DiscordClone/convex/voice.ts

"use node";

import { action } from "./_generated/server";
import { internal } from "./_generated/api";
import { v } from "convex/values";
import { AccessToken, RoomServiceClient } from "livekit-server-sdk";
import crypto from "crypto";
import type { Id } from "./_generated/dataModel";

// Shapes of the internal queries below, spelled out so voice.ts can typecheck
// without depending on `internal.auth.*` inference — otherwise the codegen
// cycle (voice exports -> _generated/api -> internal.auth typing -> voice
// usage) drops back to `any` and TS7022/TS7023 fire on the whole action.
type VoiceTokenUser = {
  userId: Id<"userProfiles">;
  username: string;
  publicSigningKey: string;
};
type VoiceTokenChannel = {
  channelId: Id<"channels">;
  type: string;
};
type VoiceTokenResult = { token: string } | { error: string };

// Signed message the client must sign with their Ed25519 private signing key
// to prove they control the userId they're asking a token for. `timestamp`
// defeats replay; the channelId binds the signature to a specific room.
function buildVoiceTokenMessage(
  userId: string,
  channelId: string,
  timestamp: number,
): string {
  return `voice-token:${userId}:${channelId}:${timestamp}`;
}

/**
 * Generate a LiveKit join token for a voice channel.
 *
 * Authorization: caller signs `voice-token:userId:channelId:timestamp` with
 * their Ed25519 signing key. The server verifies against the userProfile's
 * `publicSigningKey`, rejects stale timestamps, and confirms the channel
 * exists and is a voice/dm room. The LiveKit identity is pinned to the
 * server-resolved user so clients can't impersonate each other even with a
 * valid signature for their own account.
 *
 * LiveKit servers run with `room.auto_create: false` reject joins for
 * rooms that don't already exist — the client gets a 404 "requested
 * room does not exist" back from the /rtc/v1/validate endpoint. To
 * make this deployment-agnostic, we pre-create the room via the
 * LiveKit Server SDK before minting the token. When auto-create is
 * enabled the `createRoom` call is idempotent (409 Conflict is
 * swallowed silently), so the same code path works on both
 * configurations.
 *
 * Requires `LIVEKIT_URL` (or the frontend's `VITE_LIVEKIT_URL` as a
 * fallback) in the Convex environment so the RoomServiceClient
 * can talk to the LiveKit API.
 */
export const getToken = action({
  args: {
    channelId: v.id("channels"),
    userId: v.id("userProfiles"),
    timestamp: v.number(),
    signature: v.string(),
  },
  returns: v.union(
    v.object({ token: v.string() }),
    v.object({ error: v.string() }),
  ),
  handler: async (ctx, args): Promise<VoiceTokenResult> => {
    // Reject timestamps outside a 5-minute window on either side of the
    // server clock. Matches the recovery action's window and keeps replay
    // attempts short-lived even if a signature leaks.
    const now = Date.now();
    if (!Number.isFinite(args.timestamp) || Math.abs(now - args.timestamp) > 5 * 60 * 1000) {
      return { error: "Request expired. Please try again." };
    }

    const user: VoiceTokenUser | null = await ctx.runQuery(
      internal.auth.getUserForVoiceToken,
      { userId: args.userId },
    );
    if (!user) {
      return { error: "User not found" };
    }

    const channel: VoiceTokenChannel | null = await ctx.runQuery(
      internal.auth.getChannelForVoiceToken,
      { channelId: args.channelId },
    );
    if (!channel) {
      return { error: "Channel not found" };
    }
    if (channel.type !== "voice" && channel.type !== "dm") {
      return { error: "Not a voice channel" };
    }

    // Verify the caller actually controls `userId` by checking their
    // signature over (userId, channelId, timestamp).
    const message = buildVoiceTokenMessage(user.userId, channel.channelId, args.timestamp);
    try {
      const publicKeyObj = crypto.createPublicKey({
        key: user.publicSigningKey,
        format: "pem",
        type: "spki",
      });
      const isValid = crypto.verify(
        null,
        Buffer.from(message),
        publicKeyObj,
        Buffer.from(args.signature, "hex"),
      );
      if (!isValid) {
        return { error: "Invalid signature" };
      }
    } catch {
      return { error: "Signature verification failed" };
    }

    const apiKey = process.env.LIVEKIT_API_KEY || "devkey";
    const apiSecret = process.env.LIVEKIT_API_SECRET || "secret";
    const livekitUrl =
      process.env.LIVEKIT_URL || process.env.VITE_LIVEKIT_URL || "";

    // Ensure the room exists. The LiveKit API accepts `http(s)` URLs
    // for the management endpoint, but the frontend connect URL is a
    // `wss://` — swap the scheme when needed.
    if (livekitUrl) {
      const httpUrl = livekitUrl
        .replace(/^wss:\/\//i, "https://")
        .replace(/^ws:\/\//i, "http://");
      try {
        const roomService = new RoomServiceClient(httpUrl, apiKey, apiSecret);
        await roomService.createRoom({
          name: channel.channelId,
          // Empty rooms auto-destroy after 5 minutes with no participants,
          // matching LiveKit's own default so stale rooms from a crashed
          // client don't pile up forever.
          emptyTimeout: 5 * 60,
          // 50 participants is plenty for a voice channel in this
          // single-server deployment and keeps any runaway join loop
          // from hitting the global limit.
          maxParticipants: 50,
        });
      } catch (err: any) {
        // 409 / "already exists" is expected when a room has already
        // been created by an earlier join — swallow it and continue.
        const errMsg = String(err?.message ?? err ?? "");
        const status = err?.status ?? err?.statusCode;
        const alreadyExists =
          status === 409 ||
          /already exists/i.test(errMsg) ||
          /AlreadyExists/i.test(errMsg);
        if (!alreadyExists) {
          // Non-fatal: log and fall through to token generation. If the
          // real issue was misconfiguration the client will surface the
          // 404 it already does.
          console.warn("LiveKit createRoom failed:", errMsg);
        }
      }
    }

    // Pin identity + name to the server-resolved user. A forged `username`
    // or `userId` in the args would have already been rejected by the
    // signature check, but using the DB values is defence-in-depth.
    const at: AccessToken = new AccessToken(apiKey, apiSecret, {
      identity: user.userId,
      name: user.username,
      ttl: "24h",
    });

    at.addGrant({
      roomJoin: true,
      room: channel.channelId,
      canPublish: true,
      canSubscribe: true,
      canPublishData: true,
    });

    const token: string = await at.toJwt();
    return { token };
  },
});