feat(biz-articles): add --unread filter (one latest article per account)

只列「有未读的公众号」的最近 1 篇文章 — 与 'wx unread --filter official'
行为一致，便于扫描"哪些公众号还有未读，标题是啥"。

- ipc.rs: BizArticles 加 unread: bool 字段（serde default = false 向后兼容）
- cli/mod.rs: --unread flag
- cli/biz_articles.rs: 透传 unread
- daemon/server.rs: dispatch 加 unread 参数
- daemon/query.rs: q_biz_articles
  - 开启 --unread 时先查 session.db 拿 unread_count>0 且
    chat_type==official_account 的 username 集合
  - 与 --account 取交集（两者都给时进一步缩小范围）
  - 空交集提前 return，避免无意义全表扫
  - 解析后按 pub_time DESC 排，每个 account_username 只保留首条
  - 最后再 truncate(limit)

PR_DRAFT.md

@@ -0,0 +1,115 @@
+# feat(biz): add `wx biz-articles` command to query public account messages
+
+## Summary
+
+Adds a new `biz-articles` subcommand that queries locally cached WeChat public account (公众号) article pushes from `biz_message_0.db`.
+
+This enables a downstream workflow for downloading full article content:
+
+```bash
+wx biz-articles --since today --json | jq '.[].url' | xargs opencli weixin download
+```
+
+## Background
+
+- WeChat stores public account (官方账号) message pushes in a separate database: `message/biz_message_0.db` (SQLCipher 4 encrypted)
+- This DB was not exposed by any existing wx-cli command
+- The encryption key is already scanned and stored in `~/.wx-cli/all_keys.json` by `wx init`
+- Each public account has its own `Msg_{md5(username)}` table, following the same convention as `message_0.db`
+- Message content is zstd-compressed XML containing `<mmreader>/<item>` structures with article metadata
+
+## New CLI Interface
+
+```bash
+# Last 50 articles (default)
+wx biz-articles
+
+# More articles
+wx biz-articles -n 200
+
+# Filter by public account name (fuzzy match on display name)
+wx biz-articles --account "返朴"
+wx biz-articles --account "Datawhale"
+
+# Time filter (article publish time, YYYY-MM-DD)
+wx biz-articles --since 2026-05-10
+wx biz-articles --since 2026-05-01 --until 2026-05-10
+
+# JSON output (for downstream piping)
+wx biz-articles --json
+wx biz-articles --since 2026-05-10 --json | jq '.[].url'
+```
+
+## Output Fields
+
+Each article item includes:
+
+| Field | Description |
+|-------|-------------|
+| `time` | Article publish time (formatted) |
+| `timestamp` | Article publish timestamp (seconds) |
+| `recv_time` | Message receive time (when WeChat pushed it) |
+| `recv_time_str` | Message receive time (formatted) |
+| `account` | Public account display name |
+| `account_username` | Public account username (gh_*) |
+| `title` | Article title |
+| `url` | Article URL (mp.weixin.qq.com link) |
+| `digest` | Article summary/excerpt |
+| `cover_url` | Cover image URL |
+
+## Implementation Notes
+
+- `biz_message_0.db` is loaded on-demand via existing `DbCache` mechanism (no startup cost unless `biz-articles` is called)
+- The key for `message/biz_message_0.db` is already in `all_keys.json`, no changes to `wx init` needed
+- Multi-article pushes (图文消息) are expanded: each `<item>` in `<mmreader>` becomes a separate output row
+- Items without URL or title (e.g., payment notifications from service accounts) are filtered out
+- New `extract_cdata` helper function strips CDATA wrappers from XML content
+- Results sorted by `pub_time` DESC (article publish time, not message receive time)
+
+## Changes
+
+- `src/ipc.rs`: Add `BizArticles` IPC request variant
+- `src/cli/biz_articles.rs`: New CLI command handler (follows sns_feed pattern)  
+- `src/cli/mod.rs`: Register `BizArticles` subcommand in clap + dispatch
+- `src/daemon/query.rs`: Add `q_biz_articles` query + `parse_biz_xml_items` + `extract_cdata` helpers + 8 unit tests
+- `src/daemon/server.rs`: Add dispatch case for `BizArticles`
+
+## Test Results
+
+```
+test result: ok. 49 passed; 0 failed; 0 ignored
+```
+
+New tests (8):
+- `biz_tests::extract_cdata_normal`
+- `biz_tests::extract_cdata_empty`
+- `biz_tests::extract_cdata_url`
+- `biz_tests::extract_cdata_no_cdata_wrapper`
+- `biz_tests::parse_biz_xml_items_single_article`
+- `biz_tests::parse_biz_xml_items_skips_no_url`
+- `biz_tests::parse_biz_xml_items_multi_article`
+- `biz_tests::parse_biz_xml_items_pub_time_fallback`
+
+## Verified Output (real WeChat install with ~30 public accounts, 2026-05-10)
+
+```yaml
+- account: 返朴
+  title: 细胞生物学家俞立：从后进生到科学家，一个ADHD孩子的逆袭
+  url: http://mp.weixin.qq.com/s?__biz=Mzg2MTUyODU2NA==&mid=2247642795&...
+
+- account: Datawhale
+  title: 刚刚，Claude Code 团队这篇文章爆了！
+  url: http://mp.weixin.qq.com/s?__biz=MzIyNjM2MzQyNg==&mid=2247722630&...
+
+- account: 土猛的员外
+  title: AI时代，企业的业务底座正在从数据库变成知识引擎
+  url: http://mp.weixin.qq.com/s?__biz=MzIyOTA5NTM1OA==&mid=2247485270&...
+```
+
+## Branch
+
+`ChenyqThu/wx-cli` → `feat/biz-articles`
+
+---
+
+*Waiting for Lucien's review before opening PR.*

src/cli/biz_articles.rs

@@ -9,6 +9,7 @@ pub fn cmd_biz_articles(
     account: Option<String>,
     since: Option<String>,
     until: Option<String>,
+    unread: bool,
     json: bool,
 ) -> Result<()> {
     let since_ts = since.as_deref().map(parse_time).transpose()?;
@@ -19,6 +20,7 @@ pub fn cmd_biz_articles(
         account,
         since: since_ts,
         until: until_ts,
+        unread,
     };
     let resp = transport::send(req)?;
     let data = resp.data.get("articles")

src/cli/mod.rs

@@ -235,6 +235,9 @@ enum Commands {
         /// 结束时间 YYYY-MM-DD
         #[arg(long)]
         until: Option<String>,
+        /// 只看有未读的公众号，每个公众号取最新 1 篇
+        #[arg(long)]
+        unread: bool,
         /// 输出 JSON（默认 YAML）
         #[arg(long)]
         json: bool,
@@ -323,8 +326,8 @@ fn dispatch(cli: Cli) -> Result<()> {
         Commands::SnsSearch { keyword, limit, since, until, user, json } => {
             sns_search::cmd_sns_search(keyword, limit, since, until, user, json)
         }
-        Commands::BizArticles { limit, account, since, until, json } => {
-            biz_articles::cmd_biz_articles(limit, account, since, until, json)
+        Commands::BizArticles { limit, account, since, until, unread, json } => {
+            biz_articles::cmd_biz_articles(limit, account, since, until, unread, json)
         }
         Commands::Daemon { cmd } => daemon_cmd::cmd_daemon(cmd),
     }

src/daemon/query.rs

@@ -3046,11 +3046,41 @@ pub async fn q_biz_articles(
     account: Option<String>,
     since: Option<i64>,
     until: Option<i64>,
+    unread: bool,
 ) -> Result<Value> {
     let biz_path = db.get("message/biz_message_0.db").await?
         .context("无法解密 biz_message_0.db，请确认 all_keys.json 包含对应密鑰")?
 ;
 
+    // 开启 --unread：从 session.db 拿“公众号 + unread_count>0”的 username 子集，
+    // 作为合集过滤（与 --account 取交集），后续结果按 account_username 去重取顶 1 篇。
+    let unread_usernames: Option<std::collections::HashSet<String>> = if unread {
+        let session_path = db.get("session/session.db").await?
+            .context("无法解密 session.db")?;
+        let session_path2 = session_path.clone();
+        let unread_rows: Vec<String> = tokio::task::spawn_blocking(move || {
+            let conn = Connection::open(&session_path2)?;
+            let mut stmt = conn.prepare(
+                "SELECT username FROM SessionTable WHERE unread_count > 0"
+            )?;
+            let rows: Vec<String> = stmt.query_map([], |row| row.get::<_, String>(0))?
+                .filter_map(|r| r.ok())
+                .collect();
+            Ok::<_, anyhow::Error>(rows)
+        }).await??;
+        // 仅保留公众号类型的未读会话
+        let set: std::collections::HashSet<String> = unread_rows.into_iter()
+            .filter(|u| chat_type_of(u, names) == "official_account")
+            .collect();
+        if set.is_empty() {
+            // 没有未读公众号 → 直接空返回，避免打 biz 表扫描
+            return Ok(json!({ "count": 0, "articles": [] }));
+        }
+        Some(set)
+    } else {
+        None
+    };
+
     // 1. 从 Name2Id 表获取 rowid -> username 映射，再推导 md5 -> username
     let biz_path2 = biz_path.clone();
     let id2username: HashMap<i64, String> = tokio::task::spawn_blocking(move || {
@@ -3071,7 +3101,7 @@ pub async fn q_biz_articles(
 
     // 2. 如果 指定了 --account，找到匹配的 username 列表
     let account_low = account.as_deref().map(|s| s.to_lowercase());
-    let target_usernames: Option<Vec<String>> = account_low.as_ref().map(|low| {
+    let mut target_usernames: Option<Vec<String>> = account_low.as_ref().map(|low| {
         id2username.values()
             .filter(|u| {
                 let display = names.display(u);
@@ -3082,6 +3112,20 @@ pub async fn q_biz_articles(
             .collect()
     });
 
+    // --unread 与 --account 取交集（进一步缩小范围）
+    if let Some(ref unread_set) = unread_usernames {
+        target_usernames = Some(match target_usernames.take() {
+            Some(acc_list) => acc_list.into_iter()
+                .filter(|u| unread_set.contains(u))
+                .collect(),
+            None => unread_set.iter().cloned().collect(),
+        });
+        // 交集为空 → 提前返回
+        if target_usernames.as_ref().map(|v| v.is_empty()).unwrap_or(false) {
+            return Ok(json!({ "count": 0, "articles": [] }));
+        }
+    }
+
     // 3. 进行数据库查询
     let biz_path3 = biz_path.clone();
     let since2 = since;
@@ -3167,8 +3211,15 @@ pub async fn q_biz_articles(
         articles.extend(items);
     }
 
-    // 5. 按 pub_time DESC 排序，取前 N 条
+    // 5. 按 pub_time DESC 排序
     articles.sort_by_key(|a| std::cmp::Reverse(a.pub_time));
+
+    // --unread 语义 A：每个公众号只保留最新 1 篇（已按 pub_time 排序，取首条即可）
+    if unread {
+        let mut seen = std::collections::HashSet::<String>::new();
+        articles.retain(|a| seen.insert(a.account_username.clone()));
+    }
+
     articles.truncate(limit);
 
     let results: Vec<Value> = articles.into_iter().map(|a| {

src/daemon/server.rs

@@ -234,8 +234,8 @@ async fn dispatch(
         ReloadConfig => {
             Response::ok(serde_json::json!({ "reloading": true }))
         }
-        BizArticles { limit, account, since, until } => {
-            match query::q_biz_articles(db, &names_arc, limit, account, since, until).await {
+        BizArticles { limit, account, since, until, unread } => {
+            match query::q_biz_articles(db, &names_arc, limit, account, since, until, unread).await {
                 Ok(v) => Response::ok(v),
                 Err(e) => Response::err(e.to_string()),
             }

src/ipc.rs

@@ -113,6 +113,9 @@ pub enum Request {
         since: Option<i64>,
         #[serde(skip_serializing_if = "Option::is_none")]
         until: Option<i64>,
+        /// 只看有未读消息的公众号，每个公众号取最新 1 篇
+        #[serde(default)]
+        unread: bool,
     },
     /// 朋友圈全文搜索（匹配 contentDesc）
     SnsSearch {