SB-0MNP9QU9S0078KAT: add non-text attachment rejection test and update README to document attachment ingestion

Author: Sorra

README.md

@@ -30,6 +30,10 @@ SourceBase automatically extracts URLs from Discord messages, fetches content me
 - Behavior and limits:
   - The bot writes the provided text to a temporary file and calls the OpenBrain CLI (`ob add`) with a `file://` URL so the existing CLI-based ingestion pipeline is reused.
   - To avoid abuse, the bot enforces a conservative default size limit of 64 KiB for direct text ingestion. You can override this limit with the environment variable `OB_ADD_MAX_BYTES` (value in bytes).
+  - Attachment support:
+    - You can reply to a message that contains a text-like file attachment (for example `.md`, `.markdown`, `.txt`) with `ob add` and the bot will fetch the attachment body and ingest it.
+    - The bot accepts files with a text/* Content-Type or filenames ending in `.md`, `.markdown`, or `.txt`. Binary files (e.g. `application/octet-stream`) are rejected with a helpful error.
+    - The same size limit (default 64 KiB, configurable via `OB_ADD_MAX_BYTES`) applies to attachments; oversized attachments are rejected with a clear message.
   - If the bot cannot fetch the referenced message (reply flow), it will reply with a helpful message explaining the permission issue and how to proceed.
   - If the CLI is unavailable, the bot will notify the user with a friendly error message.
 

tests/index.ob-add.test.ts

@@ -273,6 +273,37 @@ describe("ob add message handler failure modes", () => {
     expect(replies[0]).toContain("too large to ingest");
   });
 
+  it("rejects non-text/binary attachment from referenced message", async () => {
+    const onMonitoredMessage = await loadMonitoredMessageHandler(async () => {
+      await vi.doMock("../src/bot/cli-runner.js", async (importOriginal) => {
+        const actual: any = await importOriginal();
+        return {
+          ...actual,
+          runAddCommand: vi.fn(() => createAddGenerator([], { success: false, error: "", url: "", id: undefined } as any)),
+          runCliCommand: vi.fn(),
+          runQueueCommand: vi.fn(),
+          runSummaryCommand: vi.fn(),
+          runStatsCommand: vi.fn(async () => ({ totalLinks: 0, processedCount: 0, pendingCount: 0, failedCount: 0 })),
+          isCliAvailable: vi.fn(async () => true),
+          CliRunnerError: class MockCliRunnerError extends Error {},
+        };
+      });
+    });
+
+    // Mock fetch to return a non-text Content-Type
+    global.fetch = vi.fn(async () => ({ ok: true, headers: { get: (_: string) => "application/octet-stream" }, text: async () => "\u0000\u0001binary" } as any));
+
+    const { message, replies } = createFakeMessage("ob add");
+    const att = { url: "https://example.test/file.bin", name: "file.bin" };
+    message.reference = { messageId: "ref-bin" };
+    message.channel = { messages: { fetch: vi.fn().mockResolvedValue({ content: "", attachments: { size: 1, first: () => att } }) } };
+
+    await onMonitoredMessage(message);
+
+    expect(replies.length).toBeGreaterThan(0);
+    expect(replies[0]).toContain("does not appear to be a text file");
+  });
+
   it("handles inline attachment on same message as ob add", async () => {
     const onMonitoredMessage = await loadMonitoredMessageHandler(async () => {
       await vi.doMock("../src/bot/cli-runner.js", async (importOriginal) => {