chore: prepare repository for publishing

scripts/apply_migration.sh

@@ -0,0 +1,24 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT_DIR"
+
+if [[ -f ".env.local" ]]; then
+  # shellcheck disable=SC1091
+  source ".env.local"
+fi
+if [[ -f ".env" ]]; then
+  # shellcheck disable=SC1091
+  source ".env"
+fi
+
+if [[ -z "${DATABASE_URL:-}" ]]; then
+  echo "DATABASE_URL 未设置" >&2
+  exit 1
+fi
+
+find "db/migrations" -maxdepth 1 -type f -name "*.sql" | sort | while read -r migration; do
+  psql "$DATABASE_URL" -v ON_ERROR_STOP=1 -f "$migration"
+  echo "migration 已应用: $migration"
+done

scripts/backup.sh

@@ -0,0 +1,28 @@
+#!/bin/bash
+# backup.sh - 数据库备份 + OSS上传
+set -euo pipefail
+
+DB_URL="${DATABASE_URL:-host=/var/run/postgresql dbname=llm_intelligence user=long sslmode=disable}"
+BACKUP_DIR="/tmp/llm_hub_backups"
+DATE=$(date +%Y%m%d_%H%M%S)
+BACKUP_FILE="${BACKUP_DIR}/llm_intelligence_${DATE}.sql"
+
+mkdir -p "$BACKUP_DIR"
+
+# 1. pg_dump 备份
+echo "[$(date)] 开始备份..."
+pg_dump "$DB_URL" > "$BACKUP_FILE"
+gzip "$BACKUP_FILE"
+
+# 2. 保留最近7天
+echo "[$(date)] 清理过期备份..."
+find "$BACKUP_DIR" -name "*.sql.gz" -mtime +7 -delete
+
+# 3. 如果有 OSS 工具则上传
+if command -v ossutil &> /dev/null; then
+    echo "[$(date)] 上传至 OSS..."
+    ossutil cp "${BACKUP_FILE}.gz" "oss://your-bucket/llm-hub-backups/" || true
+fi
+
+echo "[$(date)] 备份完成: ${BACKUP_FILE}.gz"
+ls -lh "${BACKUP_FILE}.gz"

scripts/feishu_alert.sh

@@ -0,0 +1,30 @@
+#!/bin/bash
+# feishu_alert.sh - 飞书告警脚本
+# 用法: ./feishu_alert.sh "告警消息"
+
+WEBHOOK_URL="${FEISHU_WEBHOOK:-}"
+if [ -z "$WEBHOOK_URL" ]; then
+    echo "错误: FEISHU_WEBHOOK 环境变量未设置" >&2
+    exit 1
+fi
+
+MESSAGE="${1:-LLM Hub 告警}"
+DATE=$(date '+%Y-%m-%d %H:%M:%S')
+
+# 构造飞书文本消息
+PAYLOAD=$(cat <<EOF
+{
+  "msg_type": "text",
+  "content": {
+    "text": "🚨 LLM Intelligence Hub 告警\\n时间: ${DATE}\\n消息: ${MESSAGE}\\n请检查系统状态。"
+  }
+}
+EOF
+)
+
+curl -s -X POST \
+    -H "Content-Type: application/json" \
+    -d "$PAYLOAD" \
+    "$WEBHOOK_URL"
+
+echo "告警已发送"

scripts/fetch_multi_source.go

@@ -0,0 +1,631 @@
+//go:build llm_script
+
+// fetch_multi_source.go - 多源 LLM 定价采集器
+// 支持: OpenRouter, Moonshot, DeepSeek, OpenAI 等
+package main
+
+import (
+	"database/sql"
+	"encoding/json"
+	"flag"
+	"fmt"
+	"io"
+	"log/slog"
+	"net/http"
+	"os"
+	"sort"
+	"strings"
+	"time"
+
+	_ "github.com/lib/pq"
+)
+
+var logger *slog.Logger
+
+func init() {
+	logger = slog.New(slog.NewJSONHandler(os.Stderr, &slog.HandlerOptions{Level: slog.LevelInfo}))
+}
+
+// ============ 统一数据接口 ============
+
+type ModelPricing struct {
+	ModelID         string
+	ModelName       string
+	ProviderName    string
+	ProviderCountry string
+	OperatorName    string
+	OperatorType    string // official / reseller / cloud
+	Region          string
+	Currency        string
+	InputPrice      float64 // per 1M tokens
+	OutputPrice     float64 // per 1M tokens
+	ContextLength   int
+	IsFree          bool
+	SourceURL       string
+	Modality        string
+	SceneTags       []string
+}
+
+// DataSource 统一采集接口
+type DataSource interface {
+	Name() string
+	FetchPricing() ([]ModelPricing, error)
+	SourceType() string // official / reseller
+}
+
+type runConfig struct {
+	DryRun bool
+}
+
+type sourceDefinition struct {
+	Key     string
+	Name    string
+	Factory func() DataSource
+}
+
+type runSummary struct {
+	SelectedSources   int
+	SuccessfulSources int
+	TotalModels       int
+	DomesticModels    int
+	CurrencyCounts    map[string]int
+}
+
+type pricingMetadataFields struct {
+	SourceType      string
+	FreeQuota       string
+	FreeLimitations string
+	RateLimit       string
+}
+
+// ============ OpenRouter 采集器 ============
+
+type OpenRouterSource struct {
+	APIKey string
+}
+
+func (s *OpenRouterSource) Name() string       { return "OpenRouter" }
+func (s *OpenRouterSource) SourceType() string { return "reseller" }
+
+func (s *OpenRouterSource) FetchPricing() ([]ModelPricing, error) {
+	url := "https://openrouter.ai/api/v1/models"
+	req, err := http.NewRequest("GET", url, nil)
+	if err != nil {
+		return nil, err
+	}
+	if s.APIKey != "" {
+		req.Header.Set("Authorization", "Bearer "+s.APIKey)
+	}
+
+	client := &http.Client{Timeout: 30 * time.Second}
+	resp, err := client.Do(req)
+	if err != nil {
+		return nil, err
+	}
+	defer resp.Body.Close()
+
+	body, _ := io.ReadAll(resp.Body)
+
+	var result struct {
+		Data []struct {
+			ID      string `json:"id"`
+			Name    string `json:"name"`
+			Pricing struct {
+				Prompt     string `json:"prompt"`
+				Completion string `json:"completion"`
+			} `json:"pricing"`
+			ContextLength int `json:"context_length"`
+		} `json:"data"`
+	}
+
+	if err := json.Unmarshal(body, &result); err != nil {
+		return nil, fmt.Errorf("parse error: %w", err)
+	}
+
+	var prices []ModelPricing
+	for _, m := range result.Data {
+		inputPrice := parsePrice(m.Pricing.Prompt)
+		outputPrice := parsePrice(m.Pricing.Completion)
+
+		prices = append(prices, ModelPricing{
+			ModelID:         m.ID,
+			ModelName:       m.Name,
+			ProviderName:    extractProvider(m.ID),
+			ProviderCountry: "US",
+			OperatorName:    "OpenRouter",
+			OperatorType:    "reseller",
+			Region:          "global",
+			Currency:        "USD",
+			InputPrice:      inputPrice * 1000000,
+			OutputPrice:     outputPrice * 1000000,
+			ContextLength:   m.ContextLength,
+			IsFree:          inputPrice == 0 && outputPrice == 0,
+			SourceURL:       "https://openrouter.ai/docs#models",
+			Modality:        "text",
+		})
+	}
+
+	logger.Info("OpenRouter采集完成", "models", len(prices))
+	return prices, nil
+}
+
+func parsePrice(s string) float64 {
+	var f float64
+	fmt.Sscanf(s, "%f", &f)
+	if f < 0 {
+		return 0
+	}
+	return f
+}
+
+func extractProvider(modelID string) string {
+	parts := strings.Split(modelID, "/")
+	if len(parts) > 1 {
+		return parts[0]
+	}
+	return "unknown"
+}
+
+func sourceDefinitions(apiKey string) []sourceDefinition {
+	return []sourceDefinition{
+		{
+			Key:  "openrouter",
+			Name: "OpenRouter",
+			Factory: func() DataSource {
+				return &OpenRouterSource{APIKey: apiKey}
+			},
+		},
+		{
+			Key:  "moonshot",
+			Name: "Moonshot",
+			Factory: func() DataSource {
+				return &MoonshotSource{}
+			},
+		},
+		{
+			Key:  "deepseek",
+			Name: "DeepSeek",
+			Factory: func() DataSource {
+				return &DeepSeekSource{}
+			},
+		},
+		{
+			Key:  "openai",
+			Name: "OpenAI",
+			Factory: func() DataSource {
+				return &OpenAISource{}
+			},
+		},
+	}
+}
+
+func parseSourceList(raw string) []string {
+	if strings.TrimSpace(raw) == "" {
+		return nil
+	}
+
+	parts := strings.Split(raw, ",")
+	sources := make([]string, 0, len(parts))
+	seen := make(map[string]struct{}, len(parts))
+	for _, part := range parts {
+		name := strings.ToLower(strings.TrimSpace(part))
+		if name == "" {
+			continue
+		}
+		if _, ok := seen[name]; ok {
+			continue
+		}
+		seen[name] = struct{}{}
+		sources = append(sources, name)
+	}
+	return sources
+}
+
+func buildSources(apiKey string, requested []string) ([]DataSource, error) {
+	definitions := sourceDefinitions(apiKey)
+	if len(requested) == 0 {
+		sources := make([]DataSource, 0, len(definitions))
+		for _, definition := range definitions {
+			sources = append(sources, definition.Factory())
+		}
+		return sources, nil
+	}
+
+	definitionByKey := make(map[string]sourceDefinition, len(definitions))
+	for _, definition := range definitions {
+		definitionByKey[definition.Key] = definition
+	}
+
+	sources := make([]DataSource, 0, len(requested))
+	for _, name := range requested {
+		definition, ok := definitionByKey[name]
+		if !ok {
+			return nil, fmt.Errorf("unknown source %q", name)
+		}
+		sources = append(sources, definition.Factory())
+	}
+	return sources, nil
+}
+
+func listSourceKeys(apiKey string) []string {
+	definitions := sourceDefinitions(apiKey)
+	keys := make([]string, 0, len(definitions))
+	for _, definition := range definitions {
+		keys = append(keys, definition.Key)
+	}
+	return keys
+}
+
+func summarizePrices(selectedSources int, successfulSources int, prices []ModelPricing) runSummary {
+	summary := runSummary{
+		SelectedSources:   selectedSources,
+		SuccessfulSources: successfulSources,
+		TotalModels:       len(prices),
+		CurrencyCounts:    make(map[string]int),
+	}
+	for _, price := range prices {
+		if strings.EqualFold(price.ProviderCountry, "CN") {
+			summary.DomesticModels++
+		}
+		summary.CurrencyCounts[strings.ToUpper(price.Currency)]++
+	}
+	return summary
+}
+
+func formatCountMap(counts map[string]int) string {
+	if len(counts) == 0 {
+		return "none"
+	}
+	keys := make([]string, 0, len(counts))
+	for key := range counts {
+		keys = append(keys, key)
+	}
+	sort.Strings(keys)
+
+	parts := make([]string, 0, len(keys))
+	for _, key := range keys {
+		parts = append(parts, fmt.Sprintf("%s:%d", key, counts[key]))
+	}
+	return strings.Join(parts, ",")
+}
+
+func printSummary(w io.Writer, summary runSummary) error {
+	if w == nil {
+		return nil
+	}
+	_, err := fmt.Fprintf(
+		w,
+		"sources=%d successful_sources=%d models=%d domestic_models=%d currencies=%s\n",
+		summary.SelectedSources,
+		summary.SuccessfulSources,
+		summary.TotalModels,
+		summary.DomesticModels,
+		formatCountMap(summary.CurrencyCounts),
+	)
+	return err
+}
+
+func pricingMetadata(p ModelPricing) pricingMetadataFields {
+	sourceType := strings.TrimSpace(strings.ToLower(p.OperatorType))
+	if sourceType == "" {
+		sourceType = "official"
+	}
+
+	fields := pricingMetadataFields{
+		SourceType:      sourceType,
+		FreeLimitations: "[]",
+		RateLimit:       "{}",
+	}
+	if p.IsFree {
+		fields.SourceType = "free_tier"
+		fields.FreeQuota = "See source_url for provider free-tier details"
+		fields.FreeLimitations = `["See source_url for current quota and policy"]`
+	}
+	return fields
+}
+
+// ============ Moonshot 采集器 ============
+
+type MoonshotSource struct{}
+
+func (s *MoonshotSource) Name() string       { return "Moonshot" }
+func (s *MoonshotSource) SourceType() string { return "official" }
+
+func (s *MoonshotSource) FetchPricing() ([]ModelPricing, error) {
+	prices := []ModelPricing{
+		{
+			ModelID: "kimi-k2.6", ModelName: "Kimi K2.6",
+			ProviderName: "Moonshot AI", ProviderCountry: "CN",
+			OperatorName: "Moonshot", OperatorType: "official",
+			Region: "CN", Currency: "CNY",
+			InputPrice: 6.50, OutputPrice: 27.00,
+			ContextLength: 262144, IsFree: false,
+			SourceURL: "https://platform.kimi.com/docs/pricing/chat-k26",
+			Modality:  "multimodal",
+			SceneTags: []string{"对话", "视觉", "代码"},
+		},
+		{
+			ModelID: "kimi-k2-0905-preview", ModelName: "Kimi K2 0905 Preview",
+			ProviderName: "Moonshot AI", ProviderCountry: "CN",
+			OperatorName: "Moonshot", OperatorType: "official",
+			Region: "CN", Currency: "CNY",
+			InputPrice: 4.00, OutputPrice: 16.00,
+			ContextLength: 262144, IsFree: false,
+			SourceURL: "https://platform.kimi.com/docs/pricing/chat-k2",
+			Modality:  "text",
+			SceneTags: []string{"代码", "对话"},
+		},
+		{
+			ModelID: "moonshot-v1-8k", ModelName: "Moonshot V1 8K",
+			ProviderName: "Moonshot AI", ProviderCountry: "CN",
+			OperatorName: "Moonshot", OperatorType: "official",
+			Region: "CN", Currency: "CNY",
+			InputPrice: 2.00, OutputPrice: 10.00,
+			ContextLength: 8192, IsFree: false,
+			SourceURL: "https://platform.kimi.com/docs/pricing/chat-v1",
+			Modality:  "text",
+			SceneTags: []string{"对话"},
+		},
+	}
+
+	logger.Info("Moonshot采集完成", "models", len(prices))
+	return prices, nil
+}
+
+// ============ DeepSeek 采集器 ============
+
+type DeepSeekSource struct{}
+
+func (s *DeepSeekSource) Name() string       { return "DeepSeek" }
+func (s *DeepSeekSource) SourceType() string { return "official" }
+
+func (s *DeepSeekSource) FetchPricing() ([]ModelPricing, error) {
+	prices := []ModelPricing{
+		{
+			ModelID: "deepseek-v4-flash", ModelName: "DeepSeek V4 Flash",
+			ProviderName: "DeepSeek", ProviderCountry: "CN",
+			OperatorName: "DeepSeek", OperatorType: "official",
+			Region: "global", Currency: "USD",
+			InputPrice: 0.14, OutputPrice: 0.28,
+			ContextLength: 1000000, IsFree: false,
+			SourceURL: "https://api-docs.deepseek.com/quick_start/pricing",
+			Modality:  "text",
+			SceneTags: []string{"对话", "推理"},
+		},
+		{
+			ModelID: "deepseek-v4-pro", ModelName: "DeepSeek V4 Pro",
+			ProviderName: "DeepSeek", ProviderCountry: "CN",
+			OperatorName: "DeepSeek", OperatorType: "official",
+			Region: "global", Currency: "USD",
+			InputPrice: 0.435, OutputPrice: 0.87,
+			ContextLength: 1000000, IsFree: false,
+			SourceURL: "https://api-docs.deepseek.com/quick_start/pricing",
+			Modality:  "code",
+			SceneTags: []string{"对话", "推理", "代码"},
+		},
+	}
+
+	logger.Info("DeepSeek采集完成", "models", len(prices))
+	return prices, nil
+}
+
+// ============ OpenAI 采集器 ============
+
+type OpenAISource struct{}
+
+func (s *OpenAISource) Name() string       { return "OpenAI" }
+func (s *OpenAISource) SourceType() string { return "official" }
+
+func (s *OpenAISource) FetchPricing() ([]ModelPricing, error) {
+	prices := []ModelPricing{
+		{
+			ModelID: "gpt-5.5", ModelName: "GPT-5.5",
+			ProviderName: "OpenAI", ProviderCountry: "US",
+			OperatorName: "OpenAI", OperatorType: "official",
+			Region: "global", Currency: "USD",
+			InputPrice: 5.00, OutputPrice: 30.00,
+			ContextLength: 200000, IsFree: false,
+			SourceURL: "https://openai.com/api/pricing/",
+			Modality:  "code",
+			SceneTags: []string{"代码", "推理", "对话"},
+		},
+		{
+			ModelID: "gpt-5.4", ModelName: "GPT-5.4",
+			ProviderName: "OpenAI", ProviderCountry: "US",
+			OperatorName: "OpenAI", OperatorType: "official",
+			Region: "global", Currency: "USD",
+			InputPrice: 2.50, OutputPrice: 15.00,
+			ContextLength: 200000, IsFree: false,
+			SourceURL: "https://openai.com/api/pricing/",
+			Modality:  "text",
+			SceneTags: []string{"代码", "对话"},
+		},
+		{
+			ModelID: "gpt-5.4-mini", ModelName: "GPT-5.4 Mini",
+			ProviderName: "OpenAI", ProviderCountry: "US",
+			OperatorName: "OpenAI", OperatorType: "official",
+			Region: "global", Currency: "USD",
+			InputPrice: 0.75, OutputPrice: 4.50,
+			ContextLength: 200000, IsFree: false,
+			SourceURL: "https://openai.com/api/pricing/",
+			Modality:  "text",
+			SceneTags: []string{"对话"},
+		},
+	}
+
+	logger.Info("OpenAI采集完成", "models", len(prices))
+	return prices, nil
+}
+
+// ============ 数据库写入 ============
+
+func saveToDatabase(db *sql.DB, prices []ModelPricing, batchID string) error {
+	for _, p := range prices {
+		// 查找或创建 provider
+		var providerID int64
+		err := db.QueryRow(
+			"SELECT id FROM model_provider WHERE name = $1",
+			p.ProviderName,
+		).Scan(&providerID)
+
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				"INSERT INTO model_provider (name, country, website, status) VALUES ($1, $2, $3, 'active') RETURNING id",
+				p.ProviderName, p.ProviderCountry, "",
+			).Scan(&providerID)
+		}
+		if err != nil {
+			logger.Warn("provider error", "name", p.ProviderName, "error", err)
+			continue
+		}
+
+		// 查找或创建 operator
+		var operatorID int64
+		err = db.QueryRow(
+			"SELECT id FROM operator WHERE name = $1",
+			p.OperatorName,
+		).Scan(&operatorID)
+
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				"INSERT INTO operator (name, country, status) VALUES ($1, $2, 'active') RETURNING id",
+				p.OperatorName, p.ProviderCountry,
+			).Scan(&operatorID)
+		}
+		if err != nil {
+			logger.Warn("operator error", "name", p.OperatorName, "error", err)
+			continue
+		}
+
+		// 查找或创建 model (使用 external_id)
+		var modelID int64
+		err = db.QueryRow(
+			"SELECT id FROM models WHERE external_id = $1",
+			p.ModelID,
+		).Scan(&modelID)
+
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				`INSERT INTO models (external_id, name, provider_id, modality, context_length, status, source, batch_id) 
+				 VALUES ($1, $2, $3, $4, $5, 'active', $6, $7) RETURNING id`,
+				p.ModelID, p.ModelName, providerID, p.Modality, p.ContextLength, p.OperatorName, batchID,
+			).Scan(&modelID)
+		}
+		if err != nil {
+			logger.Warn("model error", "id", p.ModelID, "error", err)
+			continue
+		}
+
+		// 插入定价
+		metadata := pricingMetadata(p)
+		_, err = db.Exec(
+			`INSERT INTO region_pricing 
+			 (model_id, operator_id, region, currency, input_price_per_mtok, output_price_per_mtok, is_free, effective_date, source_url, source_type, free_quota, free_limitations, rate_limit)
+			 VALUES ($1, $2, $3, $4, $5, $6, $7, CURRENT_DATE, $8, $9, $10, $11, $12)
+			 ON CONFLICT (model_id, operator_id, region, currency, effective_date) 
+			 DO UPDATE SET input_price_per_mtok = EXCLUDED.input_price_per_mtok,
+			               output_price_per_mtok = EXCLUDED.output_price_per_mtok,
+			               is_free = EXCLUDED.is_free,
+			               source_type = EXCLUDED.source_type,
+			               free_quota = EXCLUDED.free_quota,
+			               free_limitations = EXCLUDED.free_limitations,
+			               rate_limit = EXCLUDED.rate_limit,
+			               updated_at = CURRENT_TIMESTAMP`,
+			modelID, operatorID, p.Region, p.Currency, p.InputPrice, p.OutputPrice, p.IsFree, p.SourceURL,
+			metadata.SourceType, metadata.FreeQuota, metadata.FreeLimitations, metadata.RateLimit,
+		)
+		if err != nil {
+			logger.Warn("pricing insert error", "model", p.ModelID, "error", err)
+			continue
+		}
+	}
+
+	return nil
+}
+
+func savePricesToDatabase(dsn string, prices []ModelPricing) error {
+	db, err := sql.Open("postgres", dsn)
+	if err != nil {
+		return err
+	}
+	defer db.Close()
+
+	batchID := fmt.Sprintf("phase2-%s", time.Now().Format("20060102-150405"))
+	return saveToDatabase(db, prices, batchID)
+}
+
+func defaultDSN() string {
+	dsn := os.Getenv("DATABASE_URL")
+	if dsn != "" {
+		return dsn
+	}
+	return "postgres://long@/llm_intelligence?host=/var/run/postgresql"
+}
+
+func runCollector(cfg runConfig, sources []DataSource, saveFn func([]ModelPricing) error, out io.Writer) error {
+	allPrices := make([]ModelPricing, 0)
+	successfulSources := 0
+
+	for _, src := range sources {
+		prices, err := src.FetchPricing()
+		if err != nil {
+			logger.Error("采集失败", "source", src.Name(), "error", err)
+			continue
+		}
+		successfulSources++
+		allPrices = append(allPrices, prices...)
+	}
+
+	summary := summarizePrices(len(sources), successfulSources, allPrices)
+	if err := printSummary(out, summary); err != nil {
+		return err
+	}
+	if successfulSources == 0 {
+		return fmt.Errorf("no data source collected successfully")
+	}
+	if cfg.DryRun {
+		return nil
+	}
+	if saveFn == nil {
+		return fmt.Errorf("save function is required when dry-run is disabled")
+	}
+	if err := saveFn(allPrices); err != nil {
+		return err
+	}
+
+	logger.Info("多源采集完成", "total_models", len(allPrices), "sources", successfulSources)
+	return nil
+}
+
+// ============ 主程序 ============
+
+func main() {
+	var sourcesFlag string
+	var dryRun bool
+	var listSources bool
+
+	flag.StringVar(&sourcesFlag, "sources", "", "comma-separated source keys: openrouter,moonshot,deepseek,openai")
+	flag.BoolVar(&dryRun, "dry-run", false, "collect and print summary without writing to database")
+	flag.BoolVar(&listSources, "list-sources", false, "print available source keys and exit")
+	flag.Parse()
+
+	apiKey := os.Getenv("OPENROUTER_API_KEY")
+	if listSources {
+		fmt.Println(strings.Join(listSourceKeys(apiKey), ","))
+		return
+	}
+
+	sources, err := buildSources(apiKey, parseSourceList(sourcesFlag))
+	if err != nil {
+		logger.Error("数据源参数非法", "error", err)
+		os.Exit(1)
+	}
+
+	cfg := runConfig{DryRun: dryRun}
+	if err := runCollector(cfg, sources, func(prices []ModelPricing) error {
+		return savePricesToDatabase(defaultDSN(), prices)
+	}, os.Stdout); err != nil {
+		logger.Error("多源采集失败", "error", err)
+		os.Exit(1)
+	}
+}

scripts/fetch_multi_source_test.go

@@ -0,0 +1,108 @@
+//go:build llm_script
+
+package main
+
+import (
+	"bytes"
+	"testing"
+)
+
+type fakeSource struct {
+	name   string
+	prices []ModelPricing
+	err    error
+}
+
+func (s fakeSource) Name() string { return s.name }
+
+func (s fakeSource) FetchPricing() ([]ModelPricing, error) { return s.prices, s.err }
+
+func (s fakeSource) SourceType() string { return "official" }
+
+func TestBuildSourcesFiltersRequestedNames(t *testing.T) {
+	sources, err := buildSources("", []string{"moonshot", "openai"})
+	if err != nil {
+		t.Fatalf("buildSources returned error: %v", err)
+	}
+
+	if len(sources) != 2 {
+		t.Fatalf("expected 2 sources, got %d", len(sources))
+	}
+
+	if sources[0].Name() != "Moonshot" || sources[1].Name() != "OpenAI" {
+		t.Fatalf("unexpected source order: %s, %s", sources[0].Name(), sources[1].Name())
+	}
+}
+
+func TestBuildSourcesRejectsUnknownNames(t *testing.T) {
+	_, err := buildSources("", []string{"moonshot", "unknown"})
+	if err == nil {
+		t.Fatal("expected error for unknown source")
+	}
+}
+
+func TestRunCollectorDryRunSkipsDatabaseWrite(t *testing.T) {
+	cfg := runConfig{DryRun: true}
+	var out bytes.Buffer
+	writeCalled := false
+
+	err := runCollector(
+		cfg,
+		[]DataSource{
+			fakeSource{
+				name: "Moonshot",
+				prices: []ModelPricing{
+					{ModelID: "kimi-k2.6", ProviderCountry: "CN", Currency: "CNY"},
+					{ModelID: "kimi-k2-0905-preview", ProviderCountry: "CN", Currency: "CNY"},
+				},
+			},
+			fakeSource{
+				name: "OpenAI",
+				prices: []ModelPricing{
+					{ModelID: "gpt-5.5", ProviderCountry: "US", Currency: "USD"},
+				},
+			},
+		},
+		func([]ModelPricing) error {
+			writeCalled = true
+			return nil
+		},
+		&out,
+	)
+	if err != nil {
+		t.Fatalf("runCollector returned error: %v", err)
+	}
+
+	if writeCalled {
+		t.Fatal("expected dry-run to skip database write")
+	}
+
+	output := out.String()
+	if output == "" {
+		t.Fatal("expected dry-run summary output")
+	}
+	if !bytes.Contains(out.Bytes(), []byte("sources=2")) {
+		t.Fatalf("expected sources summary, got %q", output)
+	}
+	if !bytes.Contains(out.Bytes(), []byte("models=3")) {
+		t.Fatalf("expected model summary, got %q", output)
+	}
+	if !bytes.Contains(out.Bytes(), []byte("currencies=CNY:2,USD:1")) {
+		t.Fatalf("expected currency summary, got %q", output)
+	}
+}
+
+func TestPricingMetadataClassifiesSourceType(t *testing.T) {
+	freeTier := pricingMetadata(ModelPricing{OperatorType: "official", IsFree: true})
+	if freeTier.SourceType != "free_tier" {
+		t.Fatalf("expected free_tier, got %q", freeTier.SourceType)
+	}
+	if freeTier.FreeQuota == "" {
+		t.Fatal("expected free tier quota description")
+	}
+
+	reseller := pricingMetadata(ModelPricing{OperatorType: "reseller"})
+	if reseller.SourceType != "reseller" {
+		t.Fatalf("expected reseller, got %q", reseller.SourceType)
+	}
+}

scripts/fetch_openrouter.go

@@ -1,17 +1,26 @@
-// fetch_openrouter.go - OpenRouter 模型数据采集器
-// Phase 1 单数据源采集器，抓取模型基础信息与价格信息
+//go:build llm_script
+
+// fetch_openrouter.go - OpenRouter 模型数据采集器 v2.0
+// Sprint 2 增强版：指数退避重试 + 批量插入 + ProviderMapper + audit_log + 价格变动检测 + slog
 package main
 
 import (
+	"bufio"
+	"context"
 	"database/sql"
 	"encoding/json"
 	"flag"
 	"fmt"
 	"io"
+	"log/slog"
 	"net/http"
 	"os"
+	"strings"
 	"time"
 
+	"llm-intelligence/internal/collectors"
+	"llm-intelligence/internal/retry"
+
 	_ "github.com/lib/pq"
 )
 
@@ -22,23 +31,19 @@ type Config struct {
 	OutPath    string
 	MaxRetries int
 	TimeoutSec int
-	// PostgreSQL 连接参数（新增）
-	DBConn string // e.g. "host=/var/run/postgresql dbname=llm_intelligence sslmode=disable"
-}
-
-// OpenRouter API 响应结构（仅关键字段）
-type APIResponse struct {
-	Data []ModelInfo `json:"data"`
+	BatchSize  int
+	DBConn     string
 }
 
+// ModelInfo 模型信息（与 collectors 包兼容）
 type ModelInfo struct {
-	ID            string         `json:"id"`
-	Name          string         `json:"name,omitempty"`
-	Created       int64          `json:"created,omitempty"`
-	Description   string         `json:"description,omitempty"`
-	ContextLength int            `json:"context_length,omitempty"`
-	Capabilities  []string      `json:"capabilities,omitempty"`
-	Pricing       ModelPricing   `json:"pricing,omitempty"`
+	ID            string       `json:"id"`
+	Name          string       `json:"name,omitempty"`
+	Created       int64        `json:"created,omitempty"`
+	Description   string       `json:"description,omitempty"`
+	ContextLength int          `json:"context_length,omitempty"`
+	Capabilities  []string     `json:"capabilities,omitempty"`
+	Pricing       ModelPricing `json:"pricing,omitempty"`
 }
 
 type ModelPricing struct {
@@ -46,21 +51,54 @@ type ModelPricing struct {
 	Output float64 `json:"output,omitempty"`
 }
 
+var (
+	collectorVersion = "v2.0"
+	logger           *slog.Logger
+)
+
+func init() {
+	logger = slog.New(slog.NewJSONHandler(os.Stderr, &slog.HandlerOptions{
+		Level: slog.LevelInfo,
+	}))
+}
+
 func main() {
 	cfg := parseArgs()
+	start := time.Now()
+
+	logger.Info("采集器启动", "collector", "openrouter", "version", collectorVersion, "batch_size", cfg.BatchSize)
+
+	var runErr error
 	if err := run(cfg); err != nil {
-		fmt.Fprintf(os.Stderr, "采集失败: %v\n", err)
+		logger.Error("采集失败", "error", err, "duration", time.Since(start))
+		runErr = err
+	}
+
+	duration := time.Since(start)
+
+	// 写入采集统计
+	if cfg.DBConn != "" {
+		if err := recordCollectorStats(cfg.DBConn, runErr, duration); err != nil {
+			logger.Warn("采集统计写入失败", "error", err)
+		}
+	}
+
+	if runErr != nil {
 		os.Exit(1)
 	}
+
+	logger.Info("采集完成", "collector", "openrouter", "duration_ms", duration.Milliseconds())
 }
 
 func parseArgs() Config {
-	apiKey := flag.String("api-key", "", "OpenRouter API Key（建议通过环境变量注入）")
+	loadProjectEnv()
+	apiKey := flag.String("api-key", "", "OpenRouter API Key")
 	apiURL := flag.String("api-url", "https://openrouter.ai/api/v1/models", "API 地址")
 	outPath := flag.String("out", "models.json", "输出文件路径")
 	maxRetries := flag.Int("retry", 3, "最大重试次数")
 	timeoutSec := flag.Int("timeout", 30, "请求超时（秒）")
-	dbConn := flag.String("db", os.Getenv("DATABASE_URL"), "PostgreSQL 连接字符串（默认从 DATABASE_URL 环境变量读取）")
+	batchSize := flag.Int("batch", 100, "批量插入批次大小")
+	dbConn := flag.String("db", os.Getenv("DATABASE_URL"), "PostgreSQL 连接字符串")
 	flag.Parse()
 	return Config{
 		APIKey:     *apiKey,
@@ -68,82 +106,131 @@ func parseArgs() Config {
 		OutPath:    *outPath,
 		MaxRetries: *maxRetries,
 		TimeoutSec: *timeoutSec,
+		BatchSize:  *batchSize,
 		DBConn:     *dbConn,
 	}
 }
 
+func loadProjectEnv() {
+	for _, path := range []string{".env.local", ".env"} {
+		loadEnvFile(path)
+	}
+}
+
+func loadEnvFile(path string) {
+	f, err := os.Open(path)
+	if err != nil {
+		return
+	}
+	defer f.Close()
+
+	scanner := bufio.NewScanner(f)
+	for scanner.Scan() {
+		line := strings.TrimSpace(scanner.Text())
+		if line == "" || strings.HasPrefix(line, "#") {
+			continue
+		}
+		key, value, ok := strings.Cut(line, "=")
+		if !ok {
+			continue
+		}
+		key = strings.TrimSpace(key)
+		value = strings.TrimSpace(value)
+		value = strings.Trim(value, `"'`)
+		if key == "" {
+			continue
+		}
+		if _, exists := os.LookupEnv(key); exists {
+			continue
+		}
+		_ = os.Setenv(key, value)
+	}
+}
+
 func run(cfg Config) error {
 	models, err := fetchModels(cfg)
 	if err != nil {
 		return err
 	}
-	// 优先写入 PostgreSQL；若配置了 DBConn 则入库
+
+	logger.Info("API 数据获取完成", "records", len(models))
+
 	if cfg.DBConn != "" {
-		if err := summarizeDB(cfg.DBConn, models); err != nil {
-			fmt.Fprintf(os.Stderr, "警告: PostgreSQL 写入失败: %v\n", err)
-			fmt.Fprintln(os.Stderr, "降级为仅写入 JSON")
+		if err := summarizeDB(cfg.DBConn, models, cfg.BatchSize); err != nil {
+			logger.Error("PostgreSQL 写入失败", "error", err)
+			logger.Warn("降级为仅写入 JSON")
+		} else {
+			logger.Info("PostgreSQL 写入完成", "records", len(models))
 		}
 	}
 	return summarize(cfg.OutPath, models)
 }
 
-// fetchModels 抓取 OpenRouter 模型列表
+// fetchModels 抓取 OpenRouter 模型列表（集成指数退避重试）
 func fetchModels(cfg Config) ([]ModelInfo, error) {
-	// 无 API Key 时返回模拟数据（写入由后续 summarize 统一处理）
 	if cfg.APIKey == "" {
-		fmt.Println("警告: 未提供 API Key，使用模拟数据")
+		logger.Warn("未提供 API Key，使用模拟数据")
 		return []ModelInfo{
-			{ID: "openai/gpt-4o", ContextLength: 128000,
-				Pricing: ModelPricing{Input: 2.5, Output: 10.0}},
-			{ID: "anthropic/claude-3.5-sonnet:free", ContextLength: 200000,
-				Pricing: ModelPricing{}},
+			{ID: "openai/gpt-4o", ContextLength: 128000, Pricing: ModelPricing{Input: 2.5, Output: 10.0}},
+			{ID: "anthropic/claude-3.5-sonnet:free", ContextLength: 200000, Pricing: ModelPricing{}},
 		}, nil
 	}
 
-	client := &http.Client{Timeout: time.Duration(cfg.TimeoutSec) * time.Second}
-	req, err := http.NewRequest("GET", cfg.APIURL, nil)
-	if err != nil {
-		return nil, fmt.Errorf("构造请求失败: %w", err)
+	strategy := retry.Strategy{
+		MaxRetries: cfg.MaxRetries,
+		BaseDelay:  1 * time.Second,
+		MaxDelay:   30 * time.Second,
+		Multiplier: 2.0,
+		Jitter:     true,
+		Retryable:  retry.IsRetryable,
 	}
-	req.Header.Set("Authorization", "Bearer "+cfg.APIKey)
-	req.Header.Set("Content-Type", "application/json")
 
-	var resp *http.Response
-	for i := 0; i <= cfg.MaxRetries; i++ {
-		resp, err = client.Do(req)
-		if err == nil {
-			break
+	var models []ModelInfo
+	var lastErr error
+
+	err := retry.Do(context.Background(), strategy, func() error {
+		client := &http.Client{Timeout: time.Duration(cfg.TimeoutSec) * time.Second}
+		req, err := http.NewRequest("GET", cfg.APIURL, nil)
+		if err != nil {
+			return fmt.Errorf("构造请求失败: %w", err)
 		}
-		if i < cfg.MaxRetries {
-			time.Sleep(time.Duration(i+1) * time.Second)
+		req.Header.Set("Authorization", "Bearer "+cfg.APIKey)
+		req.Header.Set("Content-Type", "application/json")
+
+		resp, err := client.Do(req)
+		if err != nil {
+			lastErr = err
+			return fmt.Errorf("请求失败: %w", err)
 		}
-	}
+		defer resp.Body.Close()
+
+		if resp.StatusCode != http.StatusOK {
+			body, _ := io.ReadAll(resp.Body)
+			lastErr = fmt.Errorf("非 200 响应: %d %s", resp.StatusCode, string(body))
+			return lastErr
+		}
+
+		body, err := io.ReadAll(resp.Body)
+		if err != nil {
+			lastErr = err
+			return fmt.Errorf("读取响应失败: %w", err)
+		}
+
+		models, err = parseModels(body)
+		if err != nil {
+			lastErr = err
+			return fmt.Errorf("JSON 解析失败: %w", err)
+		}
+		return nil
+	})
+
 	if err != nil {
-		return nil, fmt.Errorf("请求失败: %w", err)
-	}
-	defer resp.Body.Close()
-
-	if resp.StatusCode != http.StatusOK {
-		body, _ := io.ReadAll(resp.Body)
-		return nil, fmt.Errorf("非 200 响应: %d %s", resp.StatusCode, string(body))
+		return nil, fmt.Errorf("采集失败（%d次尝试）: %w", strategy.MaxRetries+1, lastErr)
 	}
 
-	body, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return nil, fmt.Errorf("读取响应失败: %w", err)
-	}
-
-	// 健壮解析，兼容字段缺失和结构差异
-	models, err := parseModels(body)
-	if err != nil {
-		return nil, fmt.Errorf("JSON 解析失败: %w", err)
-	}
-
-	// TODO: 字段标准化映射（OpenRouter id → 标准厂商名、模型名）
 	return models, nil
 }
 
-// parseModels 健壮解析模型列表，兼容字段缺失/类型不一致/嵌套结构差异
 func parseModels(raw []byte) ([]ModelInfo, error) {
 	var wrapper struct {
 		Data json.RawMessage `json:"data"`
@@ -151,7 +238,6 @@ func parseModels(raw []byte) ([]ModelInfo, error) {
 	if err := json.Unmarshal(raw, &wrapper); err != nil {
 		return nil, fmt.Errorf("解析 data 字段失败: %w", err)
 	}
-	// data 为数组，每元素字段可能不同，统一用 map[string]any 兼容
 	var rawItems []any
 	if err := json.Unmarshal(wrapper.Data, &rawItems); err != nil {
 		return nil, fmt.Errorf("解析模型数组失败: %w", err)
@@ -161,17 +247,16 @@ func parseModels(raw []byte) ([]ModelInfo, error) {
 	for _, item := range rawItems {
 		m, ok := item.(map[string]any)
 		if !ok {
-			continue // 跳过非法条目
+			continue
 		}
 		model := ModelInfo{
-		ID:   getString(m, "id"),
-		Name: getString(m, "name"),
-	}
+			ID:   getString(m, "id"),
+			Name: getString(m, "name"),
+		}
 		if model.ID == "" {
-			continue // id 为必填
+			continue
 		}
 
-		// pricing 可能为嵌套对象（如 {openrouter: {input: 1}}），尝试多路径取值
 		if p, ok := m["pricing"].(map[string]any); ok {
 			model.Pricing.Input = getPrice(p, "input", "prompt")
 			model.Pricing.Output = getPrice(p, "output", "completion")
@@ -214,7 +299,6 @@ func getInt64(m map[string]any, key string) int64 {
 	return 0
 }
 
-// getPrice 多路径取值，兼容不同嵌套结构（如 {input:1} 或 {openrouter:{input:1}}）
 func getPrice(m map[string]any, keys ...string) float64 {
 	for _, k := range keys {
 		if v, ok := m[k].(float64); ok {
@@ -224,13 +308,12 @@ func getPrice(m map[string]any, keys ...string) float64 {
 	return 0
 }
 
-// summarize 输出采集摘要到 JSON 文件（保持向后兼容）
 func summarize(outPath string, models []ModelInfo) error {
 	return writeJSON(outPath, models)
 }
 
-// summarizeDB 将采集结果写入 PostgreSQL（models + model_prices 表）
-func summarizeDB(connStr string, models []ModelInfo) error {
+// summarizeDB 将采集结果写入 PostgreSQL（批量插入 + ProviderMapper + 价格变动检测 + audit_log）
+func summarizeDB(connStr string, models []ModelInfo, batchSize int) error {
 	db, err := sql.Open("postgres", connStr)
 	if err != nil {
 		return fmt.Errorf("连接数据库失败: %w", err)
@@ -241,66 +324,229 @@ func summarizeDB(connStr string, models []ModelInfo) error {
 		return fmt.Errorf("ping 数据库失败: %w", err)
 	}
 
-	tx, err := db.Begin()
-	if err != nil {
-		return fmt.Errorf("开启事务失败: %w", err)
-	}
-	defer tx.Rollback()
-
+	batchID := fmt.Sprintf("batch-%d", time.Now().Unix())
 	now := time.Now()
+	effectiveDate := now.Format("2006-01-02")
+
+	// 获取默认 operator（OpenRouter）
+	var operatorID int64
+	err = db.QueryRow("SELECT id FROM operator WHERE name = 'OpenRouter' LIMIT 1").Scan(&operatorID)
+	if err != nil {
+		logger.Warn("未找到 OpenRouter operator，使用 NULL", "error", err)
+		operatorID = 0
+	}
+
+	// 获取上次价格数据（用于变动检测）
+	lastPrices := make(map[int64]ModelPricing)
+	rows, err := db.Query(`
+		SELECT model_id, input_price_per_mtok, output_price_per_mtok
+		FROM region_pricing
+		WHERE operator_id = $1 AND effective_date = (
+			SELECT MAX(effective_date) FROM region_pricing WHERE operator_id = $1
+		)
+	`, operatorID)
+	if err == nil {
+		for rows.Next() {
+			var mid int64
+			var p ModelPricing
+			if err := rows.Scan(&mid, &p.Input, &p.Output); err == nil {
+				lastPrices[mid] = p
+			}
+		}
+		rows.Close()
+	}
+
 	insertedModels := 0
 	insertedPrices := 0
+	priceChanges := 0
 
-	for _, m := range models {
-		isFree := len(m.ID) > 5 && m.ID[len(m.ID)-5:] == ":free"
-		// upsert models 表
-		var modelID int64
-		err := tx.QueryRow(`
-			INSERT INTO models (source, external_id, name, description, context_length, capabilities, created_at_source, is_free, status, raw_payload, created_at, updated_at)
-			VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12)
-			ON CONFLICT (external_id) DO UPDATE SET
-				name = EXCLUDED.name,
-				description = EXCLUDED.description,
-				context_length = EXCLUDED.context_length,
-				capabilities = EXCLUDED.capabilities,
-				created_at_source = EXCLUDED.created_at_source,
-				is_free = EXCLUDED.is_free,
-				status = EXCLUDED.status,
-				raw_payload = EXCLUDED.raw_payload,
-				updated_at = $12
-			RETURNING id
-		`, "openrouter", m.ID, m.Name, m.Description, m.ContextLength,
-			jsonCapabilities(m.Capabilities), m.Created, isFree, "active",
-			rawPayload(m), now, now).Scan(&modelID)
-		if err != nil {
-			return fmt.Errorf("写入 models 失败 (%s): %w", m.ID, err)
+	// 批量处理
+	for i := 0; i < len(models); i += batchSize {
+		end := i + batchSize
+		if end > len(models) {
+			end = len(models)
 		}
-		insertedModels++
+		batch := models[i:end]
 
-		// upsert model_prices 表（当天有效日期）
-		effectiveDate := now.Format("2006-01-02")
-		_, err = tx.Exec(`
-			INSERT INTO model_prices (model_id, source, currency, input_price_per_mtok, output_price_per_mtok, effective_date, source_url, created_at)
-			VALUES ($1, $2, $3, $4, $5, $6, $7, $8)
-			ON CONFLICT (model_id, source, currency, effective_date) DO UPDATE SET
-				input_price_per_mtok = EXCLUDED.input_price_per_mtok,
-				output_price_per_mtok = EXCLUDED.output_price_per_mtok,
-				created_at = EXCLUDED.created_at
-		`, modelID, "openrouter", "USD", m.Pricing.Input, m.Pricing.Output, effectiveDate, "https://openrouter.ai/api/v1/models", now)
+		tx, err := db.Begin()
 		if err != nil {
-			return fmt.Errorf("写入 model_prices 失败 (%s): %w", m.ID, err)
+			return fmt.Errorf("开启事务失败: %w", err)
 		}
-		insertedPrices++
+
+		for _, m := range batch {
+			// 使用 ProviderMapper 映射厂商
+			mapping, err := collectors.MapOpenRouterID(m.ID)
+			if err != nil {
+				logger.Warn("Provider 映射失败", "id", m.ID, "error", err)
+				mapping = collectors.ModelMapping{
+					Provider:  collectors.ProviderInfo{ID: "unknown", Name: "Unknown"},
+					ModelName: m.Name,
+					RawID:     m.ID,
+					IsFree:    false,
+				}
+			}
+
+			// 查找或创建 provider_id
+			var providerID int64
+			err = tx.QueryRow("SELECT id FROM model_provider WHERE name = $1 LIMIT 1", mapping.Provider.Name).Scan(&providerID)
+			if err != nil {
+				// 未知厂商，插入
+				err = tx.QueryRow(`
+					INSERT INTO model_provider (name, name_cn, country, status)
+					VALUES ($1, $2, $3, 'active')
+					ON CONFLICT (name) DO UPDATE SET name = EXCLUDED.name
+					RETURNING id
+				`, mapping.Provider.Name, mapping.Provider.NameCN, mapping.Provider.Country).Scan(&providerID)
+				if err != nil {
+					logger.Warn("创建 provider 失败", "name", mapping.Provider.Name, "error", err)
+					providerID = 0
+				}
+			}
+
+			isFree := mapping.IsFree || (m.Pricing.Input == 0 && m.Pricing.Output == 0)
+
+			// upsert models 表（带新字段）
+			var modelID int64
+			err = tx.QueryRow(`
+				INSERT INTO models (
+					source, external_id, name, description, context_length,
+					capabilities, created_at_source, is_free, status,
+					raw_payload, provider_id, version, modality,
+					data_confidence, retrieved_at, batch_id, collector_version,
+					source_url, created_at, updated_at
+				) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $19)
+				ON CONFLICT (external_id) DO UPDATE SET
+					name = EXCLUDED.name,
+					description = EXCLUDED.description,
+					context_length = EXCLUDED.context_length,
+					capabilities = EXCLUDED.capabilities,
+					created_at_source = EXCLUDED.created_at_source,
+					is_free = EXCLUDED.is_free,
+					status = EXCLUDED.status,
+					raw_payload = EXCLUDED.raw_payload,
+					provider_id = EXCLUDED.provider_id,
+					data_confidence = 'official',
+					retrieved_at = EXCLUDED.retrieved_at,
+					batch_id = EXCLUDED.batch_id,
+					collector_version = EXCLUDED.collector_version,
+					updated_at = EXCLUDED.updated_at
+				RETURNING id
+			`,
+				"openrouter", m.ID, m.Name, m.Description, m.ContextLength,
+				jsonCapabilities(m.Capabilities), m.Created, isFree, "active",
+				rawPayload(m), providerID, "", "text",
+				"official", now, batchID, collectorVersion,
+				"https://openrouter.ai/api/v1/models", now).Scan(&modelID)
+			if err != nil {
+				_ = tx.Rollback()
+				return fmt.Errorf("写入 models 失败 (%s): %w", m.ID, err)
+			}
+			insertedModels++
+
+			// 写入 audit_log
+			_, _ = tx.Exec(`
+				INSERT INTO audit_log (table_name, record_id, field_name, old_value, new_value, operation, operator, batch_id, source_url)
+				VALUES ('models', $1, 'external_id', NULL, $2, 'INSERT', 'fetch_openrouter', $3, $4)
+			`, modelID, m.ID, batchID, "https://openrouter.ai/api/v1/models")
+
+			// upsert region_pricing 表（替代 model_prices）
+			sourceType := "reseller"
+			freeQuota := ""
+			freeLimitations := "[]"
+			rateLimit := "{}"
+			if isFree {
+				sourceType = "free_tier"
+				freeQuota = "Imported free-tier pricing entry"
+				freeLimitations = `["See source_url for current quota and policy"]`
+			}
+			var pricingID int64
+			err = tx.QueryRow(`
+				INSERT INTO region_pricing (
+					model_id, operator_id, region, currency,
+					input_price_per_mtok, output_price_per_mtok,
+					is_free, effective_date, source_url, source_type,
+					free_quota, free_limitations, rate_limit,
+					created_at, updated_at
+				) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $14)
+				ON CONFLICT (model_id, operator_id, region, currency, effective_date) DO UPDATE SET
+					input_price_per_mtok = EXCLUDED.input_price_per_mtok,
+					output_price_per_mtok = EXCLUDED.output_price_per_mtok,
+					is_free = EXCLUDED.is_free,
+					source_type = EXCLUDED.source_type,
+					free_quota = EXCLUDED.free_quota,
+					free_limitations = EXCLUDED.free_limitations,
+					rate_limit = EXCLUDED.rate_limit,
+					updated_at = EXCLUDED.updated_at
+				RETURNING id
+			`, modelID, operatorID, "global", "USD", m.Pricing.Input, m.Pricing.Output,
+				isFree, effectiveDate, "https://openrouter.ai/api/v1/models", sourceType,
+				freeQuota, freeLimitations, rateLimit, now).Scan(&pricingID)
+			if err != nil {
+				_ = tx.Rollback()
+				return fmt.Errorf("写入 region_pricing 失败 (%s): %w", m.ID, err)
+			}
+			insertedPrices++
+
+			// 价格变动检测（>5%）
+			if lastPrice, ok := lastPrices[modelID]; ok {
+				inputChange := calcChangePercent(lastPrice.Input, m.Pricing.Input)
+				outputChange := calcChangePercent(lastPrice.Output, m.Pricing.Output)
+				if abs(inputChange) > 5 || abs(outputChange) > 5 {
+					_, _ = tx.Exec(`
+						INSERT INTO pricing_history (
+							model_id, region, currency,
+							old_input_price, new_input_price,
+							old_output_price, new_output_price,
+							change_percent, changed_at
+						) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9)
+					`, modelID, "global", "USD",
+						lastPrice.Input, m.Pricing.Input,
+						lastPrice.Output, m.Pricing.Output,
+						max(abs(inputChange), abs(outputChange)), now)
+					priceChanges++
+				}
+			}
+		}
+
+		if err := tx.Commit(); err != nil {
+			return fmt.Errorf("提交事务失败: %w", err)
+		}
+
+		logger.Info("批次完成", "batch", i/batchSize+1, "records", len(batch))
 	}
 
-	if err := tx.Commit(); err != nil {
-		return fmt.Errorf("提交事务失败: %w", err)
-	}
-
-	fmt.Printf("PostgreSQL 写入完成: %d models, %d prices\n", insertedModels, insertedPrices)
+	logger.Info("PostgreSQL 写入完成",
+		"models", insertedModels,
+		"prices", insertedPrices,
+		"price_changes", priceChanges,
+		"batch_id", batchID)
 	return nil
 }
 
+func calcChangePercent(old, new float64) float64 {
+	if old == 0 {
+		if new == 0 {
+			return 0
+		}
+		return 100
+	}
+	return ((new - old) / old) * 100
+}
+
+func abs(v float64) float64 {
+	if v < 0 {
+		return -v
+	}
+	return v
+}
+
+func max(a, b float64) float64 {
+	if a > b {
+		return a
+	}
+	return b
+}
+
 func jsonCapabilities(caps []string) []byte {
 	if len(caps) == 0 {
 		return []byte("[]")
@@ -314,7 +560,6 @@ func rawPayload(m ModelInfo) []byte {
 	return b
 }
 
-// writeJSON 统一写入 JSON 文件（含摘要信息）
 func writeJSON(outPath string, models []ModelInfo) error {
 	total := len(models)
 	var freeCnt, paidCnt int
@@ -349,3 +594,24 @@ func writeJSON(outPath string, models []ModelInfo) error {
 	fmt.Printf("结果已写入: %s\n", outPath)
 	return nil
 }
+
+// recordCollectorStats 记录采集统计到 collector_stats 表
+func recordCollectorStats(connStr string, runErr error, duration time.Duration) error {
+	db, err := sql.Open("postgres", connStr)
+	if err != nil {
+		return err
+	}
+	defer db.Close()
+
+	success := runErr == nil
+	errMsg := ""
+	if runErr != nil {
+		errMsg = runErr.Error()
+	}
+
+	_, err = db.Exec(`
+		INSERT INTO collector_stats (source, batch_id, success, duration_ms, error_message, created_at)
+		VALUES ('openrouter', $1, $2, $3, $4, $5)
+	`, fmt.Sprintf("batch-%d", time.Now().Unix()), success, int(duration.Milliseconds()), errMsg, time.Now())
+	return err
+}

scripts/fetch_openrouter_test.go

@@ -1,3 +1,5 @@
+//go:build llm_script
+
 package main
 
 import (

scripts/fetch_tencent_catalog.go

@@ -0,0 +1,61 @@
+//go:build llm_script
+
+package main
+
+import (
+	"flag"
+	"fmt"
+	"io"
+	"net/http"
+	"os"
+	"time"
+)
+
+func main() {
+	var rawURL string
+	var dryRun bool
+	var timeoutSeconds int
+	var fixturePath string
+
+	flag.StringVar(&rawURL, "url", defaultTencentCatalogURL, "腾讯云公开目录 URL")
+	flag.BoolVar(&dryRun, "dry-run", false, "仅抓取并打印摘要，不写入数据库")
+	flag.IntVar(&timeoutSeconds, "timeout", int(defaultTencentCatalogTimeout/time.Second), "请求超时（秒）")
+	flag.StringVar(&fixturePath, "fixture", "", "本地 HTML/Text 样例文件，优先用于离线 dry-run")
+	flag.Parse()
+
+	cfg := fetchTencentCatalogConfig{
+		URL:     rawURL,
+		DryRun:  dryRun,
+		Timeout: time.Duration(timeoutSeconds) * time.Second,
+		Fixture: fixturePath,
+	}
+
+	client := &http.Client{Timeout: cfg.Timeout}
+	if err := runTencentCatalog(cfg, client, os.Stdout); err != nil {
+		fmt.Fprintf(os.Stderr, "fetch_tencent_catalog: %v\n", err)
+		os.Exit(1)
+	}
+}
+
+func runTencentCatalog(cfg fetchTencentCatalogConfig, client *http.Client, out io.Writer) error {
+	raw, err := fetchTencentCatalogContent(cfg, client)
+	if err != nil {
+		return err
+	}
+
+	catalog, err := parseTencentCatalog(raw)
+	if err != nil {
+		return err
+	}
+
+	_, err = fmt.Fprintf(
+		out,
+		"source=tencent-public-catalog updated_at=%s plans=%d models=%d series=%s dry_run=%t\n",
+		catalog.UpdatedAt,
+		len(catalog.Plans),
+		len(catalog.Models),
+		formatSeriesSummary(catalog.Plans),
+		cfg.DryRun,
+	)
+	return err
+}

scripts/fetch_tencent_catalog_test.go

@@ -0,0 +1,98 @@
+//go:build llm_script
+
+package main
+
+import (
+	"bytes"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+func TestParseTencentCatalogExtractsPlansAndModels(t *testing.T) {
+	raw, err := os.ReadFile(filepath.Join("testdata", "tencent_token_plan_sample.txt"))
+	if err != nil {
+		t.Fatalf("读取样例失败: %v", err)
+	}
+
+	catalog, err := parseTencentCatalog(string(raw))
+	if err != nil {
+		t.Fatalf("parseTencentCatalog 失败: %v", err)
+	}
+
+	if catalog.UpdatedAt != "2026-04-27 17:18:02" {
+		t.Fatalf("更新时间错误: %q", catalog.UpdatedAt)
+	}
+
+	if len(catalog.Plans) != 8 {
+		t.Fatalf("期望 8 个套餐，实际 %d", len(catalog.Plans))
+	}
+
+	if len(catalog.Models) != 11 {
+		t.Fatalf("期望 11 个模型，实际 %d", len(catalog.Models))
+	}
+
+	firstPlan := catalog.Plans[0]
+	if firstPlan.Series != "通用 Token Plan" {
+		t.Fatalf("套餐系列错误: %q", firstPlan.Series)
+	}
+	if firstPlan.Tier != "Lite" {
+		t.Fatalf("套餐档位错误: %q", firstPlan.Tier)
+	}
+	if firstPlan.Price != "39元/月" {
+		t.Fatalf("套餐价格错误: %q", firstPlan.Price)
+	}
+	if firstPlan.Quota != "3500万 Tokens" {
+		t.Fatalf("套餐额度错误: %q", firstPlan.Quota)
+	}
+
+	lastModel := catalog.Models[len(catalog.Models)-1]
+	if lastModel.Name != "Hy3 preview" {
+		t.Fatalf("最后一个模型错误: %q", lastModel.Name)
+	}
+	if lastModel.ModelID != "hy3-preview" {
+		t.Fatalf("最后一个模型 ID 错误: %q", lastModel.ModelID)
+	}
+	if lastModel.ContextLength != 262144 {
+		t.Fatalf("Hy3 preview 上下文长度错误: %d", lastModel.ContextLength)
+	}
+}
+
+func TestRunTencentCatalogDryRunPrintsSummary(t *testing.T) {
+	raw, err := os.ReadFile(filepath.Join("testdata", "tencent_token_plan_sample.txt"))
+	if err != nil {
+		t.Fatalf("读取样例失败: %v", err)
+	}
+
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "text/html; charset=utf-8")
+		_, _ = w.Write([]byte("<html><body><main><pre>" + string(raw) + "</pre></main></body></html>"))
+	}))
+	defer server.Close()
+
+	var out bytes.Buffer
+	err = runTencentCatalog(fetchTencentCatalogConfig{
+		URL:     server.URL,
+		DryRun:  true,
+		Timeout: defaultTencentCatalogTimeout,
+	}, server.Client(), &out)
+	if err != nil {
+		t.Fatalf("runTencentCatalog 失败: %v", err)
+	}
+
+	output := out.String()
+	for _, want := range []string{
+		"source=tencent-public-catalog",
+		"plans=8",
+		"models=11",
+		"series=Hy Token Plan:4,通用 Token Plan:4",
+		"updated_at=2026-04-27 17:18:02",
+	} {
+		if !strings.Contains(output, want) {
+			t.Fatalf("输出缺少 %q，实际: %q", want, output)
+		}
+	}
+}

scripts/generate_daily_report_test.go

@@ -0,0 +1,115 @@
+//go:build llm_script
+
+package main
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+func TestGenerateMarkdownV3IncludesTencentSubscriptionSection(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "daily_report.md")
+	report := &ReportV3{
+		Date:        "2026-05-13",
+		TotalModels: 502,
+		QualitySummary: DataQualitySummary{
+			Total: 502,
+			Fresh: 490,
+			CNY:   126,
+			USD:   376,
+		},
+		TencentSubscriptionPlans: []SubscriptionPlanInfo{
+			{
+				PlanName:      "通用 Token Plan Lite",
+				PlanFamily:    "token_plan",
+				Tier:          "Lite",
+				Currency:      "CNY",
+				ListPrice:     39,
+				QuotaValue:    35000000,
+				QuotaUnit:     "tokens/month",
+				ContextWindow: 0,
+				ModelCount:    10,
+				ModelPreview:  "tc-code-latest, glm-5, glm-5.1",
+			},
+			{
+				PlanName:      "Hy Token Plan Max",
+				PlanFamily:    "token_plan",
+				Tier:          "Max",
+				Currency:      "CNY",
+				ListPrice:     468,
+				QuotaValue:    650000000,
+				QuotaUnit:     "tokens/month",
+				ContextWindow: 262144,
+				ModelCount:    1,
+				ModelPreview:  "hy3-preview",
+			},
+		},
+	}
+
+	if err := generateMarkdownV3(report, path); err != nil {
+		t.Fatalf("generateMarkdownV3 returned error: %v", err)
+	}
+
+	body, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("read markdown output: %v", err)
+	}
+
+	content := string(body)
+	for _, want := range []string{
+		"## 💳 腾讯云套餐订阅价",
+		"通用 Token Plan Lite",
+		"Hy Token Plan Max",
+		"¥39.00/月",
+		"3500万 Tokens/月",
+		"256K",
+	} {
+		if !strings.Contains(content, want) {
+			t.Fatalf("markdown missing %q\n%s", want, content)
+		}
+	}
+}
+
+func TestGenerateHTMLV3IncludesTencentSubscriptionSection(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "daily_report.html")
+	report := &ReportV3{
+		Date:        "2026-05-13",
+		TotalModels: 502,
+		TencentSubscriptionPlans: []SubscriptionPlanInfo{
+			{
+				PlanName:     "通用 Token Plan Lite",
+				PlanFamily:   "token_plan",
+				Tier:         "Lite",
+				Currency:     "CNY",
+				ListPrice:    39,
+				QuotaValue:   35000000,
+				QuotaUnit:    "tokens/month",
+				ModelCount:   10,
+				ModelPreview: "tc-code-latest, glm-5, glm-5.1",
+			},
+		},
+	}
+
+	if err := generateHTMLV3(report, path); err != nil {
+		t.Fatalf("generateHTMLV3 returned error: %v", err)
+	}
+
+	body, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("read html output: %v", err)
+	}
+
+	content := string(body)
+	for _, want := range []string{
+		"💳 腾讯云套餐订阅价",
+		"通用 Token Plan Lite",
+		"¥39.00/月",
+		"3500万 Tokens/月",
+	} {
+		if !strings.Contains(content, want) {
+			t.Fatalf("html missing %q\n%s", want, content)
+		}
+	}
+}

scripts/generate_video_digest.go

@@ -0,0 +1,617 @@
+//go:build llm_script
+
+package main
+
+import (
+	"bytes"
+	"encoding/binary"
+	"encoding/json"
+	"errors"
+	"flag"
+	"fmt"
+	"image"
+	"image/color"
+	"image/draw"
+	"image/gif"
+	"image/png"
+	"math"
+	"os"
+	"path/filepath"
+	"sort"
+	"strings"
+	"time"
+)
+
+type reportRow struct {
+	Model    string `json:"model"`
+	Provider string `json:"provider"`
+	Scene    string `json:"scene"`
+	Input    string `json:"input,omitempty"`
+	Output   string `json:"output,omitempty"`
+	Context  string `json:"context,omitempty"`
+}
+
+type dailyReport struct {
+	ReportDate      string            `json:"report_date"`
+	Stats           map[string]string `json:"stats"`
+	International   []reportRow       `json:"international"`
+	Domestic        []reportRow       `json:"domestic"`
+	SourceReport    string            `json:"source_report"`
+	GeneratedAt     string            `json:"generated_at,omitempty"`
+	SelectedSection string            `json:"selected_section,omitempty"`
+}
+
+type DigestCard struct {
+	Slug        string   `json:"slug"`
+	Title       string   `json:"title"`
+	Headline    string   `json:"headline"`
+	BulletLines []string `json:"bullet_lines"`
+	Narration   string   `json:"narration"`
+	FramePath   string   `json:"frame_path,omitempty"`
+	ScriptPath  string   `json:"script_path,omitempty"`
+}
+
+type digestManifest struct {
+	ReportDate   string       `json:"report_date"`
+	SourceReport string       `json:"source_report"`
+	GeneratedAt  string       `json:"generated_at"`
+	OutputDir    string       `json:"output_dir"`
+	VideoPath    string       `json:"video_path"`
+	AudioPath    string       `json:"audio_path"`
+	Cards        []DigestCard `json:"cards"`
+}
+
+var framePalette = color.Palette{
+	color.RGBA{12, 18, 28, 255},
+	color.RGBA{32, 48, 74, 255},
+	color.RGBA{91, 192, 190, 255},
+	color.RGBA{245, 247, 250, 255},
+	color.RGBA{255, 196, 61, 255},
+	color.RGBA{255, 107, 107, 255},
+}
+
+var slideBackgrounds = []uint8{1, 2, 1, 4, 5}
+
+func main() {
+	var reportPath string
+	var reportDate string
+	var outputDir string
+
+	flag.StringVar(&reportPath, "report", "", "path to daily markdown report")
+	flag.StringVar(&reportDate, "date", "", "report date in YYYY-MM-DD")
+	flag.StringVar(&outputDir, "output-dir", "", "output directory for video digest artifacts")
+	flag.Parse()
+
+	resolvedReport, err := resolveReportPath(reportPath, reportDate)
+	if err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		os.Exit(1)
+	}
+
+	content, err := os.ReadFile(resolvedReport)
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "read report failed: %v\n", err)
+		os.Exit(1)
+	}
+
+	report, err := parseDailyReport(content)
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "parse report failed: %v\n", err)
+		os.Exit(1)
+	}
+	report.SourceReport = resolvedReport
+	if reportDate == "" {
+		reportDate = report.ReportDate
+	}
+	if reportDate == "" {
+		reportDate = time.Now().Format("2006-01-02")
+	}
+
+	cards := buildDigestCards(report)
+	if len(cards) == 0 {
+		fmt.Fprintln(os.Stderr, "no digest cards generated")
+		os.Exit(1)
+	}
+
+	if outputDir == "" {
+		outputDir = filepath.Join(filepath.Dir(resolvedReport), "video", reportDate)
+	}
+
+	manifest, err := generateDigestArtifacts(report, cards, outputDir)
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "generate digest artifacts failed: %v\n", err)
+		os.Exit(1)
+	}
+
+	fmt.Printf("report=%s\n", manifest.SourceReport)
+	fmt.Printf("output=%s\n", manifest.OutputDir)
+	fmt.Printf("cards=%d\n", len(manifest.Cards))
+	fmt.Printf("video=%s\n", manifest.VideoPath)
+	fmt.Printf("audio=%s\n", manifest.AudioPath)
+}
+
+func resolveReportPath(explicitPath string, reportDate string) (string, error) {
+	if explicitPath != "" {
+		return explicitPath, nil
+	}
+	root := "/home/long/project/llm-intelligence/reports/daily"
+	if reportDate != "" {
+		return filepath.Join(root, fmt.Sprintf("daily_report_%s.md", reportDate)), nil
+	}
+
+	matches, err := filepath.Glob(filepath.Join(root, "daily_report_*.md"))
+	if err != nil {
+		return "", err
+	}
+	if len(matches) == 0 {
+		return "", errors.New("no daily report markdown files found")
+	}
+	sort.Strings(matches)
+	return matches[len(matches)-1], nil
+}
+
+func parseDailyReport(content []byte) (dailyReport, error) {
+	report := dailyReport{
+		Stats: make(map[string]string),
+	}
+	lines := strings.Split(string(content), "\n")
+	section := ""
+	for _, rawLine := range lines {
+		line := strings.TrimSpace(rawLine)
+		if strings.HasPrefix(line, "**报告日期**:") {
+			report.ReportDate = strings.TrimSpace(strings.TrimPrefix(line, "**报告日期**:"))
+			continue
+		}
+		switch line {
+		case "## 📊 数据质量摘要":
+			section = "stats"
+			continue
+		case "## 🌍 国际推荐模型 TOP 5":
+			section = "international"
+			continue
+		case "## 🇨🇳 国内模型 TOP 10":
+			section = "domestic"
+			continue
+		}
+		if strings.HasPrefix(line, "## ") {
+			section = ""
+			continue
+		}
+		if !strings.HasPrefix(line, "|") || strings.Contains(line, "------") {
+			continue
+		}
+
+		parts := splitMarkdownTableLine(line)
+		switch section {
+		case "stats":
+			if len(parts) >= 2 && parts[0] != "指标" {
+				report.Stats[parts[0]] = parts[1]
+			}
+		case "international":
+			if row, ok := parseModelRow(parts); ok {
+				report.International = append(report.International, row)
+			}
+		case "domestic":
+			if row, ok := parseModelRow(parts); ok {
+				report.Domestic = append(report.Domestic, row)
+			}
+		}
+	}
+	if report.ReportDate == "" {
+		return report, errors.New("report date not found")
+	}
+	return report, nil
+}
+
+func splitMarkdownTableLine(line string) []string {
+	trimmed := strings.Trim(line, "|")
+	parts := strings.Split(trimmed, "|")
+	out := make([]string, 0, len(parts))
+	for _, part := range parts {
+		out = append(out, strings.TrimSpace(part))
+	}
+	return out
+}
+
+func parseModelRow(parts []string) (reportRow, bool) {
+	if len(parts) < 7 || parts[0] == "排名" {
+		return reportRow{}, false
+	}
+	return reportRow{
+		Model:    parts[1],
+		Provider: parts[2],
+		Scene:    parts[3],
+		Input:    parts[4],
+		Output:   parts[5],
+		Context:  parts[6],
+	}, true
+}
+
+func buildDigestCards(report dailyReport) []DigestCard {
+	total := report.Stats["模型总数"]
+	cny := report.Stats["CNY定价"]
+	usd := report.Stats["USD定价"]
+
+	codeRows := pickSceneRows(report, "代码")
+	reasoningRows := pickSceneRows(report, "推理")
+	visionRows := pickSceneRows(report, "视觉")
+
+	cards := []DigestCard{
+		newDigestCard(
+			"code",
+			"Code Digest",
+			fmt.Sprintf("%s total models tracked", total),
+			codeRows,
+			fmt.Sprintf("Code digest highlights %d candidate models. CNY priced entries %s.", len(codeRows), cny),
+		),
+		newDigestCard(
+			"reasoning",
+			"Reasoning Digest",
+			fmt.Sprintf("USD priced entries %s", usd),
+			reasoningRows,
+			fmt.Sprintf("Reasoning digest focuses on %d reasoning oriented models.", len(reasoningRows)),
+		),
+		newDigestCard(
+			"vision",
+			"Vision Digest",
+			fmt.Sprintf("CNY priced entries %s", cny),
+			visionRows,
+			fmt.Sprintf("Vision digest contains %d multimodal candidates from the latest report.", len(visionRows)),
+		),
+		newDigestCard(
+			"domestic",
+			"Domestic Digest",
+			fmt.Sprintf("Domestic pricing entries %s", cny),
+			firstRows(report.Domestic, 3),
+			fmt.Sprintf("Domestic digest summarizes top local platforms with %d highlighted entries.", min(3, len(report.Domestic))),
+		),
+		newDigestCard(
+			"global",
+			"Global Digest",
+			fmt.Sprintf("Global pricing entries %s", usd),
+			firstRows(report.International, 3),
+			fmt.Sprintf("Global digest summarizes top international recommendations with %d highlighted entries.", min(3, len(report.International))),
+		),
+	}
+	return cards
+}
+
+func newDigestCard(slug string, title string, headline string, rows []reportRow, narration string) DigestCard {
+	bullets := make([]string, 0, len(rows))
+	for _, row := range rows {
+		bullets = append(bullets, fmt.Sprintf("%s - %s - %s", row.Model, row.Provider, row.Scene))
+	}
+	if len(bullets) == 0 {
+		bullets = append(bullets, "No matching models in current report")
+	}
+	return DigestCard{
+		Slug:        slug,
+		Title:       title,
+		Headline:    headline,
+		BulletLines: bullets,
+		Narration:   narration,
+	}
+}
+
+func pickSceneRows(report dailyReport, scene string) []reportRow {
+	rows := make([]reportRow, 0, 3)
+	for _, source := range [][]reportRow{report.Domestic, report.International} {
+		for _, row := range source {
+			if strings.Contains(row.Scene, scene) {
+				rows = append(rows, row)
+			}
+			if len(rows) == 3 {
+				return rows
+			}
+		}
+	}
+	return rows
+}
+
+func firstRows(rows []reportRow, n int) []reportRow {
+	if len(rows) < n {
+		n = len(rows)
+	}
+	out := make([]reportRow, 0, n)
+	for i := 0; i < n; i++ {
+		out = append(out, rows[i])
+	}
+	return out
+}
+
+func generateDigestArtifacts(report dailyReport, cards []DigestCard, outputDir string) (digestManifest, error) {
+	scriptDir := filepath.Join(outputDir, "scripts")
+	frameDir := filepath.Join(outputDir, "frames")
+	if err := os.MkdirAll(scriptDir, 0o755); err != nil {
+		return digestManifest{}, err
+	}
+	if err := os.MkdirAll(frameDir, 0o755); err != nil {
+		return digestManifest{}, err
+	}
+
+	frames := make([]*image.Paletted, 0, len(cards))
+	delays := make([]int, 0, len(cards))
+	for i, card := range cards {
+		frame := renderCardFrame(card, i)
+		framePath := filepath.Join(frameDir, fmt.Sprintf("%02d_%s.png", i+1, card.Slug))
+		if err := writePNG(framePath, frame); err != nil {
+			return digestManifest{}, err
+		}
+
+		scriptPath := filepath.Join(scriptDir, fmt.Sprintf("%02d_%s.md", i+1, card.Slug))
+		if err := os.WriteFile(scriptPath, []byte(renderCardScript(card)), 0o644); err != nil {
+			return digestManifest{}, err
+		}
+
+		card.FramePath = framePath
+		card.ScriptPath = scriptPath
+		cards[i] = card
+		frames = append(frames, frame)
+		delays = append(delays, 120)
+	}
+
+	videoPath := filepath.Join(outputDir, "video_digest.gif")
+	if err := writeAnimatedGIF(videoPath, frames, delays); err != nil {
+		return digestManifest{}, err
+	}
+
+	audioData, err := buildNarrationAudio(cards)
+	if err != nil {
+		return digestManifest{}, err
+	}
+	audioPath := filepath.Join(outputDir, "narration.wav")
+	if err := os.WriteFile(audioPath, audioData, 0o644); err != nil {
+		return digestManifest{}, err
+	}
+
+	manifest := digestManifest{
+		ReportDate:   report.ReportDate,
+		SourceReport: report.SourceReport,
+		GeneratedAt:  time.Now().Format(time.RFC3339),
+		OutputDir:    outputDir,
+		VideoPath:    videoPath,
+		AudioPath:    audioPath,
+		Cards:        cards,
+	}
+
+	manifestPath := filepath.Join(outputDir, "manifest.json")
+	payload, err := json.MarshalIndent(manifest, "", "  ")
+	if err != nil {
+		return digestManifest{}, err
+	}
+	if err := os.WriteFile(manifestPath, payload, 0o644); err != nil {
+		return digestManifest{}, err
+	}
+
+	return manifest, nil
+}
+
+func renderCardScript(card DigestCard) string {
+	var b strings.Builder
+	b.WriteString("# " + card.Title + "\n\n")
+	b.WriteString("## Headline\n")
+	b.WriteString("- " + card.Headline + "\n\n")
+	b.WriteString("## Narration\n")
+	b.WriteString("- " + card.Narration + "\n\n")
+	b.WriteString("## Bullet Lines\n")
+	for _, line := range card.BulletLines {
+		b.WriteString("- " + line + "\n")
+	}
+	return b.String()
+}
+
+func writePNG(path string, img image.Image) error {
+	f, err := os.Create(path)
+	if err != nil {
+		return err
+	}
+	defer f.Close()
+	return png.Encode(f, img)
+}
+
+func writeAnimatedGIF(path string, frames []*image.Paletted, delays []int) error {
+	f, err := os.Create(path)
+	if err != nil {
+		return err
+	}
+	defer f.Close()
+	return gif.EncodeAll(f, &gif.GIF{Image: frames, Delay: delays, LoopCount: 0})
+}
+
+func renderCardFrame(card DigestCard, index int) *image.Paletted {
+	rect := image.Rect(0, 0, 640, 360)
+	img := image.NewPaletted(rect, framePalette)
+	bg := slideBackgrounds[index%len(slideBackgrounds)]
+	draw.Draw(img, rect, &image.Uniform{framePalette[bg]}, image.Point{}, draw.Src)
+
+	fillRect(img, 18, 18, 622, 342, 0)
+	fillRect(img, 28, 28, 612, 88, bg)
+	fillRect(img, 28, 104, 612, 332, 1)
+
+	drawRasterText(img, 40, 42, 3, sanitizeFrameText(card.Title), 3)
+	drawRasterText(img, 40, 116, 2, sanitizeFrameText(card.Headline), 4)
+	for i, line := range firstStrings(card.BulletLines, 3) {
+		drawRasterText(img, 40, 160+i*42, 2, sanitizeFrameText(line), 3)
+	}
+	drawRasterText(img, 40, 302, 1, sanitizeFrameText("LLM INTELLIGENCE VIDEO DIGEST"), 4)
+	return img
+}
+
+func firstStrings(lines []string, n int) []string {
+	if len(lines) < n {
+		n = len(lines)
+	}
+	out := make([]string, 0, n)
+	for i := 0; i < n; i++ {
+		out = append(out, lines[i])
+	}
+	return out
+}
+
+func sanitizeFrameText(input string) string {
+	upper := strings.ToUpper(input)
+	var b strings.Builder
+	for _, r := range upper {
+		if _, ok := glyphs[r]; ok {
+			b.WriteRune(r)
+			continue
+		}
+		switch {
+		case r >= 'A' && r <= 'Z':
+			b.WriteRune(r)
+		case r >= '0' && r <= '9':
+			b.WriteRune(r)
+		default:
+			b.WriteRune(' ')
+		}
+	}
+	return strings.Join(strings.Fields(b.String()), " ")
+}
+
+func fillRect(img *image.Paletted, x1 int, y1 int, x2 int, y2 int, idx uint8) {
+	for y := y1; y < y2; y++ {
+		for x := x1; x < x2; x++ {
+			img.SetColorIndex(x, y, idx)
+		}
+	}
+}
+
+func drawRasterText(img *image.Paletted, x int, y int, scale int, text string, idx uint8) {
+	cursor := x
+	for _, r := range text {
+		pattern, ok := glyphs[r]
+		if !ok {
+			pattern = glyphs[' ']
+		}
+		for row, bits := range pattern {
+			for col := 0; col < 5; col++ {
+				if bits&(1<<(4-col)) == 0 {
+					continue
+				}
+				fillRect(img, cursor+col*scale, y+row*scale, cursor+(col+1)*scale, y+(row+1)*scale, idx)
+			}
+		}
+		cursor += 6 * scale
+	}
+}
+
+func buildNarrationAudio(cards []DigestCard) ([]byte, error) {
+	const sampleRate = 16000
+	var pcm []int16
+	for i, card := range cards {
+		freq := 330.0 + float64(i)*55.0
+		duration := 0.9 + float64(len(card.BulletLines))*0.18
+		pcm = append(pcm, synthTone(freq, duration, sampleRate)...)
+		pcm = append(pcm, make([]int16, sampleRate/5)...)
+	}
+	return encodeWAV(pcm, sampleRate), nil
+}
+
+func synthTone(freq float64, duration float64, sampleRate int) []int16 {
+	samples := int(duration * float64(sampleRate))
+	out := make([]int16, 0, samples)
+	for i := 0; i < samples; i++ {
+		t := float64(i) / float64(sampleRate)
+		envelope := 1.0
+		if i < sampleRate/50 {
+			envelope = float64(i) / float64(sampleRate/50)
+		}
+		if i > samples-sampleRate/25 {
+			remaining := samples - i
+			if remaining > 0 {
+				envelope = minFloat(envelope, float64(remaining)/float64(sampleRate/25))
+			}
+		}
+		value := math.Sin(2*math.Pi*freq*t) + 0.35*math.Sin(2*math.Pi*(freq/2)*t)
+		out = append(out, int16(value*envelope*12000))
+	}
+	return out
+}
+
+func encodeWAV(samples []int16, sampleRate int) []byte {
+	const channels = 1
+	const bitsPerSample = 16
+	dataSize := len(samples) * 2
+	byteRate := sampleRate * channels * bitsPerSample / 8
+	blockAlign := channels * bitsPerSample / 8
+
+	var buf bytes.Buffer
+	buf.WriteString("RIFF")
+	_ = binary.Write(&buf, binary.LittleEndian, uint32(36+dataSize))
+	buf.WriteString("WAVE")
+	buf.WriteString("fmt ")
+	_ = binary.Write(&buf, binary.LittleEndian, uint32(16))
+	_ = binary.Write(&buf, binary.LittleEndian, uint16(1))
+	_ = binary.Write(&buf, binary.LittleEndian, uint16(channels))
+	_ = binary.Write(&buf, binary.LittleEndian, uint32(sampleRate))
+	_ = binary.Write(&buf, binary.LittleEndian, uint32(byteRate))
+	_ = binary.Write(&buf, binary.LittleEndian, uint16(blockAlign))
+	_ = binary.Write(&buf, binary.LittleEndian, uint16(bitsPerSample))
+	buf.WriteString("data")
+	_ = binary.Write(&buf, binary.LittleEndian, uint32(dataSize))
+	for _, sample := range samples {
+		_ = binary.Write(&buf, binary.LittleEndian, sample)
+	}
+	return buf.Bytes()
+}
+
+func min(a int, b int) int {
+	if a < b {
+		return a
+	}
+	return b
+}
+
+func minFloat(a float64, b float64) float64 {
+	if a < b {
+		return a
+	}
+	return b
+}
+
+var glyphs = map[rune][7]uint8{
+	' ': {0, 0, 0, 0, 0, 0, 0},
+	'-': {0, 0, 0, 31, 0, 0, 0},
+	'.': {0, 0, 0, 0, 0, 12, 12},
+	':': {0, 12, 12, 0, 12, 12, 0},
+	'/': {1, 2, 4, 8, 16, 0, 0},
+	'+': {0, 4, 4, 31, 4, 4, 0},
+	'(': {2, 4, 8, 8, 8, 4, 2},
+	')': {8, 4, 2, 2, 2, 4, 8},
+	'0': {14, 17, 19, 21, 25, 17, 14},
+	'1': {4, 12, 4, 4, 4, 4, 14},
+	'2': {14, 17, 1, 2, 4, 8, 31},
+	'3': {30, 1, 1, 14, 1, 1, 30},
+	'4': {2, 6, 10, 18, 31, 2, 2},
+	'5': {31, 16, 16, 30, 1, 1, 30},
+	'6': {14, 16, 16, 30, 17, 17, 14},
+	'7': {31, 1, 2, 4, 8, 8, 8},
+	'8': {14, 17, 17, 14, 17, 17, 14},
+	'9': {14, 17, 17, 15, 1, 1, 14},
+	'A': {14, 17, 17, 31, 17, 17, 17},
+	'B': {30, 17, 17, 30, 17, 17, 30},
+	'C': {14, 17, 16, 16, 16, 17, 14},
+	'D': {28, 18, 17, 17, 17, 18, 28},
+	'E': {31, 16, 16, 30, 16, 16, 31},
+	'F': {31, 16, 16, 30, 16, 16, 16},
+	'G': {14, 17, 16, 16, 19, 17, 15},
+	'H': {17, 17, 17, 31, 17, 17, 17},
+	'I': {14, 4, 4, 4, 4, 4, 14},
+	'J': {7, 2, 2, 2, 18, 18, 12},
+	'K': {17, 18, 20, 24, 20, 18, 17},
+	'L': {16, 16, 16, 16, 16, 16, 31},
+	'M': {17, 27, 21, 17, 17, 17, 17},
+	'N': {17, 25, 21, 19, 17, 17, 17},
+	'O': {14, 17, 17, 17, 17, 17, 14},
+	'P': {30, 17, 17, 30, 16, 16, 16},
+	'Q': {14, 17, 17, 17, 21, 18, 13},
+	'R': {30, 17, 17, 30, 20, 18, 17},
+	'S': {15, 16, 16, 14, 1, 1, 30},
+	'T': {31, 4, 4, 4, 4, 4, 4},
+	'U': {17, 17, 17, 17, 17, 17, 14},
+	'V': {17, 17, 17, 17, 17, 10, 4},
+	'W': {17, 17, 17, 17, 21, 27, 17},
+	'X': {17, 17, 10, 4, 10, 17, 17},
+	'Y': {17, 17, 10, 4, 4, 4, 4},
+	'Z': {31, 1, 2, 4, 8, 16, 31},
+}

scripts/generate_video_digest_test.go

@@ -0,0 +1,77 @@
+//go:build llm_script
+
+package main
+
+import (
+	"bytes"
+	"testing"
+)
+
+const sampleDailyReport = `# Daily Report
+
+**报告日期**: 2026-05-11
+
+## 📊 数据质量摘要
+
+| 指标 | 数值 |
+|------|------|
+| 模型总数 | 501 |
+| CNY定价 | 126 |
+| USD定价 | 375 |
+
+## 🌍 国际推荐模型 TOP 5
+
+| 排名 | 模型 | 厂商 | 场景 | 输入(原价) | 输出(原价) | 上下文 |
+|------|------|------|------|-----------|-----------|--------|
+| 1 | GPT-5.4 Mini | OpenAI | 代码 | $0.75 | $4.50 | 200000 |
+| 2 | DeepSeek-V3 | DeepSeek | 推理 | ¥1.00 | ¥2.00 | 64000 |
+| 3 | Qwen3-VL-32B | Alibaba | 视觉 | ¥0.50 | ¥1.00 | 32000 |
+
+## 🇨🇳 国内模型 TOP 10
+
+| 排名 | 模型 | 厂商 | 场景 | 输入(CNY) | 输出(CNY) | 上下文 |
+|------|------|------|------|-----------|-----------|--------|
+| 1 | DeepSeek V4 Flash | DeepSeek | 对话 | ¥1.02 | ¥2.03 | 1000000 |
+| 2 | GLM-4.6V-FlashX | Zhipu AI | 视觉 | ¥0.15 | ¥1.50 | 8000 |
+| 3 | doubao-seed-code | ByteDance | 代码 | ¥1.20 | ¥2.40 | 32000 |
+| 4 | deepseek-r1 | ByteDance | 推理 | ¥4.00 | ¥8.00 | 32000 |
+`
+
+func TestExtractDigestCardsBuildsFiveCategories(t *testing.T) {
+	report, err := parseDailyReport([]byte(sampleDailyReport))
+	if err != nil {
+		t.Fatalf("parseDailyReport returned error: %v", err)
+	}
+
+	cards := buildDigestCards(report)
+	if len(cards) != 5 {
+		t.Fatalf("expected 5 digest cards, got %d", len(cards))
+	}
+
+	if cards[0].Slug != "code" {
+		t.Fatalf("expected first card slug code, got %q", cards[0].Slug)
+	}
+	if len(cards[0].BulletLines) == 0 {
+		t.Fatal("expected code card to contain bullet lines")
+	}
+	if cards[4].Slug != "global" {
+		t.Fatalf("expected last card slug global, got %q", cards[4].Slug)
+	}
+}
+
+func TestBuildNarrationAudioProducesWAV(t *testing.T) {
+	audio, err := buildNarrationAudio([]DigestCard{
+		{Slug: "code", Narration: "Code digest update"},
+		{Slug: "global", Narration: "Global digest update"},
+	})
+	if err != nil {
+		t.Fatalf("buildNarrationAudio returned error: %v", err)
+	}
+
+	if len(audio) < 44 {
+		t.Fatalf("expected wav payload, got %d bytes", len(audio))
+	}
+	if !bytes.HasPrefix(audio, []byte("RIFF")) {
+		t.Fatalf("expected RIFF header, got %q", audio[:4])
+	}
+}

scripts/import_bytedance_data.go

@@ -0,0 +1,162 @@
+//go:build llm_script
+
+package main
+
+import (
+	"database/sql"
+	"encoding/json"
+	"log"
+	"os"
+
+	_ "github.com/lib/pq"
+)
+
+type ModelPricing struct {
+	ModelID         string
+	ModelName       string
+	ProviderName    string
+	ProviderCountry string
+	OperatorName    string
+	OperatorType    string
+	Region          string
+	Currency        string
+	InputPrice      float64
+	OutputPrice     float64
+	ContextLength   int
+	IsFree          bool
+	SourceURL       string
+	Modality        string
+}
+
+func main() {
+	dsn := os.Getenv("DATABASE_URL")
+	if dsn == "" {
+		dsn = "postgres://long@/llm_intelligence?host=/var/run/postgresql"
+	}
+
+	db, err := sql.Open("postgres", dsn)
+	if err != nil {
+		log.Fatal(err)
+	}
+	defer db.Close()
+
+	// Read raw data
+	data, err := os.ReadFile("/tmp/bytedance_raw.json")
+	if err != nil {
+		log.Fatal("Failed to read raw data:", err)
+	}
+
+	var raw struct {
+		Bytedance []struct {
+			Model         string  `json:"model"`
+			InputPrice    float64 `json:"inputPrice"`
+			OutputPrice   float64 `json:"outputPrice"`
+			ContextLength int     `json:"contextLength"`
+			Operator      string  `json:"operator"`
+			Region        string  `json:"region"`
+			Currency      string  `json:"currency"`
+		} `json:"bytedance"`
+	}
+
+	if err := json.Unmarshal(data, &raw); err != nil {
+		log.Fatal("Failed to parse raw data:", err)
+	}
+
+	log.Printf("Importing %d ByteDance models...", len(raw.Bytedance))
+	batchID := "manual-seed"
+
+	for _, b := range raw.Bytedance {
+		p := ModelPricing{
+			ModelID:         "bytedance-" + b.Model,
+			ModelName:       b.Model,
+			ProviderName:    "ByteDance",
+			ProviderCountry: "CN",
+			OperatorName:    "ByteDance Volcano",
+			OperatorType:    "official",
+			Region:          "CN",
+			Currency:        "CNY",
+			InputPrice:      b.InputPrice,
+			OutputPrice:     b.OutputPrice,
+			ContextLength:   b.ContextLength,
+			IsFree:          b.InputPrice == 0,
+			SourceURL:       "https://www.volcengine.com/docs/82379/1099320",
+			Modality:        "text",
+		}
+
+		// Find or create provider
+		var providerID int64
+		err := db.QueryRow("SELECT id FROM model_provider WHERE name = $1", p.ProviderName).Scan(&providerID)
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				"INSERT INTO model_provider (name, country, website, status) VALUES ($1, $2, $3, 'active') RETURNING id",
+				p.ProviderName, p.ProviderCountry, "",
+			).Scan(&providerID)
+		}
+		if err != nil {
+			log.Printf("Provider error: %v", err)
+			continue
+		}
+
+		// Find or create operator
+		var operatorID int64
+		err = db.QueryRow("SELECT id FROM operator WHERE name = $1", p.OperatorName).Scan(&operatorID)
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				"INSERT INTO operator (name, country, status) VALUES ($1, $2, 'active') RETURNING id",
+				p.OperatorName, p.ProviderCountry,
+			).Scan(&operatorID)
+		}
+		if err != nil {
+			log.Printf("Operator error: %v", err)
+			continue
+		}
+
+		// Find or create model
+		var modelID int64
+		err = db.QueryRow("SELECT id FROM models WHERE external_id = $1", p.ModelID).Scan(&modelID)
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				`INSERT INTO models (external_id, name, provider_id, modality, context_length, status, source, batch_id) 
+				 VALUES ($1, $2, $3, $4, $5, 'active', $6, $7) RETURNING id`,
+				p.ModelID, p.ModelName, providerID, p.Modality, p.ContextLength, p.OperatorName, batchID,
+			).Scan(&modelID)
+		}
+		if err != nil {
+			log.Printf("Model error for %s: %v", p.ModelID, err)
+			continue
+		}
+
+		// Insert pricing
+		sourceType := p.OperatorType
+		freeQuota := ""
+		freeLimitations := "[]"
+		rateLimit := "{}"
+		if p.IsFree {
+			sourceType = "free_tier"
+			freeQuota = "Imported free-tier pricing entry"
+			freeLimitations = `["See source_url for current quota and policy"]`
+		}
+		_, err = db.Exec(
+			`INSERT INTO region_pricing 
+			 (model_id, operator_id, region, currency, input_price_per_mtok, output_price_per_mtok, is_free, effective_date, source_url, source_type, free_quota, free_limitations, rate_limit)
+			 VALUES ($1, $2, $3, $4, $5, $6, $7, CURRENT_DATE, $8, $9, $10, $11, $12)
+			 ON CONFLICT (model_id, operator_id, region, currency, effective_date) 
+			 DO UPDATE SET input_price_per_mtok = EXCLUDED.input_price_per_mtok,
+			               output_price_per_mtok = EXCLUDED.output_price_per_mtok,
+			               is_free = EXCLUDED.is_free,
+			               source_type = EXCLUDED.source_type,
+			               free_quota = EXCLUDED.free_quota,
+			               free_limitations = EXCLUDED.free_limitations,
+			               rate_limit = EXCLUDED.rate_limit,
+			               updated_at = CURRENT_TIMESTAMP`,
+			modelID, operatorID, p.Region, p.Currency, p.InputPrice, p.OutputPrice, p.IsFree, p.SourceURL,
+			sourceType, freeQuota, freeLimitations, rateLimit,
+		)
+		if err != nil {
+			log.Printf("Pricing error for %s: %v", p.ModelID, err)
+			continue
+		}
+	}
+
+	log.Printf("Successfully imported %d ByteDance models", len(raw.Bytedance))
+}

scripts/import_phase2_data.go

@@ -0,0 +1,235 @@
+//go:build llm_script
+
+package main
+
+import (
+	"database/sql"
+	"encoding/json"
+	"fmt"
+	"log"
+	"os"
+	"strings"
+
+	_ "github.com/lib/pq"
+)
+
+type RawData struct {
+	Zhipu []struct {
+		Model       string `json:"model"`
+		Context     string `json:"context"`
+		InputPrice  string `json:"inputPrice"`
+		OutputPrice string `json:"outputPrice"`
+		Operator    string `json:"operator"`
+		Region      string `json:"region"`
+		Currency    string `json:"currency"`
+	} `json:"zhipu"`
+	Baidu []struct {
+		Model       string   `json:"model"`
+		Type        string   `json:"type"`
+		InputPrice  *float64 `json:"inputPrice"`
+		OutputPrice *float64 `json:"outputPrice"`
+		Operator    string   `json:"operator"`
+		Region      string   `json:"region"`
+		Currency    string   `json:"currency"`
+	} `json:"baidu"`
+}
+
+type ModelPricing struct {
+	ModelID         string
+	ModelName       string
+	ProviderName    string
+	ProviderCountry string
+	OperatorName    string
+	OperatorType    string
+	Region          string
+	Currency        string
+	InputPrice      float64
+	OutputPrice     float64
+	ContextLength   int
+	IsFree          bool
+	SourceURL       string
+	Modality        string
+	SceneTags       []string
+}
+
+func parseZhipuPrice(s string) float64 {
+	// Extract price from strings like "6元", "免费", "限时免费"
+	if strings.Contains(s, "免费") {
+		return 0
+	}
+	var f float64
+	fmt.Sscanf(s, "%f", &f)
+	return f
+}
+
+func extractContextLength(context string) int {
+	if strings.Contains(context, "1M") || strings.Contains(context, "1000K") {
+		return 1000000
+	}
+	if strings.Contains(context, "200K") {
+		return 200000
+	}
+	if strings.Contains(context, "128K") {
+		return 128000
+	}
+	if strings.Contains(context, "32K") {
+		return 32000
+	}
+	if strings.Contains(context, "8K") {
+		return 8000
+	}
+	if strings.Contains(context, "262144") || strings.Contains(context, "256K") {
+		return 262144
+	}
+	if strings.Contains(context, "8192") {
+		return 8192
+	}
+	return 0
+}
+
+func main() {
+	dsn := os.Getenv("DATABASE_URL")
+	if dsn == "" {
+		dsn = "postgres://long@/llm_intelligence?host=/var/run/postgresql"
+	}
+
+	db, err := sql.Open("postgres", dsn)
+	if err != nil {
+		log.Fatal(err)
+	}
+	defer db.Close()
+
+	// Read raw data
+	data, err := os.ReadFile("/tmp/phase2_raw_data.json")
+	if err != nil {
+		log.Fatal("Failed to read raw data:", err)
+	}
+
+	var raw RawData
+	if err := json.Unmarshal(data, &raw); err != nil {
+		log.Fatal("Failed to parse raw data:", err)
+	}
+
+	var prices []ModelPricing
+	batchID := "manual-seed"
+
+	// Process Baidu data
+	modelPrices := make(map[string]map[string]float64) // model -> type -> price
+	for _, b := range raw.Baidu {
+		if modelPrices[b.Model] == nil {
+			modelPrices[b.Model] = make(map[string]float64)
+		}
+		if b.InputPrice != nil {
+			if strings.Contains(b.Type, "输入") {
+				modelPrices[b.Model]["input"] = *b.InputPrice * 1000000 // Convert to per 1M
+			}
+			if strings.Contains(b.Type, "输出") {
+				modelPrices[b.Model]["output"] = *b.InputPrice * 1000000
+			}
+		}
+		if b.OutputPrice != nil {
+			if strings.Contains(b.Type, "输出") {
+				modelPrices[b.Model]["output"] = *b.OutputPrice * 1000000
+			}
+		}
+	}
+
+	for model, pricesMap := range modelPrices {
+		prices = append(prices, ModelPricing{
+			ModelID:         "baidu-" + strings.ToLower(strings.ReplaceAll(model, " ", "-")),
+			ModelName:       model,
+			ProviderName:    "Baidu",
+			ProviderCountry: "CN",
+			OperatorName:    "Baidu Qianfan",
+			OperatorType:    "official",
+			Region:          "CN",
+			Currency:        "CNY",
+			InputPrice:      pricesMap["input"],
+			OutputPrice:     pricesMap["output"],
+			IsFree:          pricesMap["input"] == 0 && pricesMap["output"] == 0,
+			SourceURL:       "https://cloud.baidu.com/doc/qianfan/s/wmh4sv6ya",
+			Modality:        "text",
+		})
+	}
+
+	log.Printf("Parsed %d unique models from Baidu", len(prices))
+
+	// Save to database
+	for _, p := range prices {
+		// Find or create provider
+		var providerID int64
+		err := db.QueryRow("SELECT id FROM model_provider WHERE name = $1", p.ProviderName).Scan(&providerID)
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				"INSERT INTO model_provider (name, country, website, status) VALUES ($1, $2, $3, 'active') RETURNING id",
+				p.ProviderName, p.ProviderCountry, "",
+			).Scan(&providerID)
+		}
+		if err != nil {
+			log.Printf("Provider error: %v", err)
+			continue
+		}
+
+		// Find or create operator
+		var operatorID int64
+		err = db.QueryRow("SELECT id FROM operator WHERE name = $1", p.OperatorName).Scan(&operatorID)
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				"INSERT INTO operator (name, country, status) VALUES ($1, $2, 'active') RETURNING id",
+				p.OperatorName, p.ProviderCountry,
+			).Scan(&operatorID)
+		}
+		if err != nil {
+			log.Printf("Operator error: %v", err)
+			continue
+		}
+
+		// Find or create model
+		var modelID int64
+		err = db.QueryRow("SELECT id FROM models WHERE external_id = $1", p.ModelID).Scan(&modelID)
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				`INSERT INTO models (external_id, name, provider_id, modality, context_length, status, source, batch_id) 
+				 VALUES ($1, $2, $3, $4, $5, 'active', $6, $7) RETURNING id`,
+				p.ModelID, p.ModelName, providerID, p.Modality, p.ContextLength, p.OperatorName, batchID,
+			).Scan(&modelID)
+		}
+		if err != nil {
+			log.Printf("Model error: %v", err)
+			continue
+		}
+
+		// Insert pricing
+		sourceType := p.OperatorType
+		freeQuota := ""
+		freeLimitations := "[]"
+		rateLimit := "{}"
+		if p.IsFree {
+			sourceType = "free_tier"
+			freeQuota = "Imported free-tier pricing entry"
+			freeLimitations = `["See source_url for current quota and policy"]`
+		}
+		_, err = db.Exec(
+			`INSERT INTO region_pricing 
+			 (model_id, operator_id, region, currency, input_price_per_mtok, output_price_per_mtok, is_free, effective_date, source_url, source_type, free_quota, free_limitations, rate_limit)
+			 VALUES ($1, $2, $3, $4, $5, $6, $7, CURRENT_DATE, $8, $9, $10, $11, $12)
+			 ON CONFLICT (model_id, operator_id, region, currency, effective_date) 
+			 DO UPDATE SET input_price_per_mtok = EXCLUDED.input_price_per_mtok,
+			               output_price_per_mtok = EXCLUDED.output_price_per_mtok,
+			               is_free = EXCLUDED.is_free,
+			               source_type = EXCLUDED.source_type,
+			               free_quota = EXCLUDED.free_quota,
+			               free_limitations = EXCLUDED.free_limitations,
+			               rate_limit = EXCLUDED.rate_limit,
+			               updated_at = CURRENT_TIMESTAMP`,
+			modelID, operatorID, p.Region, p.Currency, p.InputPrice, p.OutputPrice, p.IsFree, p.SourceURL,
+			sourceType, freeQuota, freeLimitations, rateLimit,
+		)
+		if err != nil {
+			log.Printf("Pricing error for %s: %v", p.ModelID, err)
+			continue
+		}
+	}
+
+	log.Printf("Successfully imported %d models into database", len(prices))
+}

scripts/import_tencent_subscription.go

@@ -0,0 +1,420 @@
+//go:build llm_script
+
+package main
+
+import (
+	"database/sql"
+	"encoding/json"
+	"flag"
+	"fmt"
+	"io"
+	"net/http"
+	"os"
+	"regexp"
+	"strconv"
+	"strings"
+	"time"
+
+	_ "github.com/lib/pq"
+)
+
+type importTencentSubscriptionConfig struct {
+	URL     string
+	Fixture string
+	DryRun  bool
+	Timeout time.Duration
+}
+
+type subscriptionPlanRow struct {
+	ProviderName   string
+	ProviderCN     string
+	ProviderCountry string
+	OperatorName   string
+	OperatorCN     string
+	OperatorCountry string
+	OperatorType   string
+	PlanFamily     string
+	PlanCode       string
+	PlanName       string
+	Tier           string
+	BillingCycle   string
+	Currency       string
+	ListPrice      float64
+	PriceUnit      string
+	QuotaValue     int64
+	QuotaUnit      string
+	ContextWindow  int
+	PlanScope      string
+	ModelScope     string
+	SourceURL      string
+	PublishedAt    string
+	EffectiveDate  string
+	Notes          string
+}
+
+func main() {
+	loadImportProjectEnv()
+
+	var rawURL string
+	var fixturePath string
+	var dryRun bool
+	var timeoutSeconds int
+
+	flag.StringVar(&rawURL, "url", defaultTencentCatalogURL, "腾讯云公开目录 URL")
+	flag.StringVar(&fixturePath, "fixture", "", "本地 HTML/Text 样例文件，优先用于离线导入")
+	flag.BoolVar(&dryRun, "dry-run", false, "仅解析并打印摘要，不写入数据库")
+	flag.IntVar(&timeoutSeconds, "timeout", int(defaultTencentCatalogTimeout/time.Second), "请求超时（秒）")
+	flag.Parse()
+
+	cfg := importTencentSubscriptionConfig{
+		URL:     rawURL,
+		Fixture: fixturePath,
+		DryRun:  dryRun,
+		Timeout: time.Duration(timeoutSeconds) * time.Second,
+	}
+
+	var db *sql.DB
+	var err error
+	if !cfg.DryRun {
+		dsn := os.Getenv("DATABASE_URL")
+		if dsn == "" {
+			dsn = "postgres://long@/llm_intelligence?host=/var/run/postgresql"
+		}
+		db, err = sql.Open("postgres", dsn)
+		if err != nil {
+			fmt.Fprintf(os.Stderr, "open db: %v\n", err)
+			os.Exit(1)
+		}
+		defer db.Close()
+	}
+
+	if err := runTencentSubscriptionImport(cfg, db, os.Stdout); err != nil {
+		fmt.Fprintf(os.Stderr, "import_tencent_subscription: %v\n", err)
+		os.Exit(1)
+	}
+}
+
+func loadImportProjectEnv() {
+	for _, path := range []string{".env.local", ".env"} {
+		loadImportEnvFile(path)
+	}
+}
+
+func loadImportEnvFile(path string) {
+	data, err := os.ReadFile(path)
+	if err != nil {
+		return
+	}
+
+	for _, line := range strings.Split(string(data), "\n") {
+		line = strings.TrimSpace(line)
+		if line == "" || strings.HasPrefix(line, "#") {
+			continue
+		}
+		key, value, ok := strings.Cut(line, "=")
+		if !ok {
+			continue
+		}
+		key = strings.TrimSpace(key)
+		value = strings.Trim(strings.TrimSpace(value), `"'`)
+		if key == "" {
+			continue
+		}
+		if _, exists := os.LookupEnv(key); exists {
+			continue
+		}
+		_ = os.Setenv(key, value)
+	}
+}
+
+func runTencentSubscriptionImport(cfg importTencentSubscriptionConfig, db *sql.DB, out io.Writer) error {
+	raw, err := fetchTencentCatalogContent(fetchTencentCatalogConfig{
+		URL:     cfg.URL,
+		DryRun:  cfg.DryRun,
+		Timeout: cfg.Timeout,
+		Fixture: cfg.Fixture,
+	}, &http.Client{Timeout: cfg.Timeout})
+	if err != nil {
+		return err
+	}
+
+	catalog, err := parseTencentCatalog(raw)
+	if err != nil {
+		return err
+	}
+
+	plans := buildSubscriptionPlans(catalog, cfg.URL)
+	if cfg.DryRun {
+		_, err = fmt.Fprintf(
+			out,
+			"source=tencent-subscription-import updated_at=%s plans=%d provider=%s operator=%s dry_run=true\n",
+			catalog.UpdatedAt,
+			len(plans),
+			plans[0].ProviderName,
+			plans[0].OperatorName,
+		)
+		return err
+	}
+
+	if db == nil {
+		return fmt.Errorf("db is required when dry-run=false")
+	}
+
+	if err := upsertSubscriptionPlans(db, plans); err != nil {
+		return err
+	}
+
+	var tableRows int
+	if err := db.QueryRow(`SELECT COUNT(*) FROM subscription_plan`).Scan(&tableRows); err != nil {
+		return fmt.Errorf("count subscription_plan: %w", err)
+	}
+
+	summary := fmt.Sprintf(
+		"source=tencent-subscription-import updated_at=%s plans=%d provider=%s operator=%s table_rows=%d dry_run=false\n",
+		catalog.UpdatedAt,
+		len(plans),
+		plans[0].ProviderName,
+		plans[0].OperatorName,
+		tableRows,
+	)
+	if _, err := io.WriteString(out, summary); err != nil {
+		return err
+	}
+	if err := writeTencentImportSummary(summary); err != nil {
+		return err
+	}
+	return nil
+}
+
+func buildSubscriptionPlans(catalog tencentCatalog, sourceURL string) []subscriptionPlanRow {
+	modelsBySeries := make(map[string][]tencentModel)
+	for _, model := range catalog.Models {
+		modelsBySeries[model.Series] = append(modelsBySeries[model.Series], model)
+	}
+
+	plans := make([]subscriptionPlanRow, 0, len(catalog.Plans))
+	for _, plan := range catalog.Plans {
+		models := modelsBySeries[plan.Series]
+		plans = append(plans, subscriptionPlanRow{
+			ProviderName:    "Tencent",
+			ProviderCN:      "腾讯",
+			ProviderCountry: "CN",
+			OperatorName:    "Tencent Cloud",
+			OperatorCN:      "腾讯云",
+			OperatorCountry: "CN",
+			OperatorType:    "cloud",
+			PlanFamily:      inferPlanFamily(plan.Series),
+			PlanCode:        slugifyPlanCode(plan.Series, plan.Tier),
+			PlanName:        fmt.Sprintf("%s %s", plan.Series, plan.Tier),
+			Tier:            plan.Tier,
+			BillingCycle:    normalizeBillingCycle(plan.BillingCycle),
+			Currency:        "CNY",
+			ListPrice:       parsePlanPrice(plan.Price),
+			PriceUnit:       "CNY/month",
+			QuotaValue:      parseQuotaValue(plan.Quota),
+			QuotaUnit:       "tokens/month",
+			ContextWindow:   maxContextWindow(models),
+			PlanScope:       plan.Series,
+			ModelScope:      encodeModelScope(models),
+			SourceURL:       sourceURL,
+			PublishedAt:     catalog.UpdatedAt,
+			EffectiveDate:   extractEffectiveDate(catalog.UpdatedAt),
+			Notes:           strings.TrimSpace(plan.Scene),
+		})
+	}
+	return plans
+}
+
+func inferPlanFamily(series string) string {
+	lower := strings.ToLower(series)
+	if strings.Contains(lower, "coding plan") {
+		return "coding_plan"
+	}
+	return "token_plan"
+}
+
+func slugifyPlanCode(series string, tier string) string {
+	seriesCode := strings.TrimSpace(series)
+	switch seriesCode {
+	case "通用 Token Plan":
+		seriesCode = "token-plan"
+	case "Hy Token Plan":
+		seriesCode = "hy-token-plan"
+	}
+
+	raw := strings.ToLower(strings.TrimSpace(seriesCode + "-" + tier))
+	replacer := strings.NewReplacer(" ", "-", "/", "-", "_", "-", ".", "-", "（", "", "）", "", "(", "", ")", "", ":", "-", "--", "-")
+	raw = replacer.Replace(raw)
+	raw = strings.Trim(raw, "-")
+	return raw
+}
+
+func normalizeBillingCycle(raw string) string {
+	if strings.Contains(raw, "月") {
+		return "monthly"
+	}
+	return strings.TrimSpace(raw)
+}
+
+func parsePlanPrice(raw string) float64 {
+	value := strings.TrimSpace(strings.TrimSuffix(raw, "元/月"))
+	f, _ := strconv.ParseFloat(value, 64)
+	return f
+}
+
+func parseQuotaValue(raw string) int64 {
+	quotaPattern := regexp.MustCompile(`([\d.]+)\s*([万亿]?)\s*Tokens`)
+	matches := quotaPattern.FindStringSubmatch(raw)
+	if len(matches) != 3 {
+		return 0
+	}
+	base, _ := strconv.ParseFloat(matches[1], 64)
+	switch matches[2] {
+	case "万":
+		base *= 10000
+	case "亿":
+		base *= 100000000
+	}
+	return int64(base)
+}
+
+func maxContextWindow(models []tencentModel) int {
+	max := 0
+	for _, model := range models {
+		if model.ContextLength > max {
+			max = model.ContextLength
+		}
+	}
+	return max
+}
+
+func encodeModelScope(models []tencentModel) string {
+	ids := make([]string, 0, len(models))
+	for _, model := range models {
+		ids = append(ids, model.ModelID)
+	}
+	data, _ := json.Marshal(ids)
+	return string(data)
+}
+
+func extractEffectiveDate(updatedAt string) string {
+	if len(updatedAt) >= len("2006-01-02") {
+		return updatedAt[:10]
+	}
+	return time.Now().Format("2006-01-02")
+}
+
+func upsertSubscriptionPlans(db *sql.DB, plans []subscriptionPlanRow) error {
+	providerID, err := ensureModelProvider(db, plans[0])
+	if err != nil {
+		return err
+	}
+	operatorID, err := ensureOperator(db, plans[0])
+	if err != nil {
+		return err
+	}
+
+	for _, plan := range plans {
+		publishedAt, err := time.Parse("2006-01-02 15:04:05", plan.PublishedAt)
+		if err != nil {
+			return fmt.Errorf("parse published_at for %s: %w", plan.PlanCode, err)
+		}
+		effectiveDate, err := time.Parse("2006-01-02", plan.EffectiveDate)
+		if err != nil {
+			return fmt.Errorf("parse effective_date for %s: %w", plan.PlanCode, err)
+		}
+
+		_, err = db.Exec(
+			`INSERT INTO subscription_plan (
+				provider_id, operator_id, plan_family, plan_code, plan_name, tier,
+				billing_cycle, currency, list_price, price_unit, quota_value, quota_unit,
+				context_window, plan_scope, model_scope, source_url, published_at, effective_date, notes
+			) VALUES (
+				$1, $2, $3, $4, $5, $6,
+				$7, $8, $9, $10, $11, $12,
+				$13, $14, $15, $16, $17, $18, $19
+			)
+			ON CONFLICT (provider_id, plan_code, effective_date)
+			DO UPDATE SET
+				operator_id = EXCLUDED.operator_id,
+				plan_family = EXCLUDED.plan_family,
+				plan_name = EXCLUDED.plan_name,
+				tier = EXCLUDED.tier,
+				billing_cycle = EXCLUDED.billing_cycle,
+				currency = EXCLUDED.currency,
+				list_price = EXCLUDED.list_price,
+				price_unit = EXCLUDED.price_unit,
+				quota_value = EXCLUDED.quota_value,
+				quota_unit = EXCLUDED.quota_unit,
+				context_window = EXCLUDED.context_window,
+				plan_scope = EXCLUDED.plan_scope,
+				model_scope = EXCLUDED.model_scope,
+				source_url = EXCLUDED.source_url,
+				published_at = EXCLUDED.published_at,
+				notes = EXCLUDED.notes,
+				updated_at = CURRENT_TIMESTAMP`,
+			providerID, operatorID, plan.PlanFamily, plan.PlanCode, plan.PlanName, plan.Tier,
+			plan.BillingCycle, plan.Currency, plan.ListPrice, plan.PriceUnit, plan.QuotaValue, plan.QuotaUnit,
+			nullIfZero(plan.ContextWindow), plan.PlanScope, plan.ModelScope, plan.SourceURL, publishedAt, effectiveDate, plan.Notes,
+		)
+		if err != nil {
+			return fmt.Errorf("upsert subscription_plan %s: %w", plan.PlanCode, err)
+		}
+	}
+
+	return nil
+}
+
+func ensureModelProvider(db *sql.DB, plan subscriptionPlanRow) (int64, error) {
+	var providerID int64
+	err := db.QueryRow(`SELECT id FROM model_provider WHERE name = $1`, plan.ProviderName).Scan(&providerID)
+	if err == nil {
+		return providerID, nil
+	}
+	if err != sql.ErrNoRows {
+		return 0, err
+	}
+
+	err = db.QueryRow(
+		`INSERT INTO model_provider (name, name_cn, country, website, status)
+		 VALUES ($1, $2, $3, $4, 'active')
+		 RETURNING id`,
+		plan.ProviderName, plan.ProviderCN, plan.ProviderCountry, "https://cloud.tencent.com",
+	).Scan(&providerID)
+	return providerID, err
+}
+
+func ensureOperator(db *sql.DB, plan subscriptionPlanRow) (int64, error) {
+	var operatorID int64
+	err := db.QueryRow(`SELECT id FROM operator WHERE name = $1`, plan.OperatorName).Scan(&operatorID)
+	if err == nil {
+		return operatorID, nil
+	}
+	if err != sql.ErrNoRows {
+		return 0, err
+	}
+
+	err = db.QueryRow(
+		`INSERT INTO operator (name, name_cn, country, website, description, status, type)
+		 VALUES ($1, $2, $3, $4, $5, 'active', $6)
+		 RETURNING id`,
+		plan.OperatorName, plan.OperatorCN, plan.OperatorCountry, "https://cloud.tencent.com",
+		"Tencent Cloud subscription plans", plan.OperatorType,
+	).Scan(&operatorID)
+	return operatorID, err
+}
+
+func nullIfZero(value int) any {
+	if value == 0 {
+		return nil
+	}
+	return value
+}
+
+func writeTencentImportSummary(summary string) error {
+	const summaryPath = "reports/verification/tencent_subscription_import_latest.txt"
+	if err := os.MkdirAll("reports/verification", 0755); err != nil {
+		return err
+	}
+	return os.WriteFile(summaryPath, []byte(summary), 0644)
+}

scripts/import_tencent_subscription_test.go

@@ -0,0 +1,90 @@
+//go:build llm_script
+
+package main
+
+import (
+	"bytes"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+func TestBuildSubscriptionPlansFromCatalog(t *testing.T) {
+	raw, err := os.ReadFile(filepath.Join("testdata", "tencent_token_plan_sample.txt"))
+	if err != nil {
+		t.Fatalf("读取样例失败: %v", err)
+	}
+
+	catalog, err := parseTencentCatalog(string(raw))
+	if err != nil {
+		t.Fatalf("parseTencentCatalog 失败: %v", err)
+	}
+
+	plans := buildSubscriptionPlans(catalog, defaultTencentCatalogURL)
+	if len(plans) != 8 {
+		t.Fatalf("期望 8 条套餐记录，实际 %d", len(plans))
+	}
+
+	first := plans[0]
+	if first.ProviderName != "Tencent" {
+		t.Fatalf("provider 错误: %q", first.ProviderName)
+	}
+	if first.OperatorName != "Tencent Cloud" {
+		t.Fatalf("operator 错误: %q", first.OperatorName)
+	}
+	if first.PlanFamily != "token_plan" {
+		t.Fatalf("plan family 错误: %q", first.PlanFamily)
+	}
+	if first.PlanCode != "token-plan-lite" {
+		t.Fatalf("plan code 错误: %q", first.PlanCode)
+	}
+	if first.ListPrice != 39 {
+		t.Fatalf("list price 错误: %v", first.ListPrice)
+	}
+	if first.QuotaValue != 35000000 {
+		t.Fatalf("quota value 错误: %d", first.QuotaValue)
+	}
+	if !strings.Contains(first.ModelScope, "\"glm-5\"") {
+		t.Fatalf("model_scope 缺少 glm-5: %q", first.ModelScope)
+	}
+	if first.PublishedAt != "2026-04-27 17:18:02" {
+		t.Fatalf("published_at 错误: %q", first.PublishedAt)
+	}
+
+	last := plans[len(plans)-1]
+	if last.PlanFamily != "token_plan" {
+		t.Fatalf("Hy Token Plan family 错误: %q", last.PlanFamily)
+	}
+	if last.PlanCode != "hy-token-plan-max" {
+		t.Fatalf("Hy Token Plan code 错误: %q", last.PlanCode)
+	}
+	if last.ContextWindow != 262144 {
+		t.Fatalf("Hy Token Plan context 错误: %d", last.ContextWindow)
+	}
+}
+
+func TestRunTencentSubscriptionImportDryRunPrintsSummary(t *testing.T) {
+	var out bytes.Buffer
+	err := runTencentSubscriptionImport(importTencentSubscriptionConfig{
+		Fixture: filepath.Join("testdata", "tencent_token_plan_sample.txt"),
+		DryRun:  true,
+		URL:     defaultTencentCatalogURL,
+	}, nil, &out)
+	if err != nil {
+		t.Fatalf("runTencentSubscriptionImport 失败: %v", err)
+	}
+
+	output := out.String()
+	for _, want := range []string{
+		"source=tencent-subscription-import",
+		"plans=8",
+		"provider=Tencent",
+		"operator=Tencent Cloud",
+		"dry_run=true",
+	} {
+		if !strings.Contains(output, want) {
+			t.Fatalf("输出缺少 %q，实际: %q", want, output)
+		}
+	}
+}

scripts/import_zhipu_data.go

@@ -0,0 +1,192 @@
+//go:build llm_script
+
+package main
+
+import (
+	"database/sql"
+	"log"
+	"os"
+
+	_ "github.com/lib/pq"
+)
+
+type ModelPricing struct {
+	ModelID         string
+	ModelName       string
+	ProviderName    string
+	ProviderCountry string
+	OperatorName    string
+	OperatorType    string
+	Region          string
+	Currency        string
+	InputPrice      float64
+	OutputPrice     float64
+	ContextLength   int
+	IsFree          bool
+	SourceURL       string
+	Modality        string
+	SceneTags       []string
+}
+
+func main() {
+	dsn := os.Getenv("DATABASE_URL")
+	if dsn == "" {
+		dsn = "postgres://long@/llm_intelligence?host=/var/run/postgresql"
+	}
+
+	db, err := sql.Open("postgres", dsn)
+	if err != nil {
+		log.Fatal(err)
+	}
+	defer db.Close()
+
+	// 智谱AI定价数据（从第一次无头浏览器抓取中提取）
+	prices := []ModelPricing{
+		// GLM-5.1系列
+		{ModelID: "glm-5.1", ModelName: "GLM-5.1", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 6.0, OutputPrice: 24.0, ContextLength: 32000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "推理", "代码"}},
+		{ModelID: "glm-5.1-32k", ModelName: "GLM-5.1 (32K+)", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 8.0, OutputPrice: 28.0, ContextLength: 200000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "推理", "代码"}},
+
+		// GLM-5-Turbo
+		{ModelID: "glm-5-turbo", ModelName: "GLM-5-Turbo", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 5.0, OutputPrice: 22.0, ContextLength: 32000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "推理"}},
+		{ModelID: "glm-5-turbo-32k", ModelName: "GLM-5-Turbo (32K+)", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 7.0, OutputPrice: 26.0, ContextLength: 200000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "推理"}},
+
+		// GLM-5
+		{ModelID: "glm-5", ModelName: "GLM-5", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 4.0, OutputPrice: 18.0, ContextLength: 32000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "推理"}},
+		{ModelID: "glm-5-32k", ModelName: "GLM-5 (32K+)", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 6.0, OutputPrice: 22.0, ContextLength: 200000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "推理"}},
+
+		// GLM-4.7
+		{ModelID: "glm-4.7", ModelName: "GLM-4.7", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 2.0, OutputPrice: 8.0, ContextLength: 32000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "推理", "代码"}},
+		{ModelID: "glm-4.7-32k", ModelName: "GLM-4.7 (32K+)", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 4.0, OutputPrice: 16.0, ContextLength: 200000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "推理", "代码"}},
+
+		// GLM-4.5-Air
+		{ModelID: "glm-4.5-air", ModelName: "GLM-4.5-Air", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 0.8, OutputPrice: 2.0, ContextLength: 32000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话"}},
+		{ModelID: "glm-4.5-air-32k", ModelName: "GLM-4.5-Air (32K+)", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 1.2, OutputPrice: 8.0, ContextLength: 128000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话"}},
+
+		// GLM-4.7-FlashX
+		{ModelID: "glm-4.7-flashx", ModelName: "GLM-4.7-FlashX", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 0.5, OutputPrice: 3.0, ContextLength: 200000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话"}},
+
+		// GLM-4.7-Flash (Free)
+		{ModelID: "glm-4.7-flash", ModelName: "GLM-4.7-Flash", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 0, OutputPrice: 0, ContextLength: 200000, IsFree: true, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话"}},
+
+		// GLM-4.6V (Vision)
+		{ModelID: "glm-4.6v", ModelName: "GLM-4.6V", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 2.0, OutputPrice: 6.0, ContextLength: 8000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "multimodal", SceneTags: []string{"视觉", "对话"}},
+
+		// GLM-4.6V-FlashX
+		{ModelID: "glm-4.6v-flashx", ModelName: "GLM-4.6V-FlashX", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 0.15, OutputPrice: 1.5, ContextLength: 8000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "multimodal", SceneTags: []string{"视觉", "对话"}},
+
+		// GLM-4.5V
+		{ModelID: "glm-4.5v", ModelName: "GLM-4.5V", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 2.0, OutputPrice: 6.0, ContextLength: 32000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "multimodal", SceneTags: []string{"视觉", "对话"}},
+
+		// GLM-4系列 (Legacy)
+		{ModelID: "glm-4-0520", ModelName: "GLM-4-0520", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 100.0, OutputPrice: 50.0, ContextLength: 128000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "推理"}},
+		{ModelID: "glm-4-air", ModelName: "GLM-4-Air", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 0.5, OutputPrice: 0.25, ContextLength: 128000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话"}},
+		{ModelID: "glm-4-airx", ModelName: "GLM-4-AirX", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 10.0, OutputPrice: 10.0, ContextLength: 8000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "极速"}},
+		{ModelID: "glm-4-long", ModelName: "GLM-4-Long", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 1.0, OutputPrice: 0.5, ContextLength: 1000000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话", "长文本"}},
+
+		// GLM-4V (Vision Legacy)
+		{ModelID: "glm-4v-plus", ModelName: "GLM-4V-Plus", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 4.0, OutputPrice: 4.0, ContextLength: 8000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "multimodal", SceneTags: []string{"视觉", "对话"}},
+		{ModelID: "glm-4v", ModelName: "GLM-4V", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 50.0, OutputPrice: 50.0, ContextLength: 2000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "multimodal", SceneTags: []string{"视觉", "对话"}},
+
+		// ChatGLM3
+		{ModelID: "chatglm3-6b", ModelName: "ChatGLM3-6B", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 0, OutputPrice: 0, ContextLength: 8000, IsFree: true, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话"}},
+
+		// GLM-4-9B
+		{ModelID: "glm-4-9b", ModelName: "GLM-4-9B", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 0, OutputPrice: 0, ContextLength: 8000, IsFree: true, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "text", SceneTags: []string{"对话"}},
+
+		// GLM-Realtime
+		{ModelID: "glm-realtime-flash", ModelName: "GLM-Realtime-Flash", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 0.18, OutputPrice: 0.18, ContextLength: 8000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "audio", SceneTags: []string{"实时", "音视频"}},
+		{ModelID: "glm-realtime-air", ModelName: "GLM-Realtime-Air", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 0.3, OutputPrice: 0.3, ContextLength: 8000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "audio", SceneTags: []string{"实时", "音视频"}},
+
+		// GLM-TTS
+		{ModelID: "glm-tts", ModelName: "GLM-TTS", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 2.0, OutputPrice: 0, ContextLength: 8000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "audio", SceneTags: []string{"语音合成"}},
+		{ModelID: "glm-tts-clone", ModelName: "GLM-TTS-Clone", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 6.0, OutputPrice: 0, ContextLength: 8000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "audio", SceneTags: []string{"音色克隆"}},
+
+		// GLM-ASR
+		{ModelID: "glm-asr-2512", ModelName: "GLM-ASR-2512", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 16.0, OutputPrice: 0, ContextLength: 8000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "audio", SceneTags: []string{"语音识别"}},
+
+		// GLM-4-Voice
+		{ModelID: "glm-4-voice", ModelName: "GLM-4-Voice", ProviderName: "Zhipu AI", ProviderCountry: "CN", OperatorName: "Zhipu", OperatorType: "official", Region: "CN", Currency: "CNY", InputPrice: 80.0, OutputPrice: 80.0, ContextLength: 8000, SourceURL: "https://open.bigmodel.cn/pricing", Modality: "audio", SceneTags: []string{"语音模型"}},
+	}
+	batchID := "manual-seed"
+
+	log.Printf("Importing %d Zhipu AI models...", len(prices))
+
+	// Save to database
+	for _, p := range prices {
+		// Find or create provider
+		var providerID int64
+		err := db.QueryRow("SELECT id FROM model_provider WHERE name = $1", p.ProviderName).Scan(&providerID)
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				"INSERT INTO model_provider (name, country, website, status) VALUES ($1, $2, $3, 'active') RETURNING id",
+				p.ProviderName, p.ProviderCountry, "",
+			).Scan(&providerID)
+		}
+		if err != nil {
+			log.Printf("Provider error: %v", err)
+			continue
+		}
+
+		// Find or create operator
+		var operatorID int64
+		err = db.QueryRow("SELECT id FROM operator WHERE name = $1", p.OperatorName).Scan(&operatorID)
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				"INSERT INTO operator (name, country, status) VALUES ($1, $2, 'active') RETURNING id",
+				p.OperatorName, p.ProviderCountry,
+			).Scan(&operatorID)
+		}
+		if err != nil {
+			log.Printf("Operator error: %v", err)
+			continue
+		}
+
+		// Find or create model
+		var modelID int64
+		err = db.QueryRow("SELECT id FROM models WHERE external_id = $1", p.ModelID).Scan(&modelID)
+		if err == sql.ErrNoRows {
+			err = db.QueryRow(
+				`INSERT INTO models (external_id, name, provider_id, modality, context_length, status, source, batch_id) 
+				 VALUES ($1, $2, $3, $4, $5, 'active', $6, $7) RETURNING id`,
+				p.ModelID, p.ModelName, providerID, p.Modality, p.ContextLength, p.OperatorName, batchID,
+			).Scan(&modelID)
+		}
+		if err != nil {
+			log.Printf("Model error for %s: %v", p.ModelID, err)
+			continue
+		}
+
+		// Insert pricing
+		sourceType := p.OperatorType
+		freeQuota := ""
+		freeLimitations := "[]"
+		rateLimit := "{}"
+		if p.IsFree {
+			sourceType = "free_tier"
+			freeQuota = "Imported free-tier pricing entry"
+			freeLimitations = `["See source_url for current quota and policy"]`
+		}
+		_, err = db.Exec(
+			`INSERT INTO region_pricing 
+			 (model_id, operator_id, region, currency, input_price_per_mtok, output_price_per_mtok, is_free, effective_date, source_url, source_type, free_quota, free_limitations, rate_limit)
+			 VALUES ($1, $2, $3, $4, $5, $6, $7, CURRENT_DATE, $8, $9, $10, $11, $12)
+			 ON CONFLICT (model_id, operator_id, region, currency, effective_date) 
+			 DO UPDATE SET input_price_per_mtok = EXCLUDED.input_price_per_mtok,
+			               output_price_per_mtok = EXCLUDED.output_price_per_mtok,
+			               is_free = EXCLUDED.is_free,
+			               source_type = EXCLUDED.source_type,
+			               free_quota = EXCLUDED.free_quota,
+			               free_limitations = EXCLUDED.free_limitations,
+			               rate_limit = EXCLUDED.rate_limit,
+			               updated_at = CURRENT_TIMESTAMP`,
+			modelID, operatorID, p.Region, p.Currency, p.InputPrice, p.OutputPrice, p.IsFree, p.SourceURL,
+			sourceType, freeQuota, freeLimitations, rateLimit,
+		)
+		if err != nil {
+			log.Printf("Pricing error for %s: %v", p.ModelID, err)
+			continue
+		}
+	}
+
+	log.Printf("Successfully imported %d Zhipu AI models", len(prices))
+}

scripts/restore.sh

@@ -0,0 +1,38 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT_DIR"
+
+if [[ "${1:-}" != "--force" || -z "${2:-}" ]]; then
+  echo "用法: bash scripts/restore.sh --force <backup.sql|backup.sql.gz>" >&2
+  exit 1
+fi
+
+BACKUP_FILE="$2"
+if [[ ! -f "$BACKUP_FILE" ]]; then
+  echo "备份文件不存在: $BACKUP_FILE" >&2
+  exit 1
+fi
+
+if [[ -f ".env.local" ]]; then
+  # shellcheck disable=SC1091
+  source ".env.local"
+fi
+if [[ -f ".env" ]]; then
+  # shellcheck disable=SC1091
+  source ".env"
+fi
+
+DB_URL="${DATABASE_URL:-host=/var/run/postgresql dbname=llm_intelligence user=long sslmode=disable}"
+
+echo "开始恢复到目标数据库..."
+psql "$DB_URL" -v ON_ERROR_STOP=1 -c "DROP SCHEMA IF EXISTS public CASCADE; CREATE SCHEMA public;"
+
+if [[ "$BACKUP_FILE" == *.gz ]]; then
+  gzip -dc "$BACKUP_FILE" | psql "$DB_URL" -v ON_ERROR_STOP=1
+else
+  psql "$DB_URL" -v ON_ERROR_STOP=1 -f "$BACKUP_FILE"
+fi
+
+echo "恢复完成: $BACKUP_FILE"

scripts/review/OPENCLAW_MULTI_REVIEW_PROMPT.md

@@ -0,0 +1,77 @@
+# OpenClaw Multi Review Prompt
+
+目标：对 `llm-intelligence` 项目执行一次高频真实状态 review，并顺手沉淀 OpenClaw 能力优化项。
+
+执行要求：
+
+1. 只基于当前仓库真实状态做判断。
+2. 如果本轮要写 `TASKS.md` 或 `GOALS.md`，必须先执行预检守卫：
+   - `bash scripts/review/preflight_task_write_guard.sh llm-intelligence-review /home/long/project/llm-intelligence/TASKS.md`
+   - `bash scripts/review/preflight_task_write_guard.sh llm-intelligence-review /home/long/project/llm-intelligence/GOALS.md`
+   - 守卫失败时，立即停止写回，不得继续尝试 `edit` 或 `write`
+3. 必须先检查：
+   - `git status --short`
+   - 最近提交记录
+   - `TASKS.md`、`GOALS.md`、`OPENCLAW_EXECUTION.md`、`reports/`
+   - 当前可执行的验证入口（例如 `Makefile`、脚本、前后端命令）
+4. 选择最合适的非破坏性验证命令执行；不要伪造“已验证”。
+5. 如果某项能力缺失，明确写成 gap，不要包装成“基本完成”。
+6. 这个 review 任务默认不改业务代码；重点是判断真实进展、识别缺口、更新 OpenClaw 优化 backlog。
+7. 默认**不要更新任何 TASKS/GOALS 状态**。review 是审查，不是任务回收。
+8. 如果用户明确要求在 review 中同步任务状态：
+   - 只能写 `/home/long/project/llm-intelligence/TASKS.md`
+   - 禁止写 `~/.openclaw/workspace/TASKS.md` 与 `~/.openclaw/workspace/GOALS.md`
+   - 写回前必须先跑一次对应目标文件的预检守卫
+   - 必须先重新读取最新文件，再决定是否 `write`
+
+输出文件：
+
+1. 单次 review 报告：
+   - 路径：`reports/openclaw/YYYY-MM-DD-HHMM-review.md`
+   - 模板：`reports/openclaw/REVIEW_TEMPLATE.md`
+2. OpenClaw 能力优化 backlog：
+   - 路径：`reports/openclaw/OPENCLAW_CAPABILITY_BACKLOG.md`
+   - 追加或更新发现的问题与建议
+
+落盘规则：
+
+1. 写输出文件前，先 `read` 现有文件内容。
+2. 生成输出文件时，统一使用 `write` 工具整文件重写。
+3. 不要使用 `edit` 工具追加、替换或局部修改文件。
+4. 如果需要更新 `OPENCLAW_CAPABILITY_BACKLOG.md`，先读完整文件，再把旧内容与本次新增内容合并后一次性 `write` 回去。
+5. 如果工具返回错误，不要原样重试同一个 `edit`；改为重新读取文件并使用 `write` 全量覆盖。
+6. 对任何共享文档，禁止连续使用同一份 stale `oldText` 重试 `edit`。
+7. 如果仓库状态与上一次 review 相比没有 delta，不要机械重复整份完成项清单；要显式写出“无 delta”并把重点转向风险老化、未提交变更、未验证项。
+
+`YYYY-MM-DD-HHMM-review.md` 必须与项目 daily memory 使用完全一致的字段命名：
+
+- 允许保留标题与 metadata block
+- 除标题与 metadata block 外，顶层 section 只允许：
+  - `## Context`
+  - `## Evidence`
+  - `## Outcome`
+  - `## Next`
+- 不要再使用 `Executive Summary`、`当前真实阶段判断`、`已完成项`、`未完成项` 作为顶层 section 标题
+- 推荐字段映射：
+  - `Context`：review ID、trigger、scope、时间窗口、当前真实阶段判断、本轮背景
+  - `Evidence`：验证命令与结果、已完成项、未完成项、伪进展/文档与实现不一致项、关键 gap 及其证据
+  - `Outcome`：执行摘要、风险判断、阶段结论、本轮最重要的落地结论
+  - `Next`：下一轮最值得推进的 3 件事、明确 owner 或建议动作
+- `Evidence` 和 `Next` 下允许继续使用二级小节或表格，但字段名必须保持上述四段式
+- `Evidence` 段中的每条关键结论，必须尽量标明证据等级：
+  - `runtime-verified`
+  - `artifact-present`
+  - `doc-claimed`
+- 如果只有 `doc-claimed`，必须直接指出“未做真实验证”，不能包装成完成
+- 新报告默认参考 `reports/openclaw/REVIEW_TEMPLATE.md` 生成，避免自由发挥
+
+`OPENCLAW_CAPABILITY_BACKLOG.md` 必须包含：
+
+- 日期时间
+- 本次 review 暴露出的 OpenClaw 能力问题
+- 问题影响
+- 优化建议
+- 优先级（P0/P1/P2）
+- 建议验证方法
+
+完成后，在最终回复中只输出简洁摘要，并列出本次生成/更新的文件。

scripts/review/preflight_task_write_guard.sh

@@ -0,0 +1,47 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+PROJECT_ROOT="/home/long/project/llm-intelligence"
+GLOBAL_GUARD="/home/long/.openclaw/workspace/scripts/preflight_task_write_guard.sh"
+
+usage() {
+  cat <<'EOF'
+Usage:
+  preflight_task_write_guard.sh <writer-role> <target-path> [target-path...]
+
+Writer roles:
+  main-session
+  llm-intelligence-agent
+  llm-intelligence-review
+  llm-intelligence-cron
+EOF
+}
+
+if [[ $# -lt 2 ]]; then
+  usage >&2
+  exit 64
+fi
+
+writer_role="$1"
+shift
+
+case "$writer_role" in
+  main-session)
+    generic_role="main-session"
+    ;;
+  llm-intelligence-agent)
+    generic_role="project-agent"
+    ;;
+  llm-intelligence-review)
+    generic_role="project-review"
+    ;;
+  llm-intelligence-cron)
+    generic_role="project-cron"
+    ;;
+  *)
+    printf '%s\n' "preflight: unsupported writer role: $writer_role" >&2
+    exit 68
+    ;;
+esac
+
+exec "$GLOBAL_GUARD" "$generic_role" "$PROJECT_ROOT" "$@"

scripts/run_daily.sh

@@ -0,0 +1,109 @@
+#!/bin/bash
+# run_daily.sh - 每日数据采集与报告生成流水线
+# Sprint 3: 完整调度脚本（采集→质量检查→报告生成→归档→通知）
+set -euo pipefail
+
+PROJECT_DIR="/home/long/project/llm-intelligence"
+DB_URL="${DATABASE_URL:-host=/var/run/postgresql dbname=llm_intelligence user=long sslmode=disable}"
+REPORT_DATE=$(date +%Y-%m-%d)
+LOG_FILE="/tmp/llm_hub_daily_${REPORT_DATE}.log"
+FEISHU_WEBHOOK="${FEISHU_WEBHOOK:-}"
+
+# 日志函数
+log() {
+    echo "[$(date '+%Y-%m-%d %H:%M:%S')] $1" | tee -a "$LOG_FILE"
+}
+
+# 错误处理
+error_exit() {
+    log "❌ 错误: $1"
+    # 降级：复制昨日报告
+    fallback_report
+    # 发送告警
+    if [ -n "$FEISHU_WEBHOOK" ]; then
+        send_alert "$1"
+    fi
+    exit 1
+}
+
+# 降级：复制昨日报告
+fallback_report() {
+    local yesterday=$(date -d "yesterday" +%Y-%m-%d)
+    local yesterday_md="${PROJECT_DIR}/reports/daily/daily_report_${yesterday}.md"
+    local today_md="${PROJECT_DIR}/reports/daily/daily_report_${REPORT_DATE}.md"
+    
+    if [ -f "$yesterday_md" ]; then
+        cp "$yesterday_md" "$today_md"
+        sed -i "s/${yesterday}/${REPORT_DATE}/g" "$today_md"
+        sed -i "1s/^/# [数据延迟] /" "$today_md"
+        log "⚠️ 已复制昨日报告并标记[数据延迟]"
+    else
+        log "⚠️ 无昨日报告可供复制"
+    fi
+}
+
+# 发送飞书告警
+send_alert() {
+    local msg="$1"
+    local payload="{\"msg_type\":\"text\",\"content\":{\"text\":\"🚨 LLM Hub 日报失败\\n日期: ${REPORT_DATE}\\n错误: ${msg}\\n请检查日志: ${LOG_FILE}\"}}"
+    
+    curl -s -X POST -H "Content-Type: application/json" \
+        -d "$payload" \
+        "$FEISHU_WEBHOOK" > /dev/null || true
+    log "📢 飞书告警已发送"
+}
+
+# 主流程
+log "🚀 开始每日流水线: ${REPORT_DATE}"
+
+cd "$PROJECT_DIR"
+
+# 1. 数据采集
+log "1️⃣ 数据采集..."
+if ! go run scripts/fetch_openrouter.go >> "$LOG_FILE" 2>&1; then
+    error_exit "数据采集失败"
+fi
+log "✅ 数据采集完成"
+
+# 2. 数据质量检查
+log "2️⃣ 数据质量检查..."
+MODEL_COUNT=$(psql "$DB_URL" -t -c "SELECT COUNT(*) FROM models WHERE deleted_at IS NULL" 2>/dev/null | tr -d ' ')
+if [ "$MODEL_COUNT" -lt 10 ]; then
+    error_exit "模型数量不足: ${MODEL_COUNT} < 10"
+fi
+log "✅ 数据质量检查通过 (模型数: ${MODEL_COUNT})"
+
+# 3. 生成日报
+log "3️⃣ 生成日报..."
+export DATABASE_URL="$DB_URL"
+if ! go run scripts/generate_daily_report.go >> "$LOG_FILE" 2>&1; then
+    error_exit "日报生成失败"
+fi
+log "✅ 日报生成完成"
+
+# 4. 归档
+log "4️⃣ 归档报告..."
+ARCHIVE_DIR="reports/daily/$(date +%Y/%m)"
+mkdir -p "$ARCHIVE_DIR"
+cp "reports/daily/daily_report_${REPORT_DATE}.md" "$ARCHIVE_DIR/" 2>/dev/null || true
+cp "reports/daily/html/daily_report_${REPORT_DATE}.html" "$ARCHIVE_DIR/" 2>/dev/null || true
+log "✅ 归档完成"
+
+# 5. 更新 daily_report 表
+log "5️⃣ 更新日报记录..."
+psql "$DB_URL" -c "
+    INSERT INTO daily_report (report_date, status, model_count, output_path, created_at, updated_at)
+    VALUES ('${REPORT_DATE}', 'generated', ${MODEL_COUNT}, 'reports/daily/daily_report_${REPORT_DATE}.md', NOW(), NOW())
+    ON CONFLICT (report_date) DO UPDATE SET
+        status = 'generated',
+        model_count = EXCLUDED.model_count,
+        output_path = EXCLUDED.output_path,
+        updated_at = NOW()
+" >> "$LOG_FILE" 2>&1
+log "✅ 日报记录更新完成"
+
+log "🎉 每日流水线全部完成！"
+log "📄 Markdown: reports/daily/daily_report_${REPORT_DATE}.md"
+log "🌐 HTML: reports/daily/html/daily_report_${REPORT_DATE}.html"
+
+exit 0

scripts/run_real_pipeline.sh

@@ -0,0 +1,38 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT_DIR"
+
+if [[ -f ".env.local" ]]; then
+  # shellcheck disable=SC1091
+  source ".env.local"
+fi
+if [[ -f ".env" ]]; then
+  # shellcheck disable=SC1091
+  source ".env"
+fi
+
+if [[ -z "${DATABASE_URL:-}" ]]; then
+  echo "DATABASE_URL 未设置" >&2
+  exit 1
+fi
+
+if [[ -z "${OPENROUTER_API_KEY:-}" ]]; then
+  echo "OPENROUTER_API_KEY 未设置，无法执行真实采集" >&2
+  exit 1
+fi
+
+"$ROOT_DIR/scripts/apply_migration.sh"
+
+go run "./scripts/fetch_openrouter.go" \
+  -api-key "$OPENROUTER_API_KEY" \
+  -db "$DATABASE_URL" \
+  -out "$ROOT_DIR/models.json"
+
+go run "./scripts/generate_daily_report.go" \
+  -json "$ROOT_DIR/models.json" \
+  -out "$ROOT_DIR/reports/daily"
+
+psql "$DATABASE_URL" -Atqc \
+  "select 'models', count(*) from models union all select 'model_prices', count(*) from model_prices union all select 'report_runs', count(*) from report_runs order by 1;"

scripts/tencent_catalog_lib.go

@@ -0,0 +1,331 @@
+//go:build llm_script
+
+package main
+
+import (
+	"fmt"
+	"html"
+	"io"
+	"net/http"
+	"os"
+	"regexp"
+	"sort"
+	"strings"
+	"time"
+)
+
+const (
+	defaultTencentCatalogURL = "https://cloud.tencent.com/document/product/1823/130060"
+)
+
+var defaultTencentCatalogTimeout = 20 * time.Second
+
+type fetchTencentCatalogConfig struct {
+	URL     string
+	DryRun  bool
+	Timeout time.Duration
+	Fixture string
+}
+
+type tencentCatalog struct {
+	UpdatedAt string
+	Plans     []tencentPlan
+	Models    []tencentModel
+}
+
+type tencentPlan struct {
+	Series       string
+	Tier         string
+	Quota        string
+	Price        string
+	BillingCycle string
+	Scene        string
+}
+
+type tencentModel struct {
+	Series        string
+	Name          string
+	ModelID       string
+	ContextLength int
+	Notes         []string
+}
+
+func fetchTencentCatalogContent(cfg fetchTencentCatalogConfig, client *http.Client) (string, error) {
+	if strings.TrimSpace(cfg.Fixture) != "" {
+		data, err := os.ReadFile(cfg.Fixture)
+		if err != nil {
+			return "", err
+		}
+		return string(data), nil
+	}
+
+	req, err := http.NewRequest(http.MethodGet, cfg.URL, nil)
+	if err != nil {
+		return "", err
+	}
+	req.Header.Set("User-Agent", "llm-intelligence/tencent-catalog-fetcher")
+
+	resp, err := client.Do(req)
+	if err != nil {
+		return "", err
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode != http.StatusOK {
+		return "", fmt.Errorf("unexpected status %d", resp.StatusCode)
+	}
+
+	body, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return "", err
+	}
+	return string(body), nil
+}
+
+func parseTencentCatalog(raw string) (tencentCatalog, error) {
+	lines := normalizeTencentCatalogLines(raw)
+
+	var catalog tencentCatalog
+	var currentSeries string
+	var currentMode string
+
+	for i := 0; i < len(lines); i++ {
+		line := lines[i]
+
+		if catalog.UpdatedAt == "" {
+			if updatedAt := extractUpdatedAt(line); updatedAt != "" {
+				catalog.UpdatedAt = updatedAt
+				continue
+			}
+		}
+
+		if series := extractSeriesHeading(line); series != "" {
+			currentSeries = series
+			currentMode = ""
+			continue
+		}
+
+		switch line {
+		case "### 套餐详情":
+			currentMode = "plans"
+			continue
+		case "### 可用模型":
+			currentMode = "models"
+			continue
+		}
+
+		switch currentMode {
+		case "plans":
+			plan, nextIndex, ok := tryParseTencentPlan(lines, i, currentSeries)
+			if ok {
+				catalog.Plans = append(catalog.Plans, plan)
+				i = nextIndex
+			}
+		case "models":
+			model, nextIndex, ok := tryParseTencentModel(lines, i, currentSeries)
+			if ok {
+				catalog.Models = append(catalog.Models, model)
+				i = nextIndex
+			}
+		}
+	}
+
+	if catalog.UpdatedAt == "" {
+		return tencentCatalog{}, fmt.Errorf("catalog updated_at not found")
+	}
+	if len(catalog.Plans) == 0 {
+		return tencentCatalog{}, fmt.Errorf("catalog plans not found")
+	}
+	if len(catalog.Models) == 0 {
+		return tencentCatalog{}, fmt.Errorf("catalog models not found")
+	}
+	return catalog, nil
+}
+
+func normalizeTencentCatalogLines(raw string) []string {
+	text := html.UnescapeString(raw)
+
+	replacements := []string{"<br>", "<br/>", "<br />", "</p>", "</div>", "</li>", "</tr>", "</td>", "</h1>", "</h2>", "</h3>", "</h4>", "</pre>", "</main>"}
+	for _, replacement := range replacements {
+		text = strings.ReplaceAll(text, replacement, "\n")
+	}
+
+	tagPattern := regexp.MustCompile(`<[^>]+>`)
+	text = tagPattern.ReplaceAllString(text, "")
+
+	text = strings.ReplaceAll(text, "\r\n", "\n")
+	text = strings.ReplaceAll(text, "\r", "\n")
+
+	rawLines := strings.Split(text, "\n")
+	lines := make([]string, 0, len(rawLines))
+	for _, rawLine := range rawLines {
+		line := strings.TrimSpace(rawLine)
+		if line == "" {
+			continue
+		}
+		lines = append(lines, line)
+	}
+	return lines
+}
+
+func extractUpdatedAt(line string) string {
+	const prefix = "最近更新时间："
+	if strings.HasPrefix(line, prefix) {
+		return strings.TrimSpace(strings.TrimPrefix(line, prefix))
+	}
+	return ""
+}
+
+func extractSeriesHeading(line string) string {
+	if !strings.HasPrefix(line, "## ") {
+		return ""
+	}
+	series := strings.TrimSpace(strings.TrimPrefix(line, "## "))
+	if strings.Contains(series, "Token Plan") || strings.Contains(series, "Coding Plan") {
+		return strings.TrimSpace(strings.TrimSuffix(series, "套餐"))
+	}
+	return ""
+}
+
+func tryParseTencentPlan(lines []string, start int, series string) (tencentPlan, int, bool) {
+	if start+4 >= len(lines) {
+		return tencentPlan{}, start, false
+	}
+	if !isTencentPlanName(lines[start]) {
+		return tencentPlan{}, start, false
+	}
+	if !isTencentPlanTier(lines[start+1]) {
+		return tencentPlan{}, start, false
+	}
+	if !strings.Contains(lines[start+2], "订阅月") {
+		return tencentPlan{}, start, false
+	}
+	if !strings.Contains(lines[start+3], "Tokens") {
+		return tencentPlan{}, start, false
+	}
+	if !strings.Contains(lines[start+4], "元/月") {
+		return tencentPlan{}, start, false
+	}
+
+	plan := tencentPlan{
+		Series:       series,
+		Tier:         strings.Trim(lines[start+1], "（）() "),
+		BillingCycle: lines[start+2],
+		Quota:        lines[start+3],
+		Price:        lines[start+4],
+	}
+
+	nextIndex := start + 4
+	if start+5 < len(lines) && !strings.HasPrefix(lines[start+5], "### ") && !isTencentPlanName(lines[start+5]) {
+		plan.Scene = lines[start+5]
+		nextIndex = start + 5
+	}
+	return plan, nextIndex, true
+}
+
+func tryParseTencentModel(lines []string, start int, series string) (tencentModel, int, bool) {
+	if start+1 >= len(lines) {
+		return tencentModel{}, start, false
+	}
+	if !isTencentModelID(lines[start+1]) {
+		return tencentModel{}, start, false
+	}
+	if isReservedTencentLine(lines[start]) {
+		return tencentModel{}, start, false
+	}
+
+	model := tencentModel{
+		Series:  series,
+		Name:    lines[start],
+		ModelID: lines[start+1],
+	}
+
+	notes := make([]string, 0, 4)
+	nextIndex := start + 1
+	for i := start + 2; i < len(lines); i++ {
+		line := lines[i]
+		if strings.HasPrefix(line, "## ") || strings.HasPrefix(line, "### ") {
+			break
+		}
+		if isTencentPlanName(line) && i+1 < len(lines) && isTencentPlanTier(lines[i+1]) {
+			break
+		}
+		if i+1 < len(lines) && isTencentModelID(lines[i+1]) && !isReservedTencentLine(line) {
+			break
+		}
+		notes = append(notes, line)
+		nextIndex = i
+	}
+
+	model.Notes = notes
+	model.ContextLength = extractContextLength(strings.Join(notes, " "))
+	return model, nextIndex, true
+}
+
+func isTencentPlanName(line string) bool {
+	switch line {
+	case "体验套餐", "基础套餐", "进阶套餐", "专业套餐":
+		return true
+	default:
+		return false
+	}
+}
+
+func isTencentPlanTier(line string) bool {
+	return strings.HasPrefix(line, "（") && strings.HasSuffix(line, "）")
+}
+
+func isReservedTencentLine(line string) bool {
+	if strings.HasPrefix(line, "#") {
+		return true
+	}
+	switch line {
+	case "Token Plan 个人版套餐概览", "套餐详情", "可用模型":
+		return true
+	default:
+		return false
+	}
+}
+
+func isTencentModelID(line string) bool {
+	modelIDPattern := regexp.MustCompile(`^[a-z0-9][a-z0-9._-]*$`)
+	return modelIDPattern.MatchString(line)
+}
+
+func extractContextLength(text string) int {
+	contextPattern := regexp.MustCompile(`(?i)(\d+)\s*([KM])\s*上下文`)
+	matches := contextPattern.FindStringSubmatch(text)
+	if len(matches) != 3 {
+		return 0
+	}
+
+	value := 0
+	fmt.Sscanf(matches[1], "%d", &value)
+	switch strings.ToUpper(matches[2]) {
+	case "K":
+		return value * 1024
+	case "M":
+		return value * 1024 * 1024
+	default:
+		return 0
+	}
+}
+
+func formatSeriesSummary(plans []tencentPlan) string {
+	counts := make(map[string]int)
+	for _, plan := range plans {
+		counts[plan.Series]++
+	}
+
+	series := make([]string, 0, len(counts))
+	for name := range counts {
+		series = append(series, name)
+	}
+	sort.Strings(series)
+
+	parts := make([]string, 0, len(series))
+	for _, name := range series {
+		parts = append(parts, fmt.Sprintf("%s:%d", name, counts[name]))
+	}
+	return strings.Join(parts, ",")
+}

scripts/testdata/tencent_token_plan_sample.txt

@@ -0,0 +1,108 @@
+# Token Plan 个人版套餐概览
+最近更新时间：2026-04-27 17:18:02
+
+## 通用 Token Plan 套餐
+### 套餐详情
+体验套餐
+（Lite）
+每订阅月
+3500万 Tokens
+39元/月
+新手尝鲜，入门首选。
+
+基础套餐
+（Standard）
+每订阅月
+1亿 Tokens
+99元/月
+日常使用，高性价比。
+
+进阶套餐
+（Pro）
+每订阅月
+3.2亿 Tokens
+299元/月
+高频 AI 开发。
+
+专业套餐
+（Max）
+每订阅月
+6.5亿 Tokens
+599元/月
+重度 AI 开发首选。
+
+### 可用模型
+Auto
+tc-code-latest
+智能路由
+
+MiniMax-M2.5
+minimax-m2.5
+深度思考、文本生成
+
+MiniMax-M2.7
+minimax-m2.7
+深度思考、文本生成
+
+GLM-5
+glm-5
+深度思考、文本生成
+
+GLM-5.1
+glm-5.1
+深度思考、文本生成
+
+kimi-k2.5
+kimi-k2.5
+深度思考、文本生成
+
+Tencent HY 2.0 Instruct
+hunyuan-2.0-instruct
+文本生成
+
+Tencent HY 2.0 Think
+hunyuan-2.0-thinking
+深度思考、文本生成
+
+Hunyuan-T1
+hunyuan-t1
+文本生成
+
+Hunyuan-TurboS
+hunyuan-turbo
+文本生成
+
+## Hy Token Plan 套餐
+### 套餐详情
+体验套餐
+（Lite）
+每订阅月
+3500万 Tokens
+28元/月
+新手尝鲜，入门首选。
+
+基础套餐
+（Standard）
+每订阅月
+1亿 Tokens
+78元/月
+日常使用，高性价比。
+
+进阶套餐
+（Pro）
+每订阅月
+3.2亿 Tokens
+238元/月
+高频 AI 开发。
+
+专业套餐
+（Max）
+每订阅月
+6.5亿 Tokens
+468元/月
+重度 AI 开发首选。
+
+### 可用模型
+Hy3 preview
+hy3-preview
+原生支持 256K 上下文。

scripts/verification_executor.go

@@ -3,6 +3,8 @@
 // matches expected_evidence, outputs pass/fail report.
 //
 // Usage: go run scripts/verification_executor.go [--dry-run] [--task T-Q2-1.1]
+//go:build llm_script
+
 package main
 
 import (
@@ -21,28 +23,36 @@ import (
 )
 
 type Verification struct {
-	Mode            string
-	Command         string
+	Mode             string
+	Command          string
 	ExpectedEvidence string
-	TimeoutSeconds  int
+	TimeoutSeconds   int
+	EvidenceGrade    string
+	TaskType         string
 }
 
 type TaskResult struct {
-	TaskID    string
-	TaskName  string
-	Verified  bool
-	Command   string
-	ExitCode  int
-	Stdout    string
-	Stderr    string
-	Error     string
-	Reason    string
+	TaskID        string
+	TaskName      string
+	Verified      bool
+	Command       string
+	ExitCode      int
+	Stdout        string
+	Stderr        string
+	StdoutSummary string
+	StderrSummary string
+	Error         string
+	Reason        string
+	EvidenceGrade string
+	TaskType      string
 }
 
 func main() {
 	dryRun := flag.Bool("dry-run", false, "print commands without executing")
 	taskFilter := flag.String("task", "", "filter by task ID (e.g. T-Q2-1.1)")
 	tasksPathFlag := flag.String("tasks", "", "path to TASKS.md")
+	statusFilter := flag.String("status", "all", "filter by normalized status: all|completed|in_progress|planned|paused|unknown")
+	completedOnly := flag.Bool("completed-only", false, "shortcut for --status completed")
 	flag.Parse()
 
 	tasksPath := resolveTasksPath(*tasksPathFlag)
@@ -65,8 +75,18 @@ func main() {
 		tasks = filtered
 	}
 
+	effectiveStatus := *statusFilter
+	if *completedOnly {
+		effectiveStatus = "completed"
+	}
+	tasks, err = filterTasksByStatus(tasks, effectiveStatus)
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "filter tasks: %v\n", err)
+		os.Exit(1)
+	}
+
 	fmt.Printf("=== Verification Report (%s) ===\n", time.Now().Format("2006-01-02 15:04"))
-	fmt.Printf("Tasks checked: %d | Dry-run: %v | TASKS: %s\n\n", len(tasks), *dryRun, tasksPath)
+	fmt.Printf("Tasks checked: %d | Dry-run: %v | Status: %s | TASKS: %s\n\n", len(tasks), *dryRun, effectiveStatus, tasksPath)
 
 	var passed, failed int
 	var results []TaskResult
@@ -87,17 +107,24 @@ func main() {
 			icon = "❌"
 		}
 		fmt.Printf("%s [%s] %s\n", icon, r.TaskID, r.TaskName)
+		if r.Command != "" {
+			fmt.Printf("   cmd: %s\n", r.Command)
+		}
+		if r.EvidenceGrade != "" || r.TaskType != "" {
+			fmt.Printf("   grade: %s | type: %s\n", r.EvidenceGrade, r.TaskType)
+		}
+		if r.StderrSummary != "" {
+			fmt.Printf("   stderr: %s\n", r.StderrSummary)
+		}
+		if r.StdoutSummary != "" && (!r.Verified || r.Reason != "" || r.Error != "") {
+			fmt.Printf("   stdout: %s\n", r.StdoutSummary)
+		}
 		if r.Error != "" {
 			fmt.Printf("   ERROR: %s\n", r.Error)
-		} else {
-			if r.Command != "" {
-				fmt.Printf("   cmd: %s\n", r.Command)
-			}
-			if r.ExitCode != 0 && r.Stdout != "" {
-				fmt.Printf("   output: %s\n", strings.TrimSpace(r.Stdout))
-			} else if r.Reason != "" {
-				fmt.Printf("   reason: %s\n", r.Reason)
-			}
+		} else if r.ExitCode != 0 && r.Stdout != "" {
+			fmt.Printf("   output: %s\n", strings.TrimSpace(r.Stdout))
+		} else if r.Reason != "" {
+			fmt.Printf("   reason: %s\n", r.Reason)
 		}
 	}
 
@@ -108,28 +135,40 @@ func main() {
 }
 
 func resolveTasksPath(flagValue string) string {
+	envValue := os.Getenv("TASKS_PATH")
+	wd := ""
+	if currentWD, err := os.Getwd(); err == nil {
+		wd = currentWD
+	}
+	sourceDir := ""
+	if _, sourcePath, _, ok := runtime.Caller(0); ok {
+		sourceDir = filepath.Dir(sourcePath)
+	}
+	return resolveTasksPathWithContext(flagValue, envValue, wd, sourceDir, "/home/long/.openclaw/workspace/TASKS.md")
+}
+
+func resolveTasksPathWithContext(flagValue, envValue, wd, sourceDir, globalTasksPath string) string {
 	candidates := []string{}
 	if flagValue != "" {
 		candidates = append(candidates, flagValue)
 	}
-	if envValue := os.Getenv("TASKS_PATH"); envValue != "" {
+	if envValue != "" {
 		candidates = append(candidates, envValue)
 	}
 
-	if wd, err := os.Getwd(); err == nil {
+	if wd != "" {
 		candidates = append(candidates,
 			filepath.Join(wd, "TASKS.md"),
 			filepath.Join(wd, "..", "TASKS.md"),
 		)
 	}
 
-	if _, sourcePath, _, ok := runtime.Caller(0); ok {
-		scriptDir := filepath.Dir(sourcePath)
-		candidates = append(candidates, filepath.Join(scriptDir, "..", "TASKS.md"))
+	defaultProjectTasks := ""
+	if sourceDir != "" {
+		defaultProjectTasks = filepath.Join(sourceDir, "..", "TASKS.md")
+		candidates = append(candidates, defaultProjectTasks)
 	}
 
-	candidates = append(candidates, "/home/long/.openclaw/workspace/TASKS.md")
-
 	seen := map[string]struct{}{}
 	for _, candidate := range candidates {
 		if candidate == "" {
@@ -148,16 +187,26 @@ func resolveTasksPath(flagValue string) string {
 	if flagValue != "" {
 		return filepath.Clean(flagValue)
 	}
-	if envValue := os.Getenv("TASKS_PATH"); envValue != "" {
+	if envValue != "" {
 		return filepath.Clean(envValue)
 	}
-	return "/home/long/.openclaw/workspace/TASKS.md"
+	if defaultProjectTasks != "" {
+		return filepath.Clean(defaultProjectTasks)
+	}
+	if wd != "" {
+		return filepath.Clean(filepath.Join(wd, "TASKS.md"))
+	}
+	if globalTasksPath != "" {
+		return filepath.Clean(globalTasksPath)
+	}
+	return "TASKS.md"
 }
 
 type taskEntry struct {
-	ID            string
-	Name          string
-	Verification  Verification
+	ID              string
+	Name            string
+	Status          string
+	Verification    Verification
 	HasVerification bool
 }
 
@@ -176,7 +225,7 @@ func parseTasks(f *os.File) []taskEntry {
 			if currentTask != nil {
 				tasks = append(tasks, *currentTask)
 			}
-			currentTask = &taskEntry{ID: m[1], Name: m[2]}
+			currentTask = &taskEntry{ID: m[1], Name: m[2], Status: normalizeStatusFromText(line)}
 			inVerification = false
 			continue
 		}
@@ -193,6 +242,10 @@ func parseTasks(f *os.File) []taskEntry {
 		}
 
 		if !inVerification {
+			statusRe := regexp.MustCompile(`^\s*-\s+\*\*状态\*\*：(.+)$`)
+			if m := statusRe.FindStringSubmatch(line); m != nil {
+				currentTask.Status = normalizeStatusFromText(m[1])
+			}
 			continue
 		}
 
@@ -216,6 +269,18 @@ func parseTasks(f *os.File) []taskEntry {
 			continue
 		}
 
+		evidenceGradeRe := regexp.MustCompile(`^\s+- evidence_grade:\s+` + "`" + `([^` + "`" + `]+)` + "`")
+		if m := evidenceGradeRe.FindStringSubmatch(line); m != nil {
+			currentTask.Verification.EvidenceGrade = m[1]
+			continue
+		}
+
+		taskTypeRe := regexp.MustCompile(`^\s+- task_type:\s+` + "`" + `([^` + "`" + `]+)` + "`")
+		if m := taskTypeRe.FindStringSubmatch(line); m != nil {
+			currentTask.Verification.TaskType = m[1]
+			continue
+		}
+
 		timeoutRe := regexp.MustCompile(`^\s+- timeout_seconds:\s+(\d+)`)
 		if m := timeoutRe.FindStringSubmatch(line); m != nil {
 			fmt.Sscanf(m[1], "%d", &currentTask.Verification.TimeoutSeconds)
@@ -244,6 +309,18 @@ func verifyTask(t taskEntry, dryRun bool) TaskResult {
 		return r
 	}
 
+	t.Verification.Mode = strings.TrimSpace(t.Verification.Mode)
+	t.Verification.TaskType = normalizeTaskType(t.Verification.TaskType)
+	t.Verification.EvidenceGrade = normalizeEvidenceGrade(t.Verification.Mode, t.Verification.EvidenceGrade)
+	r.TaskType = t.Verification.TaskType
+	r.EvidenceGrade = t.Verification.EvidenceGrade
+
+	if validationErr := validateVerification(t.Verification); validationErr != "" {
+		r.Verified = false
+		r.Reason = validationErr
+		return r
+	}
+
 	if t.Verification.Command == "" {
 		r.Reason = "verification.command is empty"
 		r.Verified = false
@@ -283,6 +360,8 @@ func verifyTask(t taskEntry, dryRun bool) TaskResult {
 
 	r.Stdout = stdout.String()
 	r.Stderr = stderr.String()
+	r.StdoutSummary = summarizeOutput(r.Stdout)
+	r.StderrSummary = summarizeOutput(r.Stderr)
 
 	if r.ExitCode != 0 && t.Verification.Mode == "test_pass" {
 		r.Verified = false
@@ -325,3 +404,134 @@ func verifyTask(t taskEntry, dryRun bool) TaskResult {
 
 	return r
 }
+
+func normalizeEvidenceGrade(mode, explicit string) string {
+	if explicit = strings.TrimSpace(explicit); explicit != "" {
+		return explicit
+	}
+
+	switch strings.TrimSpace(mode) {
+	case "test_pass":
+		return "runtime-verified"
+	case "artifact_present":
+		return "artifact-present"
+	case "semantic":
+		return "doc-claimed"
+	default:
+		return ""
+	}
+}
+
+func normalizeTaskType(raw string) string {
+	raw = strings.TrimSpace(raw)
+	if raw == "" {
+		return "unspecified"
+	}
+	return raw
+}
+
+func normalizeStatusFromText(raw string) string {
+	lower := strings.ToLower(strings.TrimSpace(raw))
+	switch {
+	case strings.Contains(raw, "✅") || strings.Contains(raw, "完成"):
+		return "completed"
+	case strings.Contains(raw, "🟡") || strings.Contains(raw, "进行中"):
+		return "in_progress"
+	case strings.Contains(raw, "🔶") || strings.Contains(raw, "🔴") || strings.Contains(raw, "待启动") || strings.Contains(raw, "未开始"):
+		return "planned"
+	case strings.Contains(raw, "⏸️") || strings.Contains(raw, "待规划") || strings.Contains(raw, "暂停"):
+		return "paused"
+	case lower == "":
+		return "unknown"
+	default:
+		return "unknown"
+	}
+}
+
+func filterTasksByStatus(tasks []taskEntry, filter string) ([]taskEntry, error) {
+	filter = strings.TrimSpace(filter)
+	if filter == "" {
+		filter = "all"
+	}
+
+	valid := map[string]struct{}{
+		"all":         {},
+		"completed":   {},
+		"in_progress": {},
+		"planned":     {},
+		"paused":      {},
+		"unknown":     {},
+	}
+	if _, ok := valid[filter]; !ok {
+		return nil, fmt.Errorf("unsupported status filter: %s", filter)
+	}
+	if filter == "all" {
+		return tasks, nil
+	}
+
+	filtered := make([]taskEntry, 0, len(tasks))
+	for _, t := range tasks {
+		status := t.Status
+		if status == "" {
+			status = "unknown"
+		}
+		if status == filter {
+			filtered = append(filtered, t)
+		}
+	}
+	return filtered, nil
+}
+
+func summarizeOutput(raw string) string {
+	cleaned := strings.TrimSpace(raw)
+	if cleaned == "" {
+		return ""
+	}
+	cleaned = strings.Join(strings.Fields(cleaned), " ")
+	const limit = 220
+	if len(cleaned) <= limit {
+		return cleaned
+	}
+	return cleaned[:limit] + "..."
+}
+
+func validateVerification(v Verification) string {
+	validModes := map[string]struct{}{
+		"test_pass":        {},
+		"artifact_present": {},
+		"semantic":         {},
+	}
+	if _, ok := validModes[v.Mode]; !ok {
+		return fmt.Sprintf("unsupported verification mode: %s", v.Mode)
+	}
+
+	validGrades := map[string]struct{}{
+		"runtime-verified": {},
+		"artifact-present": {},
+		"doc-claimed":      {},
+	}
+	if v.EvidenceGrade != "" {
+		if _, ok := validGrades[v.EvidenceGrade]; !ok {
+			return fmt.Sprintf("unsupported evidence grade: %s", v.EvidenceGrade)
+		}
+	}
+
+	validTaskTypes := map[string]struct{}{
+		"unspecified":   {},
+		"code":          {},
+		"automation":    {},
+		"documentation": {},
+		"configuration": {},
+		"data":          {},
+		"analysis":      {},
+	}
+	if _, ok := validTaskTypes[v.TaskType]; !ok {
+		return fmt.Sprintf("unsupported task type: %s", v.TaskType)
+	}
+
+	if (v.TaskType == "code" || v.TaskType == "automation") && v.Mode == "semantic" {
+		return fmt.Sprintf("semantic-only verification is not allowed for %s tasks", v.TaskType)
+	}
+
+	return ""
+}

scripts/verification_executor_test.go

@@ -0,0 +1,268 @@
+//go:build llm_script
+
+package main
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+func TestParseTasksParsesEvidenceFields(t *testing.T) {
+	md := `# Tasks
+
+### T-1 ✅ Example
+- **verification**:
+  - mode: ` + "`test_pass`" + `
+  - command: ` + "`echo ok`" + `
+  - expected_evidence: ` + "`ok`" + `
+  - evidence_grade: ` + "`runtime-verified`" + `
+  - task_type: ` + "`code`" + `
+  - timeout_seconds: 15
+`
+
+	tmpFile, err := os.CreateTemp(t.TempDir(), "tasks-*.md")
+	if err != nil {
+		t.Fatalf("create temp file: %v", err)
+	}
+	defer tmpFile.Close()
+
+	if _, err := tmpFile.WriteString(md); err != nil {
+		t.Fatalf("write temp file: %v", err)
+	}
+	if _, err := tmpFile.Seek(0, 0); err != nil {
+		t.Fatalf("seek temp file: %v", err)
+	}
+
+	tasks := parseTasks(tmpFile)
+	if len(tasks) != 1 {
+		t.Fatalf("expected 1 task, got %d", len(tasks))
+	}
+
+	got := tasks[0].Verification
+	if got.Mode != "test_pass" {
+		t.Fatalf("expected mode test_pass, got %q", got.Mode)
+	}
+	if got.Command != "echo ok" {
+		t.Fatalf("expected command echo ok, got %q", got.Command)
+	}
+	if got.ExpectedEvidence != "ok" {
+		t.Fatalf("expected evidence ok, got %q", got.ExpectedEvidence)
+	}
+	if got.EvidenceGrade != "runtime-verified" {
+		t.Fatalf("expected evidence grade runtime-verified, got %q", got.EvidenceGrade)
+	}
+	if got.TaskType != "code" {
+		t.Fatalf("expected task type code, got %q", got.TaskType)
+	}
+	if got.TimeoutSeconds != 15 {
+		t.Fatalf("expected timeout 15, got %d", got.TimeoutSeconds)
+	}
+}
+
+func TestVerifyTaskRejectsSemanticOnlyForCodeTask(t *testing.T) {
+	task := taskEntry{
+		ID:   "T-1",
+		Name: "semantic code task",
+		Verification: Verification{
+			Mode:          "semantic",
+			Command:       "echo ok",
+			TaskType:      "code",
+			EvidenceGrade: "doc-claimed",
+		},
+		HasVerification: true,
+	}
+
+	result := verifyTask(task, true)
+	if result.Verified {
+		t.Fatalf("expected semantic-only code task to fail")
+	}
+	if !strings.Contains(result.Reason, "semantic-only") {
+		t.Fatalf("expected semantic-only rejection reason, got %q", result.Reason)
+	}
+}
+
+func TestVerifyTaskDefaultsEvidenceGradeFromMode(t *testing.T) {
+	task := taskEntry{
+		ID:   "T-2",
+		Name: "artifact task",
+		Verification: Verification{
+			Mode:             "artifact_present",
+			Command:          "echo exists",
+			ExpectedEvidence: "exists",
+		},
+		HasVerification: true,
+	}
+
+	result := verifyTask(task, true)
+	if !result.Verified {
+		t.Fatalf("expected dry-run artifact task to pass, got reason %q", result.Reason)
+	}
+	if result.EvidenceGrade != "artifact-present" {
+		t.Fatalf("expected default evidence grade artifact-present, got %q", result.EvidenceGrade)
+	}
+}
+
+func TestResolveTasksPathDoesNotImplicitlyFallbackToGlobal(t *testing.T) {
+	root := t.TempDir()
+	projectDir := filepath.Join(root, "project")
+	globalDir := filepath.Join(root, "workspace")
+	scriptDir := filepath.Join(projectDir, "scripts")
+	if err := os.MkdirAll(projectDir, 0o755); err != nil {
+		t.Fatalf("mkdir project dir: %v", err)
+	}
+	if err := os.MkdirAll(globalDir, 0o755); err != nil {
+		t.Fatalf("mkdir global dir: %v", err)
+	}
+	if err := os.MkdirAll(scriptDir, 0o755); err != nil {
+		t.Fatalf("mkdir script dir: %v", err)
+	}
+
+	projectTasks := filepath.Join(projectDir, "TASKS.md")
+	globalTasks := filepath.Join(globalDir, "TASKS.md")
+	if err := os.WriteFile(projectTasks, []byte("# project"), 0o644); err != nil {
+		t.Fatalf("write project tasks: %v", err)
+	}
+	if err := os.WriteFile(globalTasks, []byte("# global"), 0o644); err != nil {
+		t.Fatalf("write global tasks: %v", err)
+	}
+
+	got := resolveTasksPathWithContext("", "", filepath.Join(root, "outside"), scriptDir, globalTasks)
+	if got != projectTasks {
+		t.Fatalf("expected project tasks path, got %q", got)
+	}
+}
+
+func TestResolveTasksPathAllowsExplicitGlobalPath(t *testing.T) {
+	root := t.TempDir()
+	projectDir := filepath.Join(root, "project")
+	globalDir := filepath.Join(root, "workspace")
+	scriptDir := filepath.Join(projectDir, "scripts")
+	if err := os.MkdirAll(projectDir, 0o755); err != nil {
+		t.Fatalf("mkdir project dir: %v", err)
+	}
+	if err := os.MkdirAll(globalDir, 0o755); err != nil {
+		t.Fatalf("mkdir global dir: %v", err)
+	}
+	if err := os.MkdirAll(scriptDir, 0o755); err != nil {
+		t.Fatalf("mkdir script dir: %v", err)
+	}
+
+	projectTasks := filepath.Join(projectDir, "TASKS.md")
+	globalTasks := filepath.Join(globalDir, "TASKS.md")
+	if err := os.WriteFile(projectTasks, []byte("# project"), 0o644); err != nil {
+		t.Fatalf("write project tasks: %v", err)
+	}
+	if err := os.WriteFile(globalTasks, []byte("# global"), 0o644); err != nil {
+		t.Fatalf("write global tasks: %v", err)
+	}
+
+	got := resolveTasksPathWithContext(globalTasks, "", filepath.Join(root, "outside"), scriptDir, globalTasks)
+	if got != globalTasks {
+		t.Fatalf("expected explicit global tasks path, got %q", got)
+	}
+}
+
+func TestVerifyTaskCapturesFailureSummaries(t *testing.T) {
+	task := taskEntry{
+		ID:   "T-3",
+		Name: "failing task",
+		Verification: Verification{
+			Mode:             "test_pass",
+			Command:          "echo standard-output && echo standard-error 1>&2 && exit 1",
+			ExpectedEvidence: "unused",
+			TaskType:         "automation",
+		},
+		HasVerification: true,
+	}
+
+	result := verifyTask(task, false)
+	if result.Verified {
+		t.Fatalf("expected failing task to fail")
+	}
+	if !strings.Contains(result.StdoutSummary, "standard-output") {
+		t.Fatalf("expected stdout summary to contain command output, got %q", result.StdoutSummary)
+	}
+	if !strings.Contains(result.StderrSummary, "standard-error") {
+		t.Fatalf("expected stderr summary to contain command error, got %q", result.StderrSummary)
+	}
+}
+
+func TestParseTasksParsesNormalizedStatus(t *testing.T) {
+	md := `# Tasks
+
+### T-1 ✅ Done task
+- **状态**：✅ 完成（2026-05-11）
+- **verification**:
+  - mode: ` + "`test_pass`" + `
+  - command: ` + "`echo ok`" + `
+  - expected_evidence: ` + "`ok`" + `
+
+### T-2 🔶 Planned task
+- **状态**：🔶 待启动
+- **verification**:
+  - mode: ` + "`test_pass`" + `
+  - command: ` + "`echo ok`" + `
+  - expected_evidence: ` + "`ok`" + `
+
+### T-3 ⏸️ Paused task
+- **状态**：⏸️ 待规划
+- **verification**:
+  - mode: ` + "`test_pass`" + `
+  - command: ` + "`echo ok`" + `
+  - expected_evidence: ` + "`ok`" + `
+`
+
+	tmpFile, err := os.CreateTemp(t.TempDir(), "tasks-status-*.md")
+	if err != nil {
+		t.Fatalf("create temp file: %v", err)
+	}
+	defer tmpFile.Close()
+
+	if _, err := tmpFile.WriteString(md); err != nil {
+		t.Fatalf("write temp file: %v", err)
+	}
+	if _, err := tmpFile.Seek(0, 0); err != nil {
+		t.Fatalf("seek temp file: %v", err)
+	}
+
+	tasks := parseTasks(tmpFile)
+	if len(tasks) != 3 {
+		t.Fatalf("expected 3 tasks, got %d", len(tasks))
+	}
+
+	if tasks[0].Status != "completed" {
+		t.Fatalf("expected first task status completed, got %q", tasks[0].Status)
+	}
+	if tasks[1].Status != "planned" {
+		t.Fatalf("expected second task status planned, got %q", tasks[1].Status)
+	}
+	if tasks[2].Status != "paused" {
+		t.Fatalf("expected third task status paused, got %q", tasks[2].Status)
+	}
+}
+
+func TestFilterTasksByStatus(t *testing.T) {
+	tasks := []taskEntry{
+		{ID: "T-1", Status: "completed"},
+		{ID: "T-2", Status: "planned"},
+		{ID: "T-3", Status: "in_progress"},
+	}
+
+	completed, err := filterTasksByStatus(tasks, "completed")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if len(completed) != 1 || completed[0].ID != "T-1" {
+		t.Fatalf("expected only completed task, got %#v", completed)
+	}
+
+	all, err := filterTasksByStatus(tasks, "all")
+	if err != nil {
+		t.Fatalf("unexpected error for all: %v", err)
+	}
+	if len(all) != 3 {
+		t.Fatalf("expected all 3 tasks, got %d", len(all))
+	}
+}

scripts/verify_common.sh

@@ -0,0 +1,124 @@
+#!/bin/bash
+
+set -u
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+PROJECT_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+VERIFY_DB_NAME="${VERIFY_DB_NAME:-llm_intelligence}"
+
+PASS_COUNT=0
+FAIL_COUNT=0
+WARN_COUNT=0
+
+pass() {
+    echo "[PASS] $1"
+    PASS_COUNT=$((PASS_COUNT + 1))
+}
+
+fail() {
+    echo "[FAIL] $1"
+    FAIL_COUNT=$((FAIL_COUNT + 1))
+}
+
+warn() {
+    echo "[WARN] $1"
+    WARN_COUNT=$((WARN_COUNT + 1))
+}
+
+sql_scalar() {
+    local sql="$1"
+    psql -d "$VERIFY_DB_NAME" -Atqc "$sql"
+}
+
+check_file() {
+    local path="$1"
+    local desc="$2"
+    if [ -f "$PROJECT_ROOT/$path" ]; then
+        pass "$desc"
+    else
+        fail "$desc (缺少文件: $path)"
+    fi
+}
+
+check_executable() {
+    local path="$1"
+    local desc="$2"
+    if [ -x "$PROJECT_ROOT/$path" ]; then
+        pass "$desc"
+    else
+        fail "$desc (不可执行: $path)"
+    fi
+}
+
+check_shell() {
+    local desc="$1"
+    local cmd="$2"
+    if bash -lc "cd \"$PROJECT_ROOT\" && $cmd" >/tmp/llm_verify_cmd.out 2>/tmp/llm_verify_cmd.err; then
+        pass "$desc"
+    else
+        local details=""
+        if [ -s /tmp/llm_verify_cmd.err ]; then
+            details="$(tr '\n' ' ' </tmp/llm_verify_cmd.err | sed 's/[[:space:]]\+/ /g' | sed 's/ $//')"
+        elif [ -s /tmp/llm_verify_cmd.out ]; then
+            details="$(tr '\n' ' ' </tmp/llm_verify_cmd.out | sed 's/[[:space:]]\+/ /g' | sed 's/ $//')"
+        fi
+        if [ -n "$details" ]; then
+            fail "$desc ($details)"
+        else
+            fail "$desc"
+        fi
+    fi
+}
+
+check_sql_int_ge() {
+    local desc="$1"
+    local sql="$2"
+    local expected="$3"
+    local actual rc details
+    set +e
+    actual="$(sql_scalar "$sql" 2>/tmp/llm_verify_sql.err | tr -d '[:space:]')"
+    rc=$?
+    set -e
+    if [ "$rc" -ne 0 ]; then
+        details="$(tr '\n' ' ' </tmp/llm_verify_sql.err | sed 's/[[:space:]]\+/ /g' | sed 's/ $//')"
+        fail "$desc (SQL执行失败: ${details:-unknown error})"
+        return
+    fi
+    if [[ "$actual" =~ ^-?[0-9]+$ ]] && [ "$actual" -ge "$expected" ]; then
+        pass "$desc (当前: $actual, 期望 >= $expected)"
+    else
+        fail "$desc (当前: ${actual:-N/A}, 期望 >= $expected)"
+    fi
+}
+
+check_sql_int_eq() {
+    local desc="$1"
+    local sql="$2"
+    local expected="$3"
+    local actual rc details
+    set +e
+    actual="$(sql_scalar "$sql" 2>/tmp/llm_verify_sql.err | tr -d '[:space:]')"
+    rc=$?
+    set -e
+    if [ "$rc" -ne 0 ]; then
+        details="$(tr '\n' ' ' </tmp/llm_verify_sql.err | sed 's/[[:space:]]\+/ /g' | sed 's/ $//')"
+        fail "$desc (SQL执行失败: ${details:-unknown error})"
+        return
+    fi
+    if [[ "$actual" =~ ^-?[0-9]+$ ]] && [ "$actual" -eq "$expected" ]; then
+        pass "$desc (当前: $actual)"
+    else
+        fail "$desc (当前: ${actual:-N/A}, 期望 = $expected)"
+    fi
+}
+
+finish_phase() {
+    echo
+    echo "SUMMARY pass=$PASS_COUNT fail=$FAIL_COUNT warn=$WARN_COUNT"
+    if [ "$FAIL_COUNT" -eq 0 ]; then
+        echo "PHASE_RESULT: PASS"
+        exit 0
+    fi
+    echo "PHASE_RESULT: FAIL"
+    exit 1
+}

scripts/verify_lighthouse.sh

@@ -0,0 +1,92 @@
+#!/usr/bin/env bash
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+PROJECT_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+FRONTEND_DIR="$PROJECT_ROOT/frontend"
+LIGHTHOUSE_PORT="${LIGHTHOUSE_PORT:-4173}"
+LIGHTHOUSE_SCORE_THRESHOLD="${LIGHTHOUSE_SCORE_THRESHOLD:-80}"
+LIGHTHOUSE_FCP_THRESHOLD_MS="${LIGHTHOUSE_FCP_THRESHOLD_MS:-2000}"
+LIGHTHOUSE_URL="http://127.0.0.1:${LIGHTHOUSE_PORT}"
+LIGHTHOUSE_JSON="${TMPDIR:-/tmp}/llm_lighthouse_report.json"
+PREVIEW_LOG="${TMPDIR:-/tmp}/llm_lighthouse_preview.log"
+PREVIEW_PID=""
+CHROME_BIN=""
+
+cleanup() {
+  if [[ -n "${PREVIEW_PID:-}" ]] && kill -0 "$PREVIEW_PID" >/dev/null 2>&1; then
+    kill "$PREVIEW_PID" >/dev/null 2>&1 || true
+    wait "$PREVIEW_PID" >/dev/null 2>&1 || true
+  fi
+}
+
+trap cleanup EXIT
+
+for candidate in /usr/bin/google-chrome-stable /usr/bin/google-chrome /snap/bin/chromium /usr/bin/chromium /usr/bin/chromium-browser; do
+  if [[ -x "$candidate" ]]; then
+    CHROME_BIN="$candidate"
+    break
+  fi
+done
+
+if [[ -z "$CHROME_BIN" ]]; then
+  echo "LIGHTHOUSE_ERROR=chrome-not-found"
+  exit 1
+fi
+
+cd "$FRONTEND_DIR"
+npm run build >/tmp/llm_lighthouse_build.out 2>/tmp/llm_lighthouse_build.err
+
+npm exec vite preview -- --host 127.0.0.1 --port "$LIGHTHOUSE_PORT" >"$PREVIEW_LOG" 2>&1 &
+PREVIEW_PID=$!
+
+for _ in $(seq 1 30); do
+  if curl -fsS "$LIGHTHOUSE_URL" >/dev/null 2>&1; then
+    break
+  fi
+  sleep 0.5
+done
+
+if ! curl -fsS "$LIGHTHOUSE_URL" >/dev/null 2>&1; then
+  echo "LIGHTHOUSE_ERROR=preview-not-ready"
+  exit 1
+fi
+
+npx lighthouse "$LIGHTHOUSE_URL" \
+  --chrome-path="$CHROME_BIN" \
+  --only-categories=performance \
+  --preset=desktop \
+  --output=json \
+  --output-path="$LIGHTHOUSE_JSON" \
+  --quiet \
+  --chrome-flags="--headless=new --no-sandbox --disable-dev-shm-usage" \
+  >/tmp/llm_lighthouse_stdout.out 2>/tmp/llm_lighthouse_stderr.err
+
+read -r score fcp speed_index lcp <<EOF
+$(node -e '
+const fs = require("fs");
+const report = JSON.parse(fs.readFileSync(process.argv[1], "utf8"));
+const perf = Math.round((report.categories.performance.score || 0) * 100);
+const fcp = Math.round(report.audits["first-contentful-paint"]?.numericValue || 0);
+const speedIndex = Math.round(report.audits["speed-index"]?.numericValue || 0);
+const lcp = Math.round(report.audits["largest-contentful-paint"]?.numericValue || 0);
+console.log(`${perf} ${fcp} ${speedIndex} ${lcp}`);
+' "$LIGHTHOUSE_JSON")
+EOF
+
+echo "LIGHTHOUSE_SCORE=${score}"
+echo "LIGHTHOUSE_FCP_MS=${fcp}"
+echo "LIGHTHOUSE_SPEED_INDEX_MS=${speed_index}"
+echo "LIGHTHOUSE_LCP_MS=${lcp}"
+
+if (( score < LIGHTHOUSE_SCORE_THRESHOLD )); then
+  echo "LIGHTHOUSE_ERROR=score-below-threshold"
+  exit 1
+fi
+
+if (( fcp >= LIGHTHOUSE_FCP_THRESHOLD_MS )); then
+  echo "LIGHTHOUSE_ERROR=fcp-above-threshold"
+  exit 1
+fi
+

scripts/verify_phase1.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+. "$SCRIPT_DIR/verify_common.sh"
+
+echo "=== Phase 1 验收检查 ==="
+
+check_file "db/migrations/001_phase1_core_tables.sql" "核心三表迁移文件存在"
+check_file "db/migrations/002_sprint1_complete_schema.sql" "Sprint 1 扩展迁移文件存在"
+check_sql_int_eq "Sprint 1 扩展表全部存在" \
+    "select count(*) from pg_tables where schemaname='public' and tablename in ('model_provider','operator','region_pricing','pricing_history','free_tier','daily_report','user_subscription','audit_log');" \
+    8
+check_sql_int_eq "models 扩展字段已落库" \
+    "select count(*) from information_schema.columns where table_name='models' and column_name in ('provider_id','modality','batch_id','data_confidence');" \
+    4
+check_sql_int_eq "关键 CHECK 约束存在" \
+    "select count(*) from pg_constraint where conname in ('chk_price_non_negative','chk_currency_valid','chk_models_context_length','chk_models_modality','chk_models_data_confidence');" \
+    5
+check_sql_int_eq "updated_at 触发器已挂载到业务表" \
+    "select count(*) from pg_trigger where tgname in ('models_updated_at','model_provider_updated_at','operator_updated_at','region_pricing_updated_at','pricing_history_updated_at','free_tier_updated_at','daily_report_updated_at','user_subscription_updated_at');" \
+    8
+check_sql_int_ge "厂商种子数据不少于 6 条" \
+    "select count(*) from model_provider;" \
+    6
+check_sql_int_ge "region_pricing 已有迁移数据" \
+    "select count(*) from region_pricing;" \
+    1
+check_sql_int_eq "血缘字段 batch_id 已完成回填" \
+    "select count(*) from models where batch_id is null;" \
+    0
+
+finish_phase

scripts/verify_phase2.sh

@@ -0,0 +1,32 @@
+#!/bin/bash
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+. "$SCRIPT_DIR/verify_common.sh"
+
+echo "=== Phase 2 验收检查 ==="
+
+check_shell "ProviderMapper 单元测试通过" "go test ./internal/collectors/..."
+check_shell "重试组件单元测试通过" "go test ./internal/retry/..."
+check_shell "OpenRouter 采集器可独立构建" "go build -o /dev/null ./scripts/fetch_openrouter.go"
+check_sql_int_ge "国内厂商种子数不少于 7 家" \
+    "select count(*) from model_provider where country='CN';" \
+    7
+check_sql_int_ge "国内厂商模型数不少于 10 条" \
+    "select count(*) from models m join model_provider p on m.provider_id = p.id where p.country='CN';" \
+    10
+check_sql_int_ge "CNY 定价记录不少于 10 条" \
+    "select count(*) from region_pricing where currency='CNY';" \
+    10
+check_sql_int_ge "采集成功统计记录至少 1 条" \
+    "select count(*) from collector_stats where success = true;" \
+    1
+check_sql_int_ge "models 总量达到 371+" \
+    "select count(*) from models where deleted_at is null;" \
+    371
+check_sql_int_ge "models 审计日志达到 371+" \
+    "select count(*) from audit_log where table_name='models';" \
+    371
+
+finish_phase

scripts/verify_phase3.sh

@@ -0,0 +1,26 @@
+#!/bin/bash
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+. "$SCRIPT_DIR/verify_common.sh"
+
+TODAY="$(date +%Y-%m-%d)"
+ARCHIVE_DIR="reports/daily/$(date +%Y/%m)"
+
+echo "=== Phase 3 验收检查 ==="
+
+check_executable "scripts/run_daily.sh" "日报流水线脚本可执行"
+check_executable "scripts/feishu_alert.sh" "飞书告警脚本可执行"
+check_shell "日报生成器可独立构建" "go build -o /dev/null ./scripts/generate_daily_report.go"
+check_shell "日报脚本包含降级逻辑" "grep -q 'fallback_report' scripts/run_daily.sh"
+check_shell "日报脚本包含飞书告警逻辑" "grep -q 'send_alert' scripts/run_daily.sh"
+check_shell "今日日报文件存在且包含数据质量摘要" "test -f reports/daily/daily_report_${TODAY}.md && grep -q '数据质量摘要' reports/daily/daily_report_${TODAY}.md"
+check_shell "今日归档报告存在" "test -f ${ARCHIVE_DIR}/daily_report_${TODAY}.md"
+check_sql_int_ge "daily_report 已写入至少 1 条 generated 记录" \
+    "select count(*) from daily_report where status='generated';" \
+    1
+check_shell "crontab 已配置每日调度" "crontab -l 2>/dev/null | grep -q 'scripts/run_daily.sh'"
+check_shell "真实采集 API Key 已配置" "([ -n \"${OPENROUTER_API_KEY:-}\" ] || ([ -f .env.local ] && grep -Eq '^OPENROUTER_API_KEY=.+' .env.local) || ([ -f .env ] && grep -Eq '^OPENROUTER_API_KEY=.+' .env))"
+
+finish_phase

scripts/verify_phase4.sh

@@ -0,0 +1,21 @@
+#!/bin/bash
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+. "$SCRIPT_DIR/verify_common.sh"
+
+echo "=== Phase 4 验收检查 ==="
+
+check_file "frontend/package.json" "前端 package.json 存在"
+check_file "frontend/vite.config.ts" "前端 vite.config.ts 存在"
+check_file "frontend/tsconfig.json" "前端 tsconfig.json 存在"
+check_shell "前端生产构建通过" "cd frontend && npm run build >/tmp/llm_phase4_build.log 2>&1"
+check_shell "App 已接入 Dashboard 和 Explorer 入口" "grep -q 'Dashboard' frontend/src/App.tsx && grep -q 'Explorer' frontend/src/App.tsx"
+check_shell "Explorer 已实现分页/排序/筛选" "grep -q 'PAGE_SIZE' frontend/src/pages/Explorer.tsx && grep -q 'toggleSort' frontend/src/pages/Explorer.tsx && grep -q 'providerFilter' frontend/src/pages/Explorer.tsx && grep -q 'modalityFilter' frontend/src/pages/Explorer.tsx"
+check_shell "Explorer 具备 API 失败回退到本地 JSON" "grep -q \"latest_models.json\" frontend/src/lib/models.ts && grep -q \"models.json\" frontend/src/lib/models.ts"
+check_shell "Dashboard 已集成 ECharts" "grep -q \"from 'echarts'\" frontend/src/pages/Dashboard.tsx"
+check_shell "Explorer 已实现 stale 状态显示" "grep -qi 'stale' frontend/src/pages/Explorer.tsx"
+check_shell "Explorer 已实现 pricing unavailable 显示" "grep -qi 'pricing unavailable' frontend/src/lib/models.ts"
+
+finish_phase

scripts/verify_phase5.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+. "$SCRIPT_DIR/verify_common.sh"
+
+echo "=== Phase 5 验收检查 ==="
+
+check_file "Dockerfile" "Dockerfile 存在"
+check_file "docker-compose.yml" "docker-compose.yml 存在"
+check_file "nginx.conf" "Nginx 配置存在"
+check_file ".env.example" ".env.example 存在"
+check_file ".github/workflows/ci.yml" "GitHub Actions CI 配置存在"
+check_executable "scripts/backup.sh" "数据库备份脚本可执行"
+check_file "healthcheck.sh" "健康检查脚本存在"
+check_file "scripts/restore.sh" "数据库恢复脚本存在"
+check_shell "CI 包含 Go 测试" "grep -Eq 'go test .*\\./internal/|go test .*\\./\\.\\.\\.' .github/workflows/ci.yml"
+check_shell "CI 包含前端构建" "grep -q 'npm run build' .github/workflows/ci.yml"
+check_shell "CI 包含 Docker 构建" "grep -q 'docker build' .github/workflows/ci.yml"
+check_shell "CI 配置了覆盖率门禁" "grep -Eqi 'coverage|80%' .github/workflows/ci.yml"
+check_shell "CI 配置了构建产物上传" "grep -Eqi 'upload-artifact|artifacts' .github/workflows/ci.yml"
+check_shell "日志轮转配置已落地" "find . -maxdepth 3 -type f | grep -Eqi 'logrotate|logrotate\\.conf'"
+
+finish_phase

scripts/verify_phase6.sh

@@ -0,0 +1,133 @@
+#!/bin/bash
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+. "$SCRIPT_DIR/verify_common.sh"
+
+DB_URL="${DATABASE_URL:-host=/var/run/postgresql dbname=llm_intelligence user=long sslmode=disable}"
+SERVER_BIN="/tmp/llm_phase6_server"
+SERVER_LOG="/tmp/llm_phase6_server.log"
+SERVER_PORT="${PHASE6_PORT:-}"
+SERVER_PID=""
+
+cleanup() {
+    if [ -n "${SERVER_PID:-}" ] && kill -0 "$SERVER_PID" >/dev/null 2>&1; then
+        kill "$SERVER_PID" >/dev/null 2>&1 || true
+        wait "$SERVER_PID" >/dev/null 2>&1 || true
+    fi
+    rm -f "$SERVER_BIN"
+}
+
+trap cleanup EXIT
+
+port_in_use() {
+    local port="$1"
+    (echo >"/dev/tcp/127.0.0.1/$port") >/dev/null 2>&1
+}
+
+reserve_server_port() {
+    if [ -n "${SERVER_PORT:-}" ]; then
+        return 0
+    fi
+    for candidate in $(seq 18080 18120); do
+        if ! port_in_use "$candidate"; then
+            SERVER_PORT="$candidate"
+            return 0
+        fi
+    done
+    return 1
+}
+
+start_server() {
+    DATABASE_URL="$DB_URL" PORT="$SERVER_PORT" "$SERVER_BIN" >"$SERVER_LOG" 2>&1 &
+    SERVER_PID=$!
+    for _ in $(seq 1 20); do
+        if ! kill -0 "$SERVER_PID" >/dev/null 2>&1; then
+            return 1
+        fi
+        if curl -fsS "http://127.0.0.1:${SERVER_PORT}/health" >/tmp/llm_phase6_health.out 2>/tmp/llm_phase6_health.err &&
+           grep -q '"status":"ok"' /tmp/llm_phase6_health.out; then
+            return 0
+        fi
+        sleep 0.5
+    done
+    return 1
+}
+
+echo "=== Phase 6 综合验收检查 ==="
+
+check_shell "Phase 1~5 总门禁通过" "bash scripts/verify_pre_phase6.sh"
+check_shell "全仓 Go 测试通过" "go test ./..."
+check_shell "脚本级采集器单测通过" "bash scripts/test.sh"
+check_shell "真实采集并输出今日日报" "bash scripts/run_real_pipeline.sh"
+check_shell "API Server 可构建" "go build -o /dev/null ./cmd/server"
+check_shell "健康检查脚本通过" "DATABASE_URL='$DB_URL' bash healthcheck.sh"
+check_shell "密钥未硬编码进源码" "grep -R -n 'sk-' cmd internal frontend/src scripts .github/workflows --include='*.go' --include='*.ts' --include='*.tsx' --include='*.sh' --include='*.yml' --include='*.yaml' --exclude='verify_phase6.sh' >/tmp/llm_phase6_secret_scan.out 2>/dev/null; test ! -s /tmp/llm_phase6_secret_scan.out"
+check_shell "最近 7 次采集成功率达到 95%" "psql \"$DB_URL\" -Atqc \"select coalesce(round(avg(case when success then 100 else 0 end),2),0) from (select success from collector_stats order by created_at desc limit 7) t;\" | awk '{ exit !(\$1 >= 95) }'"
+
+if go build -o "$SERVER_BIN" ./cmd/server >/tmp/llm_phase6_server_build.out 2>/tmp/llm_phase6_server_build.err; then
+    if reserve_server_port && start_server; then
+        pass "API /health 可用"
+
+        set +e
+        api_metrics="$(curl -sS -o /tmp/llm_phase6_models.json -w '%{http_code} %{time_total}' "http://127.0.0.1:${SERVER_PORT}/api/v1/models")"
+        api_rc=$?
+        set -e
+        if [ "$api_rc" -eq 0 ]; then
+            api_code="$(printf '%s' "$api_metrics" | awk '{print $1}')"
+            api_time="$(printf '%s' "$api_metrics" | awk '{print $2}')"
+            if [ "$api_code" = "200" ]; then
+                pass "API /api/v1/models 返回 200"
+            else
+                fail "API /api/v1/models 返回异常状态 (HTTP ${api_code:-unknown})"
+            fi
+
+            if awk "BEGIN { exit !($api_time < 0.5) }"; then
+                pass "API 响应 < 500ms (当前: ${api_time}s)"
+            else
+                fail "API 响应 >= 500ms (当前: ${api_time}s)"
+            fi
+
+            if grep -q '"data"' /tmp/llm_phase6_models.json; then
+                pass "API 返回模型数据载荷"
+            else
+                fail "API 返回体缺少 data 字段"
+            fi
+        else
+            fail "API /api/v1/models 请求失败"
+        fi
+
+        set +e
+        plan_metrics="$(curl -sS -o /tmp/llm_phase6_subscription_plans.json -w '%{http_code} %{time_total}' "http://127.0.0.1:${SERVER_PORT}/api/v1/subscription-plans")"
+        plan_rc=$?
+        set -e
+        if [ "$plan_rc" -eq 0 ]; then
+            plan_code="$(printf '%s' "$plan_metrics" | awk '{print $1}')"
+            if [ "$plan_code" = "200" ]; then
+                pass "API /api/v1/subscription-plans 返回 200"
+            else
+                fail "API /api/v1/subscription-plans 返回异常状态 (HTTP ${plan_code:-unknown})"
+            fi
+
+            if grep -q '"data"' /tmp/llm_phase6_subscription_plans.json; then
+                pass "API 返回套餐数据载荷"
+            else
+                fail "套餐 API 返回体缺少 data 字段"
+            fi
+        else
+            fail "API /api/v1/subscription-plans 请求失败"
+        fi
+    else
+        details="$(tr '\n' ' ' <"$SERVER_LOG" | sed 's/[[:space:]]\+/ /g' | sed 's/ $//')"
+        fail "API Server 启动失败 (${details:-no server log})"
+    fi
+else
+    details="$(tr '\n' ' ' </tmp/llm_phase6_server_build.err | sed 's/[[:space:]]\+/ /g' | sed 's/ $//')"
+    fail "API Server 构建失败 (${details:-unknown build error})"
+fi
+
+check_shell "Phase 6 性能文档存在" "test -f docs/PERFORMANCE_TEST.md"
+check_shell "前端已具备测试入口" "cd frontend && npm run test -- --run >/tmp/llm_phase6_frontend_test.out 2>/tmp/llm_phase6_frontend_test.err"
+
+finish_phase

scripts/verify_pre_phase6.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+
+phases=(
+    "verify_phase1.sh"
+    "verify_phase2.sh"
+    "verify_phase3.sh"
+    "verify_phase4.sh"
+    "verify_phase5.sh"
+)
+
+overall=0
+
+echo "=== Pre-Phase 6 总验收 ==="
+
+for phase in "${phases[@]}"; do
+    echo
+    echo ">>> 执行 ${phase}"
+    if bash "$SCRIPT_DIR/$phase"; then
+        echo ">>> ${phase} PASS"
+    else
+        echo ">>> ${phase} FAIL"
+        overall=1
+    fi
+done
+
+echo
+if [ "$overall" -eq 0 ]; then
+    echo "PRE_PHASE6_RESULT: PASS"
+else
+    echo "PRE_PHASE6_RESULT: FAIL"
+fi
+
+exit "$overall"

scripts/verify_video_pipeline.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+. "$SCRIPT_DIR/verify_common.sh"
+
+REPORT_PATH="${1:-$(find "$PROJECT_ROOT/reports/daily" -maxdepth 1 -type f -name 'daily_report_*.md' | sort | tail -1)}"
+if [[ -z "$REPORT_PATH" || ! -f "$REPORT_PATH" ]]; then
+    echo "未找到日报 Markdown 文件" >&2
+    exit 1
+fi
+
+REPORT_DATE="$(basename "$REPORT_PATH" | sed -E 's/^daily_report_([0-9-]+)\.md$/\1/')"
+OUTPUT_DIR="$PROJECT_ROOT/reports/daily/video/$REPORT_DATE"
+
+echo "=== 视频日报验收检查 ==="
+
+check_shell "视频日报生成器单元测试通过" "go test -tags llm_script scripts/generate_video_digest.go scripts/generate_video_digest_test.go"
+check_shell "视频日报生成器真实产物落盘" "go run -tags llm_script scripts/generate_video_digest.go --report \"$REPORT_PATH\" --output-dir \"$OUTPUT_DIR\" >/tmp/llm_video_digest.out 2>/tmp/llm_video_digest.err && grep -q 'cards=5' /tmp/llm_video_digest.out"
+check_file "reports/daily/video/$REPORT_DATE/manifest.json" "manifest 已生成"
+check_file "reports/daily/video/$REPORT_DATE/video_digest.gif" "GIF 视频原型已生成"
+check_file "reports/daily/video/$REPORT_DATE/narration.wav" "旁白音轨已生成"
+check_shell "脚本分镜文件达到 5 份" "test \"\$(find '$OUTPUT_DIR/scripts' -maxdepth 1 -type f -name '*.md' | wc -l)\" -eq 5"
+check_shell "PNG 帧达到 5 张" "test \"\$(find '$OUTPUT_DIR/frames' -maxdepth 1 -type f -name '*.png' | wc -l)\" -eq 5"
+check_shell "GIF 文件头有效" "head -c 6 \"$OUTPUT_DIR/video_digest.gif\" | grep -Eq 'GIF8[79]a'"
+check_shell "WAV 文件头有效" "head -c 4 \"$OUTPUT_DIR/narration.wav\" | grep -q 'RIFF'"
+check_shell "manifest 引用了视频与音轨产物" "grep -q 'video_digest.gif' \"$OUTPUT_DIR/manifest.json\" && grep -q 'narration.wav' \"$OUTPUT_DIR/manifest.json\""
+
+echo
+echo "SUMMARY pass=$PASS_COUNT fail=$FAIL_COUNT warn=$WARN_COUNT"
+if [ "$FAIL_COUNT" -eq 0 ]; then
+    echo "VIDEO_PIPELINE: PASS"
+    exit 0
+fi
+echo "VIDEO_PIPELINE: FAIL"
+exit 1