Troubleshooting¶

“Failed to load default platform”¶

Check that the adapter for your target is on the classpath, then initialize it before creating Nostr4J objects. Having only the core library is not enough; see platform setup.

Local relay or wallet URL is rejected¶

Loopback destinations are blocked by default. For JVM development only, start with -Dnge-platforms.allowLoopbackInURIs=true. Avoid this flag in production, unless you really know what you're doing.

A fetch returns fewer events than the limit¶

The limit only caps the number of results, so getting fewer is normal: the timeout may have expired, the filter may have matched little data, or several relays may have returned the same events. If you enabled stopping at EOSE, the fetch also returns when the relays finish their stored results. Check your fetch policy before increasing the timeout.

NWC never becomes ready¶

Check that the URI parses, at least one advertised relay is reachable, the wallet service is online and the capability has not been revoked. Use waitForReady() before balance or payment methods and close the wallet after a failed attempt.

RTC peer discovered, but no data arrives¶

Discovery means Nostr signaling worked; it does not prove the RTC data channel opened. Check whether both peers can exchange ICE candidates, whether the configured TURN service is reachable, and whether the signaling relays accept topology events. Restrictive NATs can prevent a direct connection. A successful TURN WebSocket upgrade alone does not prove its authenticated virtual-socket handshake completed.

A TURN WebSocket close code of 1011 indicates an internal error at the endpoint. Check the server logs, or try another compatible service.

Also check that every participant uses matching application and protocol IDs in RTCSettings, the same room key and compatible protocol versions. For TURN writes, a reachable server must also implement the client's authenticated delivery-receipt format.

Broadcast reports no connected mutually attested graph¶

Peer discovery can finish before the room has usable physical links and an agreed routing topology. No connected mutually attested graph for broadcast means the current graph cannot support that broadcast; a list of discovered peers is not enough.

Inspect room.getConnectionDiagnostics() for pending, established and closing resources, then check the topology and transport callbacks. Verify ICE connectivity and the TURN handshake rather than repeatedly publishing discovery announcements. Handle the broadcast task's failure, and send fresh application state after connectivity recovers.

A stream read waits indefinitely¶

Both participants must call connect() with the other's public key and the same connection ID. Finish exchanging replies before closing the output: EOF ends the remote connection. readAllBytes() waits for EOF; use a framed read such as readUTF() during an open conversation. Run blocking operations away from your UI thread; see Simple P2P streams.

SLF4J reports that no provider was found¶

Choose a matching SLF4J provider in your application when you need those logs. Nostr4J does not bundle or force a provider. Include one provider, such as slf4j-jdk14 for Java logging, rather than several conflicting providers.

TeaVM compilation fails before analysis¶

Start by checking that the platform dependencies resolve to compatible versions. For browser builds, choose the task for your output target in platform setup. See the TeaVM documentation for more.