c57c930949
数值全部在服务端算好再交给模型，模型只做解读。让模型从 CSV 里自己推 z 分数，它算错的次数足以让简报引用图表反驳它的数字。 - services/insights.py：z 分数（28 天个人基线，且**排除当天**——用一个 值参与算出来的均值去衡量它自己，会把真实离群点摊平）、13 个月趋势斜率 （按序数日期最小二乘，手表放充电器上一周不会压缩 x 轴）、近 7 天活动量 对比。 - services/coach.py：三套提示词 + 回复解析，每套都配一个规则引擎版本。 网关一次生成要几分钟，上游被限流时给一个朴素的答案，好过给一张空卡片。 - services/ai.py：多轮 chat()、SSE stream()、complete()/stream_chat()， 以及 extract_json()——上游是推理模型，可见输出以思维链开头，所以从末尾 倒着找最后一个配平的 JSON（字符串感知，扛得住引号里的 } 和转义引号）。 - 接口 briefing / trend-insight / copilot(SSE)，缓存表 ai_insights。 - 前端：今日页晨报卡（后台生成 + 轮询升级）、全局 Copilot 浮窗、指标详情 页归因面板。features.ai 打开。 实测（对着自建 ai-gateway）：晨报一次 273 秒，缓存命中 18 毫秒——所以简报 绝不能同步阻塞首屏。网关的流式通道比阻塞通道更不可靠：同一条提示词流式 139 秒后返回「所有模型均不可用」，阻塞则成功，因此 stream_chat() 在流式零 输出时对同一模型退回非流式重试。Copilot 实测 TTFB 9ms、全程 40 秒。 顺带修两处：refresh 原来只跳过缓存读、不删行，导致「重新生成」后的轮询读 到旧行、看到 cached 就停了，用户一直盯着他刚要求替换掉的那段字；基线零方差 时原来返回 z=0.0，把「和每一条观测都不同」标成「完全正常」，改为 z=null。 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
92 lines
4.1 KiB
Plaintext
92 lines
4.1 KiB
Plaintext
# --- Server ---
|
|
# BACKEND_PORT takes precedence over PORT. Prefer it: many tools inject PORT
|
|
# for the frontend, and Flask would otherwise take the React dev server's port.
|
|
BACKEND_PORT=5000
|
|
|
|
# --- Database: sqlite (default) or mariadb ---
|
|
DB_TYPE=sqlite
|
|
# SQLite file (used when DB_TYPE=sqlite)
|
|
DATABASE_PATH=./data/health.db
|
|
|
|
# MariaDB (used when DB_TYPE=mariadb) — production DB on the Oracle server
|
|
# (129.146.26.249, local MariaDB 10.3). Dedicated account over TCP 127.0.0.1;
|
|
# MARIADB_SOCKET is only needed if TCP auth is disabled for the app user.
|
|
# MARIADB_HOST=127.0.0.1
|
|
# MARIADB_PORT=3306
|
|
# MARIADB_USER=garmin
|
|
# MARIADB_PASSWORD=your_production_mariadb_password
|
|
# MARIADB_DATABASE=garmin_health_lab
|
|
|
|
# --- Auth ---
|
|
# CHANGE THIS in production! Used to sign JWTs (7-day expiry by default).
|
|
JWT_SECRET=dev_secret_change_me
|
|
JWT_EXPIRY_DAYS=7
|
|
|
|
# --- auth-hub OAuth2 provider (centralized SSO) ---
|
|
# See docs/AUTH_HUB_INTEGRATION.md for setup instructions.
|
|
#
|
|
# Base URL of the auth-hub service
|
|
AUTH_HUB_BASE_URL=http://129.146.26.249:5300
|
|
#
|
|
# OAuth2 client credentials (obtain from auth-hub.manage_clients create)
|
|
# NOTE: these must be registered against the auth-hub instance AUTH_HUB_BASE_URL
|
|
# actually points to (dev vs prod are separate databases with separate clients).
|
|
# Put the REAL values in backend/.env (gitignored) — never here.
|
|
AUTH_HUB_CLIENT_ID=your_client_id
|
|
AUTH_HUB_CLIENT_SECRET=your_client_secret
|
|
#
|
|
# Callback URL (must exactly match what's registered in auth-hub)
|
|
AUTH_HUB_REDIRECT_URI=http://129.146.26.249:8123/auth/callback
|
|
|
|
# --- CORS (comma-separated allowed front-end origins) ---
|
|
# localhost stays in the production list on purpose: CORS is not an auth
|
|
# boundary — every data route requires a valid JWT — so allowing a developer's
|
|
# dev server costs nothing and saves toggling this on every session.
|
|
CORS_ORIGIN=http://localhost:3000,http://localhost:5173
|
|
|
|
# --- AI models (text-only, large context) ---
|
|
# Put REAL keys in backend/.env — that file is gitignored. Never commit keys.
|
|
# Any model whose credentials are absent is skipped automatically.
|
|
|
|
# Self-hosted AI gateway (model id "gateway"). OpenAI-compatible; it fans out
|
|
# over nvidia/gemini/ollama itself and rotates several Gemini keys, so it
|
|
# absorbs single-vendor quota limits. Reached directly, bypassing any local
|
|
# HTTP proxy. NOTE: its NVIDIA upstream is a large reasoning model — replies
|
|
# can take 2-3 minutes, so set AI_TIMEOUT_SECONDS accordingly.
|
|
# HTTPS (Caddy, strips the /ai prefix) rather than http://…:5100 — the token
|
|
# rides in an Authorization header and should not cross the internet in clear.
|
|
AI_GATEWAY_BASE_URL=https://oracle.zichuan.xyz/ai/v1
|
|
AI_GATEWAY_TOKEN=
|
|
AI_GATEWAY_MODEL=ai-gateway-auto
|
|
|
|
# Google AI Studio -> "gemini-flash". Free-tier quota is small; 429s are common.
|
|
GEMINI_API_KEY=
|
|
|
|
# NVIDIA NIM -> "llama-70b", "nemotron-49b", "mistral-large".
|
|
# Model ids come from that account's live GET /v1/models — do not guess them.
|
|
NVIDIA_API_KEY=
|
|
# NVIDIA_BASE_URL=https://integrate.api.nvidia.com/v1
|
|
|
|
# Preference order. The first configured model answers; if it fails or times
|
|
# out, the next is tried. Read per request, so changes need no restart.
|
|
AI_MODEL_CHAIN=gateway,gemini-flash,llama-70b
|
|
|
|
# Max days of history sent (CSV-encoded). Trimmed further per model so the
|
|
# payload always fits that model's own context window.
|
|
AI_DAY_BUDGET=365
|
|
|
|
# Measured against the gateway, not guessed: a trivial prompt took 138s end to
|
|
# end, because its primary upstream emits a full chain of thought before the
|
|
# answer. Nothing user-facing blocks on this (the briefing generates in a
|
|
# background thread), but the timeout still has to clear the real latency.
|
|
AI_TIMEOUT_SECONDS=300
|
|
# Output cap. Reasoning models spend part of it thinking before they answer;
|
|
# entries that need more declare their own budget in services/ai.py.
|
|
AI_MAX_TOKENS=1024
|
|
|
|
# Output cap for the AI coach (晨报 / 趋势归因 / Copilot). Larger than
|
|
# AI_MAX_TOKENS above: the same reasoning trace is spent from this budget
|
|
# before the answer starts, and at 1024 the reply was all thinking with the
|
|
# JSON truncated away.
|
|
AI_COACH_MAX_TOKENS=4000
|