feat(kb): ingest surfaces — web paste bar, chat chip, API route, lisa kb CLI (K-G)

src/cli-args.test.ts

@@ -24,6 +24,12 @@ describe("parseArgs — raw / passthrough subcommand routing", () => {
     assert.equal(a.port, 5757);
   });
 
+  test("kb: passthrough — its --title/--tags/--force flags reach the handler verbatim", () => {
+    const a = parseArgs(["kb", "add", "https://x.dev/a", "--title", "T", "--tags", "a,b", "--force"]);
+    assert.equal(a.subcommand, "kb");
+    assert.deepEqual(a.subargs, ["add", "https://x.dev/a", "--title", "T", "--tags", "a,b", "--force"]);
+  });
+
   test("autostart: recognized global flags are parsed into the global fields (not swallowed)", () => {
     const a = parseArgs([
       "autostart", "install",

src/cli-args.ts

@@ -41,6 +41,7 @@ export interface ParsedArgs {
     | "agents"
     | "pair"
     | "mail"
+    | "kb"
     | "login"
     | "logout"
     | "billing";
@@ -67,7 +68,7 @@ const RAW_SUBCOMMANDS = new Set(["heartbeat", "autostart"]);
  * global flags (`mail connect --host/--port/--provider …`), which would
  * otherwise be swallowed as global settings and never reach the handler.
  */
-const PASSTHROUGH_SUBCOMMANDS = new Set(["mail"]);
+const PASSTHROUGH_SUBCOMMANDS = new Set(["mail", "kb"]);
 
 export function parseArgs(argv: string[]): ParsedArgs {
   const out: ParsedArgs = {
@@ -194,6 +195,7 @@ export function parseArgs(argv: string[]): ParsedArgs {
       first === "agents" ||
       first === "pair" ||
       first === "mail" ||
+      first === "kb" ||
       first === "login" ||
       first === "logout" ||
       first === "billing"

src/cli.ts

@@ -105,6 +105,14 @@ SKILLS (executable, Phase 3.1)
   lisa skills enable <slug>    Remove a disable flag.
   lisa skills audit <slug>     Show the audit trail.
 
+KNOWLEDGE BASE
+  lisa kb add <url> [--title T] [--tags a,b] [--force]
+                               Save a web page (incl. WeChat / Bilibili /
+                               YouTube) into the KB with provenance.
+  lisa kb list [wiki|sources]  List entries, newest first.
+  lisa kb search "<query>"     Search sources + wiki (TF-IDF).
+  lisa kb brief [YYYY-MM-DD]   Print a daily feeds brief (needs kb/feeds.json).
+
   lisa --help                  Show this message.
   lisa --version               Print the installed Lisa version.
 
@@ -352,6 +360,11 @@ async function main(): Promise<void> {
     process.exit(await runMailCommand(args.subargs));
   }
 
+  if (args.subcommand === "kb") {
+    const { runKbCommand } = await import("./cli/kb.js");
+    process.exit(await runKbCommand(args.subargs));
+  }
+
   if (args.subcommand === "sessions") {
     const sessions = await listSessionsOnDisk();
     for (const s of sessions) {

src/cli/kb.test.ts

@@ -0,0 +1,88 @@
+import { test, describe, after, beforeEach } from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, rmSync } from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const TMP = mkdtempSync(path.join(os.tmpdir(), "lisa-cli-kb-"));
+process.env.LISA_HOME = TMP;
+process.env.LISA_KB_NO_GIT = "1";
+
+const { runKbCommand } = await import("./kb.js");
+const store = await import("../kb/store.js");
+
+after(() => rmSync(TMP, { recursive: true, force: true }));
+
+// Capture console output per run — the CLI's contract IS its printed lines.
+let out: string[] = [];
+let err: string[] = [];
+const origLog = console.log;
+const origErr = console.error;
+beforeEach(() => {
+  out = [];
+  err = [];
+  console.log = (...a: unknown[]) => out.push(a.join(" "));
+  console.error = (...a: unknown[]) => err.push(a.join(" "));
+});
+after(() => {
+  console.log = origLog;
+  console.error = origErr;
+});
+
+describe("lisa kb CLI", () => {
+  test("bad / missing subcommand prints usage", async () => {
+    assert.equal(await runKbCommand(["bogus"]), 2);
+    assert.match(err.join("\n"), /usage: lisa kb add/);
+  });
+
+  test("add without a url errors with usage", async () => {
+    assert.equal(await runKbCommand(["add"]), 2);
+    assert.match(err.join("\n"), /lisa kb add <url>/);
+  });
+
+  test("list: empty KB, then entries with layer/date/tags", async () => {
+    assert.equal(await runKbCommand(["list"]), 0);
+    assert.match(out.join("\n"), /knowledge base is empty/);
+
+    await store.addSource({ title: "Meeting", body: "notes", tags: ["work"] });
+    await store.writeWiki({ title: "Projects", body: "current work" });
+    out = [];
+    assert.equal(await runKbCommand(["list"]), 0);
+    const text = out.join("\n");
+    assert.match(text, /wiki .*projects.*Projects/);
+    assert.match(text, /source .*meeting.*Meeting.*#work/);
+
+    out = [];
+    assert.equal(await runKbCommand(["list", "wiki"]), 0);
+    assert.doesNotMatch(out.join("\n"), /Meeting/);
+  });
+
+  test("search hits and misses", async () => {
+    assert.equal(await runKbCommand(["search", "projects"]), 0);
+    assert.match(out.join("\n"), /\[wiki\/projects\]/);
+    out = [];
+    assert.equal(await runKbCommand(["search", "zzznotoken"]), 0);
+    assert.match(out.join("\n"), /no matches/);
+  });
+
+  test("brief: none yet → pointer at feeds.json; existing brief prints", async () => {
+    assert.equal(await runKbCommand(["brief"]), 0);
+    assert.match(out.join("\n"), /no briefs yet.*feeds\.json/);
+
+    // K-H writes briefs as sources/brief-<date>.md; simulate one.
+    const { atomicWrite } = await import("../fs-utils.js");
+    const { entryFile } = await import("../kb/paths.js");
+    await atomicWrite(
+      entryFile("sources", "brief-2026-07-23"),
+      "---\ntitle: Daily brief 2026-07-23\ncreated: 2026-07-23T08:00:00Z\norigin: brief\n---\n\n- top story one\n",
+    );
+    out = [];
+    assert.equal(await runKbCommand(["brief"]), 0);
+    assert.match(out.join("\n"), /Daily brief 2026-07-23/);
+    assert.match(out.join("\n"), /top story one/);
+
+    out = [];
+    assert.equal(await runKbCommand(["brief", "2026-01-01"]), 0);
+    assert.match(out.join("\n"), /no brief for 2026-01-01/);
+  });
+});

src/cli/kb.ts

@@ -0,0 +1,134 @@
+/**
+ * `lisa kb <add|list|search|brief>` — the knowledge base from the terminal.
+ *
+ *   lisa kb add <url> [--title T] [--tags a,b] [--force]   Ingest a link (Layer 1)
+ *   lisa kb list [wiki|sources]                            Entries, newest first
+ *   lisa kb search <query>                                 TF-IDF search
+ *   lisa kb brief [date]                                   Print a daily brief (K-H)
+ *
+ * Mirrors the mail subcommand shape: full passthrough args (cli-args.ts), the
+ * handler owns all of its flags.
+ */
+
+function parseFlags(args: string[]): { flags: Record<string, string>; rest: string[] } {
+  const flags: Record<string, string> = {};
+  const rest: string[] = [];
+  for (let i = 0; i < args.length; i++) {
+    const a = args[i]!;
+    if (!a.startsWith("--")) {
+      rest.push(a);
+      continue;
+    }
+    const key = a.slice(2);
+    const next = args[i + 1];
+    if (next && !next.startsWith("--")) {
+      flags[key] = next;
+      i++;
+    } else {
+      flags[key] = "true";
+    }
+  }
+  return { flags, rest };
+}
+
+const USAGE =
+  "usage: lisa kb add <url> [--title T] [--tags a,b] [--force]\n" +
+  "       lisa kb list [wiki|sources]\n" +
+  "       lisa kb search <query>\n" +
+  "       lisa kb brief [YYYY-MM-DD]";
+
+export async function runKbCommand(args: string[]): Promise<number> {
+  const [sub, ...tail] = args;
+  switch (sub) {
+    case "add": {
+      const { flags, rest } = parseFlags(tail);
+      const url = rest[0];
+      if (!url) {
+        console.error("usage: lisa kb add <url> [--title T] [--tags a,b] [--force]");
+        return 2;
+      }
+      const { ingestUrl } = await import("../kb/ingest/index.js");
+      try {
+        const res = await ingestUrl(url, {
+          title: flags.title,
+          tags: flags.tags ? flags.tags.split(",").map((t) => t.trim()).filter(Boolean) : undefined,
+          force: flags.force === "true",
+        });
+        if (res.deduped) {
+          console.log(`already saved: sources/${res.entry.slug}  "${res.entry.title}" (--force to re-capture)`);
+        } else {
+          console.log(`saved: sources/${res.entry.slug}  "${res.entry.title}"  via ${res.via}`);
+          const t = res.entry.extra?.transcript;
+          if (t?.startsWith("unavailable")) {
+            console.log(`note: transcript ${t} — metadata + description captured; paste a transcript with kb_add if you have one`);
+          }
+        }
+        return 0;
+      } catch (err) {
+        console.error(`ingest failed: ${(err as Error).message}`);
+        return 1;
+      }
+    }
+    case "list": {
+      const layer = tail[0] === "wiki" || tail[0] === "sources" ? tail[0] : undefined;
+      const { listEntries } = await import("../kb/store.js");
+      const entries = await listEntries(layer);
+      if (entries.length === 0) {
+        console.log("(knowledge base is empty — `lisa kb add <url>` or capture from chat)");
+        return 0;
+      }
+      for (const e of entries) {
+        const date = (e.updated ?? e.created ?? "").slice(0, 10);
+        const tags = e.tags.length ? `  #${e.tags.join(" #")}` : "";
+        console.log(`${e.layer === "wiki" ? "wiki  " : "source"} ${date}  ${e.slug}  ${e.title}${tags}`);
+      }
+      return 0;
+    }
+    case "search": {
+      const query = tail.join(" ").trim();
+      if (!query) {
+        console.error("usage: lisa kb search <query>");
+        return 2;
+      }
+      const { searchKb } = await import("../kb/search.js");
+      const hits = await searchKb(query, 10);
+      if (hits.length === 0) {
+        console.log("(no matches)");
+        return 0;
+      }
+      for (const h of hits) {
+        console.log(`[${h.layer}/${h.slug}] ${h.title} (score=${h.score.toFixed(2)})\n  ${h.excerpt}\n`);
+      }
+      return 0;
+    }
+    case "brief": {
+      // Daily briefs are written by the feeds service (K-H) as
+      // sources/brief-<date>.md — searchable, linkable Layer-1 entries.
+      const { listEntries, readEntry } = await import("../kb/store.js");
+      const wanted = tail[0]; // optional YYYY-MM-DD
+      const briefs = (await listEntries("sources")).filter((e) =>
+        wanted ? e.slug.startsWith(`brief-${wanted}`) : /^brief-\d{4}-\d{2}-\d{2}/.test(e.slug),
+      );
+      const target = briefs[0];
+      if (!target) {
+        console.log(
+          wanted
+            ? `(no brief for ${wanted})`
+            : "(no briefs yet — add feeds to ~/.lisa/kb/feeds.json to enable the daily brief)",
+        );
+        return 0;
+      }
+      const entry = await readEntry("sources", target.slug);
+      if (!entry) {
+        console.log(`(brief ${target.slug} unreadable)`);
+        return 1;
+      }
+      console.log(`# ${entry.title}\n`);
+      console.log(entry.body);
+      return 0;
+    }
+    default:
+      console.error(USAGE);
+      return sub ? 2 : 0;
+  }
+}

src/web/lisa-client.ts

@@ -1126,9 +1126,46 @@ async function send(message) {
   const filesToSend = [...pendingFiles];
   pendingFiles = [];
   renderAttachPreview();
+  maybeOfferKbIngest(message);
   await runChat(message, filesToSend);
 }
 
+// ── chat → KB: a bare URL in the user's message gets a one-tap 存入知识库
+//    chip under the bubble; it calls the same /api/kb/ingest the KB view uses.
+function maybeOfferKbIngest(message) {
+  if (!message) return;
+  var m = message.match(/https?:\\/\\/[^\\s<>"')\\]]+/);
+  if (!m) return;
+  var url = m[0].replace(/[.,;:!?。，；：、]+$/, '');
+  var chip = el('div', 'kb-ingest-chip', null);
+  var btn = document.createElement('button');
+  btn.type = 'button';
+  btn.className = 'kb-ingest-btn';
+  btn.textContent = '💾 存入知识库';
+  chip.appendChild(btn);
+  btn.addEventListener('click', function () {
+    btn.disabled = true;
+    btn.textContent = '保存中…';
+    fetch('/api/kb/ingest', { method: 'POST', headers: { 'content-type': 'application/json' }, body: JSON.stringify({ url: url }) })
+      .then(function (r) { return r.json(); })
+      .then(function (d) {
+        if (d && d.ok) {
+          btn.textContent = d.deduped ? '已在知识库 ✓' : '已存入知识库 ✓';
+          if (typeof window.lisaReloadKb === 'function') window.lisaReloadKb();
+        } else {
+          btn.disabled = false;
+          btn.textContent = '💾 存入知识库';
+          if (typeof window.lisaKbToast === 'function') window.lisaKbToast((d && d.error) ? d.error : '保存失败');
+        }
+      })
+      .catch(function () {
+        btn.disabled = false;
+        btn.textContent = '💾 存入知识库';
+        if (typeof window.lisaKbToast === 'function') window.lisaKbToast('保存失败');
+      });
+  });
+}
+
 // On failure, show the error detail with a retry button that re-runs the same
 // turn. Kept separate from send() so retry never re-appends the user's bubble
 // or re-reads the (already-cleared) attachment tray.
@@ -1353,6 +1390,8 @@ input.addEventListener('input', () => {
     setTimeout(function () { t.classList.add('show'); }, 10);
     setTimeout(function () { t.classList.remove('show'); setTimeout(function () { t.remove(); }, 300); }, 2200);
   }
+  // Shared with the chat-bubble ingest chip (defined outside this closure).
+  window.lisaKbToast = kbToast;
 })();
 
 // ── PWA: register service worker + iOS install hint ─────────────────
@@ -2683,16 +2722,54 @@ if ('serviceWorker' in navigator) {
       });
     }
   }
+  function kbIngestSubmit() {
+    var urlEl = document.getElementById('kbIngestUrl');
+    var go = document.getElementById('kbIngestGo');
+    var status = document.getElementById('kbIngestStatus');
+    if (!urlEl || !go) return;
+    var url = (urlEl.value || '').trim();
+    if (!url) return;
+    go.disabled = true;
+    urlEl.disabled = true;
+    if (status) { status.className = 'kb-ingest-status'; status.textContent = 'Fetching & extracting…'; }
+    fetch('/api/kb/ingest', { method: 'POST', headers: { 'content-type': 'application/json' }, body: JSON.stringify({ url: url }) })
+      .then(function (r) { return r.json(); })
+      .then(function (d) {
+        go.disabled = false;
+        urlEl.disabled = false;
+        if (d && d.ok) {
+          urlEl.value = '';
+          var note = d.deduped ? 'Already saved — opening it' : 'Saved via ' + (d.via || 'generic');
+          if (d.transcript && String(d.transcript).indexOf('unavailable') === 0) note += ' (no transcript — you can paste one)';
+          if (status) status.textContent = note;
+          loadKbList();
+          if (d.entry) kbOpen(d.entry.layer, d.entry.slug);
+        } else if (status) {
+          status.className = 'kb-ingest-status err';
+          status.textContent = (d && d.error) ? d.error : 'Ingest failed';
+        }
+      })
+      .catch(function () {
+        go.disabled = false;
+        urlEl.disabled = false;
+        if (status) { status.className = 'kb-ingest-status err'; status.textContent = 'Ingest failed (network)'; }
+      });
+  }
   function loadKb() {
     views.kb.innerHTML =
       '<div class="view-head"><div><h2>Knowledge Base</h2><div class="vh-sub">Sources + wiki · live search</div></div></div>'
+      + '<div class="kb-ingestbar"><input id="kbIngestUrl" class="kb-ingest-url" type="url" placeholder="Paste a link to save — WeChat · Bilibili · YouTube · any article" autocomplete="off"><button type="button" id="kbIngestGo" class="kb-ingest-go">Save</button><span id="kbIngestStatus" class="kb-ingest-status"></span></div>'
       + '<div class="kb-searchbar"><input id="kbSearch" class="kb-search" type="text" placeholder="Search your knowledge base…" autocomplete="off"></div>'
       + '<div class="kb-body"><div id="kbList" class="kb-list"></div><div id="kbReader" class="kb-reader"></div></div>';
     var s = document.getElementById('kbSearch');
     if (s) s.addEventListener('input', function () {
       if (kbSearchTimer) clearTimeout(kbSearchTimer);
       kbSearchTimer = setTimeout(loadKbList, 200);
     });
+    var go = document.getElementById('kbIngestGo');
+    if (go) go.addEventListener('click', kbIngestSubmit);
+    var u = document.getElementById('kbIngestUrl');
+    if (u) u.addEventListener('keydown', function (e) { if (e.key === 'Enter') kbIngestSubmit(); });
     loadKbList();
   }
   window.lisaReloadKb = loadKbList;