能同时在同一台机器上并行启动。
- PostgreSQL 数据默认落盘到 `${HOST_DATA_ROOT}/postgres`，Redis 数据默认落盘到 `${HOST_DATA_ROOT}/redis`。
- PostgreSQL 数据默认落盘到宿主机项目根目录 `./data/postgres`，Redis 数据默认落盘到 `./data/redis`。
- 如果你只想保留前后端容器，继续使用 `deploy-prod.sh`；如果希望把依赖也打包进来，使用 `deploy-full.sh`。

1
backend/Dockerfile
View File

@ -20,6 +20,7 @@ RUN if [ -n "${PIP_INDEX_URL}" ]; then pip config set global.index-url "${PIP_IN
&& pip install -r requirements.txt
COPY backend/ /app/backend/
COPY data/ /app/data/
EXPOSE 8000

5
backend/api/system_router.py
View File

@ -1,5 +1,6 @@
from fastapi import APIRouter, HTTPException
from core.speech_service import inspect_speech_model_status
from core.utils import _get_default_system_timezone
from schemas.system import SystemTemplatesUpdateRequest
from services.platform_service import get_platform_settings_snapshot, get_speech_runtime_settings
@ -17,6 +18,7 @@ def get_system_defaults():
md_templates = get_agent_md_templates()
platform_settings = get_platform_settings_snapshot()
speech_settings = get_speech_runtime_settings()
model_status = inspect_speech_model_status()
return {
"templates": md_templates,
"limits": {
@ -41,6 +43,9 @@ def get_system_defaults():
"device": speech_settings["device"],
"max_audio_seconds": speech_settings["max_audio_seconds"],
"default_language": speech_settings["default_language"],
"ready": model_status["ready"],
"message": model_status["message"],
"expected_path": model_status["expected_path"],
},
}

24
backend/bootstrap/app_runtime.py
View File

@ -6,10 +6,12 @@ from sqlmodel import Session, select
from core.cache import cache
from core.database import engine, init_database
from core.docker_instance import docker_manager
from core.speech_service import inspect_speech_model_status
from core.settings import DATABASE_URL_DISPLAY, REDIS_ENABLED
from models.bot import BotInstance
from services.bot_storage_service import _migrate_bot_resources_store
from services.platform_service import prune_expired_activity_events
from services.default_assets_service import ensure_default_skill_market_items, ensure_runtime_data_assets
from services.platform_service import get_speech_runtime_settings, prune_expired_activity_events
from services.runtime_service import docker_callback, set_main_loop
@ -29,11 +31,31 @@ def register_app_runtime(app: FastAPI) -> None:
current_loop = asyncio.get_running_loop()
app.state.main_loop = current_loop
set_main_loop(current_loop)
asset_report = ensure_runtime_data_assets()
if asset_report["templates_initialized"] or asset_report["skills_synchronized"]:
print(
"[init] 默认资源已同步 "
f"(templates={asset_report['templates_initialized']}, skills={asset_report['skills_synchronized']})"
)
init_database()
with Session(engine) as session:
skill_report = ensure_default_skill_market_items(session)
if skill_report["created"] or skill_report["updated"]:
print(
"[init] 默认 skills 已入库 "
f"(created={len(skill_report['created'])}, updated={len(skill_report['updated'])})"
)
prune_expired_activity_events(session, force=True)
bots = session.exec(select(BotInstance)).all()
for bot in bots:
_migrate_bot_resources_store(bot.id)
docker_manager.ensure_monitor(bot.id, docker_callback)
speech_settings = get_speech_runtime_settings()
model_status = inspect_speech_model_status()
if speech_settings["enabled"]:
if model_status["ready"]:
print(f"🎙️ 语音识别模型就绪: {model_status['resolved_path']}")
else:
hint = f"，请将模型文件放到 {model_status['expected_path']}" if model_status["expected_path"] else ""
print(f"⚠️ 语音识别模型未就绪: {model_status['message']}{hint}")
print("✅ 启动自检完成")

16
backend/core/settings.py
View File

@ -117,6 +117,13 @@ DATA_ROOT: Final[str] = _normalize_dir_path(os.getenv("DATA_ROOT", str(PROJECT_R
BOTS_WORKSPACE_ROOT: Final[str] = _normalize_dir_path(
os.getenv("BOTS_WORKSPACE_ROOT", str(PROJECT_ROOT / "workspace" / "bots"))
)
BUNDLED_DATA_ROOT: Final[Path] = (PROJECT_ROOT / "data").resolve()
RUNTIME_DATA_ROOT: Final[Path] = Path(DATA_ROOT).resolve()
BUNDLED_TEMPLATES_ROOT: Final[Path] = (BUNDLED_DATA_ROOT / "templates").resolve()
RUNTIME_TEMPLATES_ROOT: Final[Path] = (RUNTIME_DATA_ROOT / "templates").resolve()
BUNDLED_SKILLS_ROOT: Final[Path] = (BUNDLED_DATA_ROOT / "skills").resolve()
RUNTIME_SKILLS_ROOT: Final[Path] = (RUNTIME_DATA_ROOT / "skills").resolve()
RUNTIME_MODEL_ROOT: Final[Path] = (RUNTIME_DATA_ROOT / "model").resolve()
def _normalize_database_url(url: str) -> str:
@ -205,7 +212,7 @@ DEFAULT_WORKSPACE_DOWNLOAD_EXTENSIONS: Final[tuple[str, ...]] = (
)
STT_ENABLED_DEFAULT: Final[bool] = True
STT_MODEL: Final[str] = str(os.getenv("STT_MODEL") or "ggml-small-q8_0.bin").strip()
_DEFAULT_STT_MODEL_DIR: Final[Path] = (Path(DATA_ROOT) / "model").resolve()
_DEFAULT_STT_MODEL_DIR: Final[Path] = RUNTIME_MODEL_ROOT
_configured_stt_model_dir = _normalize_dir_path(os.getenv("STT_MODEL_DIR", str(_DEFAULT_STT_MODEL_DIR)))
if _configured_stt_model_dir and not Path(_configured_stt_model_dir).exists() and _DEFAULT_STT_MODEL_DIR.exists():
STT_MODEL_DIR: Final[str] = str(_DEFAULT_STT_MODEL_DIR)
@ -240,6 +247,7 @@ APP_HOST: Final[str] = str(os.getenv("APP_HOST") or "0.0.0.0").strip()
APP_PORT: Final[int] = _env_int("APP_PORT", 8000, 1, 65535)
APP_RELOAD: Final[bool] = _env_bool("APP_RELOAD", False)
TEMPLATE_ROOT: Final[Path] = (BACKEND_ROOT / "templates").resolve()
AGENT_MD_TEMPLATES_FILE: Final[Path] = TEMPLATE_ROOT / "agent_md_templates.json"
TOPIC_PRESETS_TEMPLATES_FILE: Final[Path] = TEMPLATE_ROOT / "topic_presets.json"
AGENT_MD_TEMPLATES_FILE: Final[Path] = RUNTIME_TEMPLATES_ROOT / "agent_md_templates.json"
TOPIC_PRESETS_TEMPLATES_FILE: Final[Path] = RUNTIME_TEMPLATES_ROOT / "topic_presets.json"
BUNDLED_AGENT_MD_TEMPLATES_FILE: Final[Path] = BUNDLED_TEMPLATES_ROOT / "agent_md_templates.json"
BUNDLED_TOPIC_PRESETS_TEMPLATES_FILE: Final[Path] = BUNDLED_TEMPLATES_ROOT / "topic_presets.json"

33
backend/core/speech_service.py
View File

@ -24,6 +24,39 @@ class SpeechDurationError(SpeechServiceError):
pass
def inspect_speech_model_status() -> Dict[str, Any]:
service = WhisperSpeechService()
model = str(STT_MODEL or "").strip()
model_dir = str(STT_MODEL_DIR or "").strip()
expected_path = ""
if model:
if any(sep in model for sep in ("/", "\\")):
expected_path = str(Path(model).expanduser())
elif model_dir:
expected_path = str((Path(model_dir).expanduser() / model).resolve())
try:
resolved_path = service._resolve_model_source()
return {
"ready": True,
"model": model,
"model_dir": model_dir,
"expected_path": expected_path or resolved_path,
"resolved_path": resolved_path,
"message": "",
}
except SpeechServiceError as exc:
return {
"ready": False,
"model": model,
"model_dir": model_dir,
"expected_path": expected_path,
"resolved_path": "",
"message": str(exc),
}
class WhisperSpeechService:
def __init__(self) -> None:
self._model: Any = None

206
backend/services/default_assets_service.py 100644
View File

@ -0,0 +1,206 @@
from __future__ import annotations
import json
import os
import re
import shutil
import zipfile
from pathlib import Path
from typing import Any, Dict, List
from sqlmodel import Session, select
from core.settings import (
AGENT_MD_TEMPLATES_FILE,
BUNDLED_AGENT_MD_TEMPLATES_FILE,
BUNDLED_SKILLS_ROOT,
BUNDLED_TOPIC_PRESETS_TEMPLATES_FILE,
DATA_ROOT,
RUNTIME_MODEL_ROOT,
RUNTIME_SKILLS_ROOT,
RUNTIME_TEMPLATES_ROOT,
TOPIC_PRESETS_TEMPLATES_FILE,
)
from core.utils import (
_is_ignored_skill_zip_top_level,
_is_valid_top_level_skill_name,
_read_description_from_text,
_sanitize_skill_market_key,
)
from models.skill import SkillMarketItem
def _copy_if_missing(src: Path, dst: Path) -> bool:
if not src.exists() or not src.is_file():
return False
if src.resolve() == dst.resolve() if dst.exists() else False:
return False
if dst.exists():
return False
dst.parent.mkdir(parents=True, exist_ok=True)
shutil.copy2(src, dst)
return True
def _copy_if_different(src: Path, dst: Path) -> bool:
if not src.exists() or not src.is_file():
return False
if src.resolve() == dst.resolve() if dst.exists() else False:
return False
dst.parent.mkdir(parents=True, exist_ok=True)
if dst.exists():
try:
if src.stat().st_size == dst.stat().st_size and src.read_bytes() == dst.read_bytes():
return False
except Exception:
pass
shutil.copy2(src, dst)
return True
def _iter_bundled_skill_packages() -> List[Path]:
if not BUNDLED_SKILLS_ROOT.exists() or not BUNDLED_SKILLS_ROOT.is_dir():
return []
return sorted(path for path in BUNDLED_SKILLS_ROOT.iterdir() if path.is_file() and path.suffix.lower() == ".zip")
def ensure_runtime_data_assets() -> Dict[str, int]:
Path(DATA_ROOT).mkdir(parents=True, exist_ok=True)
RUNTIME_TEMPLATES_ROOT.mkdir(parents=True, exist_ok=True)
RUNTIME_SKILLS_ROOT.mkdir(parents=True, exist_ok=True)
RUNTIME_MODEL_ROOT.mkdir(parents=True, exist_ok=True)
templates_initialized = 0
skills_synchronized = 0
if _copy_if_missing(BUNDLED_AGENT_MD_TEMPLATES_FILE, AGENT_MD_TEMPLATES_FILE):
templates_initialized += 1
if _copy_if_missing(BUNDLED_TOPIC_PRESETS_TEMPLATES_FILE, TOPIC_PRESETS_TEMPLATES_FILE):
templates_initialized += 1
for src in _iter_bundled_skill_packages():
if _copy_if_different(src, RUNTIME_SKILLS_ROOT / src.name):
skills_synchronized += 1
return {
"templates_initialized": templates_initialized,
"skills_synchronized": skills_synchronized,
}
def _extract_skill_zip_summary(zip_path: Path) -> Dict[str, Any]:
entry_names: List[str] = []
description = ""
with zipfile.ZipFile(zip_path) as archive:
members = archive.infolist()
file_members = [member for member in members if not member.is_dir()]
for member in file_members:
raw_name = str(member.filename or "").replace("\\", "/").lstrip("/")
if not raw_name:
continue
first = raw_name.split("/", 1)[0].strip()
if _is_ignored_skill_zip_top_level(first):
continue
if _is_valid_top_level_skill_name(first) and first not in entry_names:
entry_names.append(first)
candidates = sorted(
[
str(member.filename or "").replace("\\", "/").lstrip("/")
for member in file_members
if str(member.filename or "").replace("\\", "/").rsplit("/", 1)[-1].lower()
in {"skill.md", "readme.md"}
],
key=lambda value: (value.count("/"), value.lower()),
)
for candidate in candidates:
try:
with archive.open(candidate, "r") as file:
preview = file.read(4096).decode("utf-8", errors="ignore")
description = _read_description_from_text(preview)
if description:
break
except Exception:
continue
return {
"entry_names": entry_names,
"description": description,
}
def _default_display_name(stem: str) -> str:
chunks = [chunk for chunk in re.split(r"[-_]+", str(stem or "").strip()) if chunk]
if not chunks:
return "Skill"
return " ".join(chunk.upper() if chunk.isupper() else chunk.capitalize() for chunk in chunks)
def _resolve_unique_skill_key(existing_keys: set[str], preferred_key: str) -> str:
base_key = _sanitize_skill_market_key(preferred_key) or "skill"
candidate = base_key
counter = 2
while candidate in existing_keys:
candidate = f"{base_key}-{counter}"
counter += 1
existing_keys.add(candidate)
return candidate
def ensure_default_skill_market_items(session: Session) -> Dict[str, List[str]]:
report: Dict[str, List[str]] = {"created": [], "updated": []}
default_packages = _iter_bundled_skill_packages()
if not default_packages:
return report
rows = session.exec(select(SkillMarketItem)).all()
existing_by_zip = {str(row.zip_filename or "").strip(): row for row in rows if str(row.zip_filename or "").strip()}
existing_keys = {str(row.skill_key or "").strip() for row in rows if str(row.skill_key or "").strip()}
for bundled_path in default_packages:
runtime_path = RUNTIME_SKILLS_ROOT / bundled_path.name
source_path = runtime_path if runtime_path.exists() else bundled_path
try:
summary = _extract_skill_zip_summary(source_path)
except Exception:
continue
zip_filename = bundled_path.name
entry_names_json = json.dumps(summary["entry_names"], ensure_ascii=False)
display_name = _default_display_name((summary["entry_names"] or [bundled_path.stem])[0])
zip_size_bytes = int(source_path.stat().st_size) if source_path.exists() else 0
row = existing_by_zip.get(zip_filename)
if row is None:
row = SkillMarketItem(
skill_key=_resolve_unique_skill_key(existing_keys, bundled_path.stem),
display_name=display_name,
description=str(summary["description"] or "").strip(),
zip_filename=zip_filename,
zip_size_bytes=zip_size_bytes,
entry_names_json=entry_names_json,
)
session.add(row)
existing_by_zip[zip_filename] = row
report["created"].append(zip_filename)
continue
changed = False
if int(row.zip_size_bytes or 0) != zip_size_bytes:
row.zip_size_bytes = zip_size_bytes
changed = True
if str(row.entry_names_json or "") != entry_names_json:
row.entry_names_json = entry_names_json
changed = True
if not str(row.display_name or "").strip():
row.display_name = display_name
changed = True
if not str(row.description or "").strip() and str(summary["description"] or "").strip():
row.description = str(summary["description"] or "").strip()
changed = True
if changed:
report["updated"].append(zip_filename)
if report["created"] or report["updated"]:
session.commit()
return report

32
backend/services/template_service.py
View File

@ -2,21 +2,30 @@ from __future__ import annotations
from typing import Any, Dict, List
from core.settings import AGENT_MD_TEMPLATES_FILE, TOPIC_PRESETS_TEMPLATES_FILE
from core.settings import (
AGENT_MD_TEMPLATES_FILE,
BUNDLED_AGENT_MD_TEMPLATES_FILE,
BUNDLED_TOPIC_PRESETS_TEMPLATES_FILE,
TOPIC_PRESETS_TEMPLATES_FILE,
)
TEMPLATE_KEYS = ("agents_md", "soul_md", "user_md", "tools_md", "identity_md")
def _load_json_object(path: str) -> Dict[str, Any]:
def _load_json_object(path: str, fallback_path: str = "") -> Dict[str, Any]:
import json
try:
with open(path, "r", encoding="utf-8") as file:
data = json.load(file)
if isinstance(data, dict):
return data
except Exception:
pass
for candidate in [path, fallback_path]:
candidate = str(candidate or "").strip()
if not candidate:
continue
try:
with open(candidate, "r", encoding="utf-8") as file:
data = json.load(file)
if isinstance(data, dict):
return data
except Exception:
continue
return {}
@ -36,12 +45,12 @@ def _write_json_atomic(path: str, payload: Dict[str, Any]) -> None:
def get_agent_md_templates() -> Dict[str, str]:
raw = _load_json_object(str(AGENT_MD_TEMPLATES_FILE))
raw = _load_json_object(str(AGENT_MD_TEMPLATES_FILE), str(BUNDLED_AGENT_MD_TEMPLATES_FILE))
return {key: _normalize_md_text(raw.get(key)) for key in TEMPLATE_KEYS}
def get_topic_presets() -> Dict[str, Any]:
raw = _load_json_object(str(TOPIC_PRESETS_TEMPLATES_FILE))
raw = _load_json_object(str(TOPIC_PRESETS_TEMPLATES_FILE), str(BUNDLED_TOPIC_PRESETS_TEMPLATES_FILE))
presets = raw.get("presets")
if not isinstance(presets, list):
return {"presets": []}
@ -68,4 +77,3 @@ def update_topic_presets(raw: Dict[str, Any]) -> Dict[str, Any]:
def get_agent_template_value(key: str) -> str:
return get_agent_md_templates().get(key, "")

9
data/model/README.md 100644
View File

@ -0,0 +1,9 @@
# Speech Model Directory
This directory is reserved for local Whisper `.bin` model files and is intentionally not committed to source control.
Put the file configured by `STT_MODEL` here, for example:
- `ggml-small-q8_0.bin`
If the model file is missing, the backend will still start, but it will print a startup warning and speech transcription requests will not work until the file is added.

BIN
data/skills/nano-banana-pro.zip 100644
View File

Binary file not shown.

BIN
data/skills/powerpoint-pptx.zip 100644
View File

Binary file not shown.

BIN
data/skills/self-improving-agent.zip 100644
View File

Binary file not shown.

BIN
data/skills/stock-analysis.zip 100644
View File

Binary file not shown.

BIN
data/skills/tavily-search.zip 100644
View File

Binary file not shown.

BIN
data/skills/writing.zip 100644
View File

Binary file not shown.

7
data/templates/agent_md_templates.json 100644
View File

@ -0,0 +1,7 @@
{
"agents_md": "- 优先完成任务目标\n- 操作前先说明意图\n- 输出必须可执行\n\n## 默认输出规范\n\n- 每次执行任务时，在 workspace 中创建新目录保存本次输出。\n- 输出内容默认采用 Markdown（.md）格式。\n- 最终报告需求.md和.htm双格式。",
"soul_md": "你是专业的企业数字员工，表达清晰、可执行。",
"user_md": "- 语言: 中文\n- 风格: 专业\n- 偏好: 简明且有步骤",
"tools_md": "- 谨慎使用 shell\n- 修改文件后复核\n- 失败时说明原因并重试策略",
"identity_md": "- 角色: 企业数字员工\n- 领域: 运维与任务执行"
}

134
data/templates/topic_presets.json 100644
View File

@ -0,0 +1,134 @@
{
"presets": [
{
"id": "politics",
"topic_key": "politics_news",
"name": "时政新闻",
"description": "沉淀国内外时政动态、政策发布与重大公共治理事件，便于集中查看。",
"routing_purpose": "收录与政府决策、政策法规、外交事务及公共治理相关的关键信息。",
"routing_include_when": [
"时政",
"政策",
"法规",
"国务院",
"政府",
"部委",
"人大",
"政协",
"外交",
"国际关系",
"白宫",
"总统",
"议会",
"election",
"policy"
],
"routing_exclude_when": [
"娱乐",
"明星",
"综艺",
"体育",
"游戏",
"购物",
"种草",
"广告"
],
"routing_examples_positive": [
"国务院发布新一轮宏观政策措施。",
"外交部就国际热点事件发布声明。",
"某国总统宣布新的对外政策方向。"
],
"routing_examples_negative": [
"某明星新剧开播引发热议。",
"某球队转会新闻与赛果分析。",
"数码产品促销与购物推荐汇总。"
],
"routing_priority": 85
},
{
"id": "finance",
"topic_key": "finance_market",
"name": "财经信息",
"description": "聚合宏观经济、市场波动、公司财报与监管政策等财经信息。",
"routing_purpose": "沉淀与资本市场、行业景气、资产价格相关的关键结论与风险提示。",
"routing_include_when": [
"财经",
"金融",
"股市",
"A股",
"港股",
"美股",
"债券",
"汇率",
"利率",
"通胀",
"GDP",
"财报",
"央行",
"market",
"earnings"
],
"routing_exclude_when": [
"娱乐",
"体育",
"游戏",
"影视",
"八卦",
"生活方式",
"旅行攻略"
],
"routing_examples_positive": [
"央行公布最新利率决议并释放政策信号。",
"上市公司发布季度财报并上调全年指引。",
"美元指数走强导致主要货币普遍承压。"
],
"routing_examples_negative": [
"某综艺节目收视排名变化。",
"某球员转会传闻引发讨论。",
"新游上线玩法测评。"
],
"routing_priority": 80
},
{
"id": "tech",
"topic_key": "tech_updates",
"name": "技术资讯",
"description": "追踪 AI、云计算、开源社区与开发工具链的最新技术资讯。",
"routing_purpose": "沉淀技术发布、版本升级、兼容性变更与工程实践建议。",
"routing_include_when": [
"技术",
"开源",
"AI",
"模型",
"大语言模型",
"MCP",
"API",
"SDK",
"发布",
"版本",
"升级",
"breaking change",
"security advisory"
],
"routing_exclude_when": [
"娱乐",
"体育",
"美食",
"旅游",
"情感",
"八卦"
],
"routing_examples_positive": [
"某主流框架发布新版本并调整默认配置。",
"开源项目披露高危安全漏洞并给出修复方案。",
"AI 模型服务更新 API，返回结构发生变化。"
],
"routing_examples_negative": [
"某艺人参加活动造型盘点。",
"旅游目的地打卡攻略合集。",
"比赛结果预测与竞猜。"
],
"routing_priority": 75
}
]
}

10
docker-compose.full.yml
View File

@ -9,7 +9,7 @@ services:
POSTGRES_PASSWORD: ${POSTGRES_SUPERPASSWORD:?POSTGRES_SUPERPASSWORD is required}
POSTGRES_DB: ${POSTGRES_BOOTSTRAP_DB:-postgres}
volumes:
- ${HOST_DATA_ROOT}/postgres:/var/lib/postgresql/data
- ./data/postgres:/var/lib/postgresql/data
expose:
- "5432"
healthcheck:
@ -32,7 +32,7 @@ services:
TZ: ${TZ:-Asia/Shanghai}
command: ["redis-server", "--appendonly", "yes", "--save", "60", "1000"]
volumes:
- ${HOST_DATA_ROOT}/redis:/data
- ./data/redis:/data
expose:
- "6379"
healthcheck:
@ -74,7 +74,7 @@ services:
DATABASE_POOL_TIMEOUT: ${DATABASE_POOL_TIMEOUT:-30}
DATABASE_POOL_RECYCLE: ${DATABASE_POOL_RECYCLE:-1800}
UPLOAD_MAX_MB: ${UPLOAD_MAX_MB:-100}
DATA_ROOT: ${HOST_DATA_ROOT}
DATA_ROOT: /app/data
BOTS_WORKSPACE_ROOT: ${HOST_BOTS_WORKSPACE_ROOT}
DATABASE_URL: postgresql+psycopg://${POSTGRES_APP_USER}:${POSTGRES_APP_PASSWORD}@postgres:5432/${POSTGRES_APP_DB}
REDIS_ENABLED: ${REDIS_ENABLED:-true}
@ -89,7 +89,7 @@ services:
WORKSPACE_DOWNLOAD_EXTENSIONS: ${WORKSPACE_DOWNLOAD_EXTENSIONS:-}
STT_ENABLED: ${STT_ENABLED:-true}
STT_MODEL: ${STT_MODEL:-ggml-small-q8_0.bin}
STT_MODEL_DIR: ${STT_MODEL_DIR:-${HOST_DATA_ROOT}/model}
STT_MODEL_DIR: ${STT_MODEL_DIR:-/app/data/model}
STT_DEVICE: ${STT_DEVICE:-cpu}
STT_MAX_AUDIO_SECONDS: ${STT_MAX_AUDIO_SECONDS:-20}
STT_DEFAULT_LANGUAGE: ${STT_DEFAULT_LANGUAGE:-zh}
@ -99,7 +99,7 @@ services:
STT_INITIAL_PROMPT: ${STT_INITIAL_PROMPT:-以下内容可能包含简体中文和英文术语。请优先输出简体中文，英文单词、缩写、品牌名和数字保持原文，不要翻译。}
volumes:
- /var/run/docker.sock:/var/run/docker.sock
- ${HOST_DATA_ROOT}:${HOST_DATA_ROOT}
- ./data:/app/data
- ${HOST_BOTS_WORKSPACE_ROOT}:${HOST_BOTS_WORKSPACE_ROOT}
expose:
- "8000"

6
docker-compose.prod.yml
View File

@ -20,7 +20,7 @@ services:
DATABASE_POOL_TIMEOUT: ${DATABASE_POOL_TIMEOUT:-30}
DATABASE_POOL_RECYCLE: ${DATABASE_POOL_RECYCLE:-1800}
UPLOAD_MAX_MB: ${UPLOAD_MAX_MB:-100}
DATA_ROOT: ${HOST_DATA_ROOT}
DATA_ROOT: /app/data
BOTS_WORKSPACE_ROOT: ${HOST_BOTS_WORKSPACE_ROOT}
DATABASE_URL: ${DATABASE_URL:-}
REDIS_ENABLED: ${REDIS_ENABLED:-false}
@ -33,7 +33,7 @@ services:
PANEL_ACCESS_PASSWORD: ${PANEL_ACCESS_PASSWORD:-}
STT_ENABLED: ${STT_ENABLED:-true}
STT_MODEL: ${STT_MODEL:-ggml-small-q8_0.bin}
STT_MODEL_DIR: ${STT_MODEL_DIR:-${HOST_DATA_ROOT}/model}
STT_MODEL_DIR: ${STT_MODEL_DIR:-/app/data/model}
STT_DEVICE: ${STT_DEVICE:-cpu}
STT_MAX_AUDIO_SECONDS: ${STT_MAX_AUDIO_SECONDS:-20}
STT_DEFAULT_LANGUAGE: ${STT_DEFAULT_LANGUAGE:-zh}
@ -43,7 +43,7 @@ services:
STT_INITIAL_PROMPT: ${STT_INITIAL_PROMPT:-以下内容可能包含简体中文和英文术语。请优先输出简体中文，英文单词、缩写、品牌名和数字保持原文，不要翻译。}
volumes:
- /var/run/docker.sock:/var/run/docker.sock
- ${HOST_DATA_ROOT}:${HOST_DATA_ROOT}
- ./data:/app/data
- ${HOST_BOTS_WORKSPACE_ROOT}:${HOST_BOTS_WORKSPACE_ROOT}
expose:
- "8000"

11
scripts/deploy-full.sh
View File

@ -4,6 +4,7 @@ set -euo pipefail
ROOT_DIR="$(cd "$(dirname "$0")/.." && pwd)"
ENV_FILE="${1:-$ROOT_DIR/.env.full}"
COMPOSE_FILE="$ROOT_DIR/docker-compose.full.yml"
DATA_DIR="$ROOT_DIR/data"
require_file() {
local path="$1"
@ -89,7 +90,6 @@ wait_for_health() {
require_file "$ENV_FILE" "Create it from: $ROOT_DIR/.env.full.example"
require_file "$COMPOSE_FILE" ""
load_env_var HOST_DATA_ROOT
load_env_var HOST_BOTS_WORKSPACE_ROOT
load_env_var POSTGRES_SUPERUSER postgres
load_env_var POSTGRES_SUPERPASSWORD
@ -99,7 +99,6 @@ load_env_var POSTGRES_APP_USER
load_env_var POSTGRES_APP_PASSWORD
load_env_var NGINX_PORT 8080
require_env HOST_DATA_ROOT
require_env HOST_BOTS_WORKSPACE_ROOT
require_env POSTGRES_SUPERUSER
require_env POSTGRES_SUPERPASSWORD
@ -111,10 +110,10 @@ require_env NGINX_PORT
echo "[deploy-full] using env: $ENV_FILE"
mkdir -p \
"$HOST_DATA_ROOT" \
"$HOST_DATA_ROOT/postgres" \
"$HOST_DATA_ROOT/redis" \
"$HOST_DATA_ROOT/model" \
"$DATA_DIR" \
"$DATA_DIR/postgres" \
"$DATA_DIR/redis" \
"$DATA_DIR/model" \
"$HOST_BOTS_WORKSPACE_ROOT"
docker compose --env-file "$ENV_FILE" -f "$COMPOSE_FILE" config -q

2
scripts/deploy-prod.sh
View File

@ -3,6 +3,7 @@ set -euo pipefail
ROOT_DIR="$(cd "$(dirname "$0")/.." && pwd)"
ENV_FILE="${1:-$ROOT_DIR/.env.prod}"
DATA_DIR="$ROOT_DIR/data"
if [[ ! -f "$ENV_FILE" ]]; then
echo "Missing env file: $ENV_FILE"
@ -11,6 +12,7 @@ if [[ ! -f "$ENV_FILE" ]]; then
fi
echo "[deploy] using env: $ENV_FILE"
mkdir -p "$DATA_DIR" "$DATA_DIR/model"
docker compose --env-file "$ENV_FILE" -f "$ROOT_DIR/docker-compose.prod.yml" config -q
docker compose --env-file "$ENV_FILE" -f "$ROOT_DIR/docker-compose.prod.yml" up -d --build