Ivan FB 7080a629da
feat: migrate the FFI layer to nim-ffi 0.2.0
nim-ffi 0.2.0 reshapes the authoring model: `.ffi.` procs take the
library value plus typed params instead of threading (ctx, callback,
userData) by hand, the macro validates the context itself, and payloads
ride the wire as CBOR rather than ad-hoc JSON strings. The old idiom no
longer compiles against it, so the whole surface moves at once.

Proc names are camelCase chosen so the generated snake_case export
matches the previous C symbol exactly (wakuRelayPublish ->
waku_relay_publish), keeping the ABI names stable.

Node lifecycle now uses the dedicated pragmas: `.ffiCtor.` for
create_node (LogosDelivery.new already returns the Future[Result[...]]
the contract wants) and `.ffiDtor.` for destroy. Contexts come from the
macro-emitted FFIContextPool, which caps live contexts at 32.

Events become typed `.ffiEvent.` procs over `.ffi.` payload objects. The
payloads carry wire-friendly scalars rather than the domain types, which
are not serialisable; byte fields stay base64. This is what makes the
generated bindings emit typed listeners instead of leaving consumers to
register by name and parse JSON themselves.

The payload fields are deliberately unexported. genBindings copies field
names verbatim, so an export marker leaks into the generated Rust as
`pub payload*: String` and the file does not parse -- a nim-ffi bug (it
strips the marker from type names but not fields, so its single-file
examples never hit it). Construction therefore lives behind the emit*
procs in declare_lib, which also keeps event emission in one place and
collapses each listener body to a single call.

`requireInitializedNode` is gone: the macro rejects a null/invalid ctx
before the handler runs, so all 14 call sites were redundant.

Relay and filter push handlers are declared `raises: [Defect]`, so the
emit call is wrapped explicitly -- the dispatch path no longer guards
the body for us.

genBindings() emits the C/C++/Rust bindings and must stay last in the
compilation root; it is a no-op without -d:ffiGenBindings. The Rust
output is checked in so consumers can vendor it directly.

Known gaps, tracked separately: the hand-written liblogosdelivery.h /
_kernel.h still declare the pre-CBOR signatures and need generating or
dropping, and nimble resolves cbor_serialization 0.4.0 while the lock
and nim-ffi both pin 0.3.0.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-16 00:55:59 +02:00

7.2 KiB

Logos Messaging API (LMAPI) Library

A C FFI library providing a simplified interface to Logos Messaging functionality.

Overview

This library wraps the high-level API functions from waku/api/api.nim and exposes them via a C FFI interface, making them accessible from C, C++, and other languages that support C FFI.

API Functions

Node Lifecycle

logosdelivery_create_node

Creates a new instance of the node from the given configuration JSON.

void *logosdelivery_create_node(
    const char *configJson,
    FFICallBack callback,
    void *userData
);

Parameters:

  • configJson: JSON string containing node configuration
  • callback: Callback function to receive the result
  • userData: User data passed to the callback

Returns: Pointer to the context needed by other API functions, or NULL on error.

Example configuration JSON:

{
  "mode": "Core",
  "preset": "logos.dev",
  "messagingOverrides": {
    "listen-address": "0.0.0.0",
    "tcp-port": 60000,
    "discv5-udp-port": 9000
  }
}

The configuration object has four optional top-level keys: mode ("Core" or "Edge", defaults to "Core"), preset, messagingOverrides (per-field node config overrides), and channelsOverrides (reliable-channel overrides). Override keys accept the config field name or its CLI switch name (e.g. "clusterId" or "cluster-id"); unknown keys are rejected. Use "preset" to select a network preset (e.g., "twn", "logos.dev", "status.prod") which auto-configures entry nodes, cluster ID, sharding, and other network-specific settings.

Available presets:

Preset Cluster ID RLN Sharding Network
twn 1 on auto (8 shards) The Waku Network
logos.dev 2 off auto (8 shards) Logos Dev Network
logos.test 2 off auto (8 shards) Logos Test Network
status.prod 16 off auto (1 shard) Status Production Network

logosdelivery_start_node

Starts the node.

int logosdelivery_start_node(
    void *ctx,
    FFICallBack callback,
    void *userData
);

logosdelivery_stop_node

Stops the node.

int logosdelivery_stop_node(
    void *ctx,
    FFICallBack callback,
    void *userData
);

logosdelivery_destroy

Destroys a node instance and frees resources.

int logosdelivery_destroy(
    void *ctx,
    FFICallBack callback,
    void *userData
);

Messaging

logosdelivery_subscribe

Subscribe to a content topic to receive messages.

int logosdelivery_subscribe(
    void *ctx,
    FFICallBack callback,
    void *userData,
    const char *contentTopic
);

Parameters:

  • ctx: Context pointer from logosdelivery_create_node
  • callback: Callback function to receive the result
  • userData: User data passed to the callback
  • contentTopic: Content topic string (e.g., "/myapp/1/chat/proto")

logosdelivery_unsubscribe

Unsubscribe from a content topic.

int logosdelivery_unsubscribe(
    void *ctx,
    FFICallBack callback,
    void *userData,
    const char *contentTopic
);

logosdelivery_send

Send a message.

int logosdelivery_send(
    void *ctx,
    FFICallBack callback,
    void *userData,
    const char *messageJson
);

Parameters:

  • messageJson: JSON string containing the message

Example message JSON:

{
  "contentTopic": "/myapp/1/chat/proto",
  "payload": "SGVsbG8gV29ybGQ=",
  "ephemeral": false
}

Note: The payload field should be base64-encoded.

Returns: Request ID in the callback message that can be used to track message delivery.

Events

logosdelivery_add_event_listener

Registers a callback invoked whenever the named event fires (e.g., message received). Listeners are registered per event name, so register once for each event you want to observe.

uint64_t logosdelivery_add_event_listener(
    void *ctx,
    const char *eventName,
    FFICallBack callback,
    void *userData
);

Returns: The listener id (> 0), or 0 if callback is NULL.

Event names: onMessageSent, onMessageError, onMessagePropagated, onMessageReceived, onConnectionStatusChange, onTopicHealthChange, onConnectionChange, onChannelMessageReceived, onChannelMessageSent, onChannelMessageError, onReceivedMessage.

logosdelivery_remove_event_listener

Unregisters a previously added listener.

int logosdelivery_remove_event_listener(
    void *ctx,
    uint64_t listenerId
);

Returns: RET_OK when a listener was removed, RET_ERR otherwise.

Important: The callback should be fast, non-blocking, and thread-safe.

Building

The library follows the same build system as the main Logos Messaging project.

Build the library

make liblogosdeliveryStatic    # Build static library
# or
make liblogosdeliveryDynamic   # Build dynamic library

Return Codes

All functions that return int use the following return codes:

  • RET_OK (0): Success
  • RET_ERR (1): Error
  • RET_MISSING_CALLBACK (2): Missing callback function

Callback Function

All API functions use the following callback signature:

typedef void (*FFICallBack)(
    int callerRet,
    const char *msg,
    size_t len,
    void *userData
);

Parameters:

  • callerRet: Return code (RET_OK, RET_ERR, etc.)
  • msg: Response message (may be empty for success)
  • len: Length of the message
  • userData: User data passed in the original call

Example Usage

#include "liblogosdelivery.h"
#include <stdio.h>

void callback(int ret, const char *msg, size_t len, void *userData) {
    if (ret == RET_OK) {
        printf("Success: %.*s\n", (int)len, msg);
    } else {
        printf("Error: %.*s\n", (int)len, msg);
    }
}

int main() {
    const char *config = "{"
        "\"logLevel\": \"INFO\","
        "\"mode\": \"Core\","
        "\"preset\": \"logos.dev\""
        "}";

    // Create node
    void *ctx = logosdelivery_create_node(config, callback, NULL);
    if (ctx == NULL) {
        return 1;
    }

    // Start node
    logosdelivery_start_node(ctx, callback, NULL);

    // Subscribe to a topic
    logosdelivery_subscribe(ctx, callback, NULL, "/myapp/1/chat/proto");

    // Send a message
    const char *msg = "{"
        "\"contentTopic\": \"/myapp/1/chat/proto\","
        "\"payload\": \"SGVsbG8gV29ybGQ=\","
        "\"ephemeral\": false"
        "}";
    logosdelivery_send(ctx, callback, NULL, msg);

    // Clean up
    logosdelivery_stop_node(ctx, callback, NULL);
    logosdelivery_destroy(ctx, callback, NULL);

    return 0;
}

Architecture

The library is structured as follows:

  • liblogosdelivery.h: C header file with function declarations
  • liblogosdelivery.nim: Main library entry point
  • declare_lib.nim: Library declaration and initialization
  • lmapi/node_api.nim: Node lifecycle API implementation
  • lmapi/messaging_api.nim: Subscribe/send API implementation

The library uses the nim-ffi framework for FFI infrastructure, which handles:

  • Thread-safe request processing
  • Async operation management
  • Memory management between C and Nim
  • Callback marshaling

See Also

  • Main API documentation: waku/api/api.nim
  • Original libwaku library: library/libwaku.nim
  • nim-ffi framework: vendor/nim-ffi/