comm-ipc 0.1.13

A lightweight IPC library with bridge, streaming, and FastAPI integration support

CommIPC: Asynchronous IPC for Linux

Installation

pip install comm-ipc

# Or with FastAPI support
pip install "comm-ipc[fastapi]"

Why CommIPC?

CommIPC bridges the gap between simple Unix sockets and complex message brokers like RabbitMQ. It gives you Type-Safe, Asynchronous communication with the ease of a local function call, and now includes seamless FastAPI Integration.

---

Quick Start

This script demonstrates a basic server and client interaction in a single file.

import asyncio
from comm_ipc.server import CommIPCServer
from comm_ipc.client import CommIPC

async def main():
    # Start server
    server = CommIPCServer(socket_path="/tmp/quickstart.sock")
    server_task = asyncio.create_task(server.run())
    await asyncio.sleep(0.1)  # Wait for server to start

    # Start client
    client = CommIPC(socket_path="/tmp/quickstart.sock")
    await client.connect()
    
    channel = await client.open("demo")

    # Register an event
    async def hello_handler(cd):
        return {"msg": f"Hello, {cd.data['name']}"}
    
    await channel.add_event("hello", hello_handler)

    # Call the event
    res = await channel.event("hello", {"name": "World"})
    print(res.data["msg"])

    # Shutdown
    await client.close()
    await server.stop()
    server_task.cancel()

if __name__ == "__main__":
    asyncio.run(main())

---

Connectivity and Transport

CommIPC supports several transport layers:

• Local: Unix Domain Sockets.
• Remote: TCP connections.
• Secure: SSL/TLS encryption for TCP.
• Gateway: Dynamic FastAPI integration for REST/SSE.

Initialization

from comm_ipc.server import CommIPCServer
from comm_ipc.client import CommIPC

# Server setup
server = CommIPCServer(socket_path="/tmp/comm_ipc.sock")
await server.run()

# Client setup
client = CommIPC(socket_path="/tmp/comm_ipc.sock")
await client.connect()

---

System Reference: CommIPCServer

CommIPCServer manages routing and security handshakes.

Constructor: CommIPCServer

• server_id (str): Unique server ID.
• socket_path (str): Path to the Unix socket. Default: /tmp/comm_ipc.sock.
• error_policy (str): Behavior on exceptions ("ignore", "raise", "broadcast").
• connection_secret (str): Secret for HMAC-SHA256 handshakes.
• system_passwords (Dict[str, str]): Pre-set channel passwords.
• channel_policy (str): Behavior on owner disconnect.
  • "terminate": Close the channel (default).
  • "promote": Promote the next earliest member to owner.
• lb_policy (str): Load balancing policy for group events.
  • "least-active": Send to provider with fewest pending calls (default).
  • "round-robin": Cycle through providers sequentially.
• idle_timeout (float): Header read timeout. Default: 60.0.
• data_timeout (float): Body read timeout. Default: 60.0.
• verbose (bool): log output.

Methods

• await run(socket_path, host, port, ssl_context): Start the server on a socket or TCP address.
• await stop(): Close all links and stop the server.

---

System Reference: CommIPC (Client)

CommIPC is the main client interface.

Constructor: CommIPC

• client_id (str): Unique client ID.
• socket_path (str): Path to the Unix socket.
• on_error (Callable): Callback for errors.
• ssl_context: SSL context for TCP.
• connection_secret (str): Handshake secret.
• auto_reconnect (bool): Automatic reconnection. Default: True.
• reconnect_max_tries (int): Retry limit. 0 means unlimited.
• idle_timeout (float): Header read timeout.
• data_timeout (float): Body read timeout.
• heartbeat_interval (float): Ping frequency. Default: 30.0.
• return_type (str): Data format ("dict" or "model"). Default: "dict".
• verbose (bool): Log output.

Methods

• await connect(host, port, ssl_context, connection_secret): Establish a link.
• await open(chan, password): Open a channel and return a CommIPCChannel.
• await set_password(chan, password): Set a channel password (owner only).
• await call(chan, ev, data): Perform a request-response call.
• stream(chan, ev, data): Async iterator for data streams.
• await add_subscription(chan, sub_name, parameters): Register a subscription endpoint.
• await remove_subscription(chan, sub_name): Remove a subscription.
• await subscribe(chan, sub_name, callback): Receive data from a stream.
• await unsubscribe(chan, sub_name): Stop receiving data.
• await publish(chan, sub_name, data): Send data to subscribers. Accepts dict or BaseModel.
• await wait_till_end(): Wait until the connection terminates.
• await close(): Disconnect from the server.
• on_msg: Callback for all incoming messages.

---

System Reference: CommIPCChannel

CommIPCChannel objects handle channel-specific interactions.

Methods

• await add_event(name, call, parameters, returns, is_group): Register an RPC handler. If is_group is True, it registers as a load-balanced provider.
• await add_stream(name, call, parameters): Register a stream handler (automatically detects async generators).
• await event(event_name, data): Call a remote RPC event.
• stream(event_name, data): Collect a remote stream.
• await broadcast(event_name, data): Send to all channel members.
• await send(event_name, data): Send to a provider without a response.
• await add_subscription(sub_name, model): Register a channel subscription schema.
• await remove_subscription(sub_name): Remove a channel subscription.
• await subscribe(sub_name, callback): Channel-level subscription.
• await unsubscribe(sub_name): Channel-level unsubscription.
• await publish(sub_name, data): Channel-level publishing. Accepts dict or BaseModel.
• on_receive(call, event_name): Attach a listener for specific or all messages.
• explore(): List discovered events and subscriptions.
• get_schema(name): Get the Pydantic schema for an endpoint.

---

System Reference: CommIPCApp

CommIPCApp provides a declarative, decorator-based interface for CommIPCChannel.

Constructor: CommIPCApp

• channel (CommIPCChannel, optional): The open channel instance to wrap. If provided, registration happens immediately.

Methods

• await app.register(channel): Binds the app to a channel and registers all buffered handlers/listeners. This is the recommended way to initialize a top-level CommIPCApp.

Decorators

• @app.provide(name, parameters, returns): Registers an asynchronous handler as an event provider. Automatically detects if the handler is an async generator for streaming support.
• @app.on(event_name): Attaches a listener to the channel. If an event_name is provided, it also automatically handles the server-side subscribe() call.
• @app.group.provide(name, parameters, returns): Registers a load-balanced group event provider.
• @app.subscription(name, model): Declares a subscription schema. This is a non-blocking way to ensure the server is aware of the subscription metadata, which is required before calling publish().
• app.group(name: str): Returns a CommIPCAppGroup helper for the specified load-balanced group.

CommIPCAppGroup Reference

• @group.provide(name, parameters, returns): Registers an event provider within the group. Calls to this event will be load-balanced across all providers in the group.

---

System Reference: CommIPCGroup

CommIPCGroup provides an interface for load-balanced event groups. It is accessed via channel.group.

Methods

• __call__(name: str): Returns a CommIPCGroup instance scoped to the specified group name.
• await provide(event, handler, parameters, returns): Registers a provider for an event within this group.
• await get(event, data): Calls an event within the group (load balanced).
• stream(event, data): Returns an async iterator for a grouped stream (load balanced).

---

Example Usage (Decoupled API)

from comm_ipc import CommIPCApp, CommIPC
from comm_ipc.comm_data import CommData
from pydantic import BaseModel
import asyncio

# 1. Define App at top level (FastAPI-style)
app = CommIPCApp()

# 2. Define a data model
class MathParams(BaseModel):
    a: int
    b: int

# 3. Register a provider
@app.provide("add", parameters=MathParams)
async def add(cd: CommData):
    return cd.data.a + cd.data.b

# 4. Register a streaming provider (detected automatically)
@app.provide("counter")
async def count_up(cd: CommData):
    for i in range(5):
        yield {"count": i}

# 5. Register a group provider
@app.group("workers").provide("mult")
async def mult(cd: CommData):
    return cd.data["a"] * cd.data["b"]

# 6. Listen to subscriptions
@app.on("updates")
async def handle_update(cd: CommData):
    print(f"System Status: {cd.data['status']}")

async def main():
    client = CommIPC(socket_path="/tmp/comm_ipc.sock")
    await client.connect()
    chan = await client.open("math")
    
    # 7. Bind and register all handlers
    await app.register(chan)
    
    await client.wait_till_end()

if __name__ == "__main__":
    asyncio.run(main())

---

Load Balancing (Event Groups)

When multiple providers register for the same event, it creates a conflict unless Groups are used. By grouping providers, the server automatically load-balances calls (using a least-active or round-robin policy).

Standard API

# Provider A
await channel.group("workers").add_event("process", handle_a)
 
# Provider B
await channel.group("workers").add_event("process", handle_b)
 
# Consumer
res = await channel.group("workers").event("process", data)

Decorator API

# Create a named group decorator
workers = app.group("workers")
 
@workers.provide("process")
async def handle(cd):
    ...

---

Code Examples

Request-Response (RPC)

import asyncio
from pydantic import BaseModel
from comm_ipc import CommIPC, CommData

class MathParams(BaseModel):
    a: int
    b: int

async def main():
    # 1. Setup client with model support
    client = CommIPC(return_type="model") 
    await client.connect()
    channel = await client.open("math_engine")

    # 2. Register a provider
    async def add_handler(cd: CommData):
        # cd.data is automatically a MathParams instance
        return {"result": cd.data.a + cd.data.b}
    
    await channel.add_event("add", add_handler, parameters=MathParams)

    # 3. Call the provider
    res = await channel.event("add", MathParams(a=10, b=20))
    print(f"Result: {res.data['result']}") 

    await client.close()

if __name__ == "__main__":
    asyncio.run(main())

Publisher-Subscriber

import asyncio
from pydantic import BaseModel
from comm_ipc import CommIPC

class User(BaseModel):
    id: int
    name: str

async def main():
    client = CommIPC(return_type="model")
    await client.connect()
    channel = await client.open("social")

    # 1. Declare and Subscribe
    async def on_user(cd):
        # cd.data is automatically an instance of User
        print(f"New user joined: {cd.data.name}") 

    await channel.subscribe("join_events", on_user)

    # 2. Publish
    # Note: add_subscription is required before publishing to register the schema
    await channel.add_subscription("join_events", model=User)
    await channel.publish("join_events", User(id=1, name="Alice"))
    
    await asyncio.sleep(0.5) # Give it time to arrive
    await client.close()

if __name__ == "__main__":
    asyncio.run(main())

Streaming

import asyncio
from comm_ipc import CommIPC

async def main():
    client = CommIPC()
    await client.connect()
    channel = await client.open("streams")

    # 1. Provider (Async Generator)
    async def count_up(cd):
        for i in range(cd.data["limit"]):
            yield {"count": i}

    await channel.add_stream("counter", count_up)

    # 2. Consumer
    async for chunk in channel.stream("counter", {"limit": 5}):
        print(f"Received: {chunk.data['count']}")

    await client.close()

if __name__ == "__main__":
    asyncio.run(main())

Messaging and Listeners

import asyncio
from comm_ipc import CommIPC

async def main():
    client = CommIPC()
    await client.connect()
    channel = await client.open("monitor")

    async def on_event(cd):
        print(f"Message from {cd.sender_id}: {cd.data}")

    # Listen for specific event
    channel.on_receive(on_event, "alert")
    
    # Broadcast to everyone
    await channel.broadcast("alert", {"msg": "System going down!"})
    
    # Send directed message (no response expected)
    await channel.send("alert", {"msg": "Individual warning"})

    await asyncio.sleep(0.1)
    await client.close()

if __name__ == "__main__":
    asyncio.run(main())

Security and Passwords

import asyncio
from comm_ipc import CommIPC

async def main():
    client = CommIPC()
    await client.connect()

    # 1. Open and set password (Owner)
    channel = await client.open("secure_vault")
    await client.set_password("secure_vault", "secret123")
    
    # 2. Open with password (Member)
    # This would typically be in a separate process
    client2 = CommIPC(client_id="guest")
    await client2.connect()
    try:
        chan2 = await client2.open("secure_vault", password="secret123")
        print("Successfully accessed vault!")
    except Exception as e:
        print(f"Access denied: {e}")

    await client2.close()
    await client.close()

if __name__ == "__main__":
    asyncio.run(main())

FastAPI Integration

from fastapi import FastAPI
from comm_ipc import CommIPC
from comm_ipc.api import CommAPI

app = FastAPI()
client = CommIPC()

@app.on_event("startup")
async def startup():
    await client.connect()
    # Expose IPC events as REST endpoints
    api = CommAPI(app, client)
    api.add_event("math_engine", "add", path="/add", method="POST")

# Now accessible via POST /add

---

System Reference: CommAPI

CommAPI bridges CommIPC with FastAPI, enabling you to expose IPC events as REST/SSE endpoints.

Constructor: CommAPI

• app (FastAPI): Your FastAPI application instance.
• client (CommIPC): A connected CommIPC client.

Methods

• add_event(channel, event_name, path, method, tags): Exposes a specific IPC event at the given path. Automatically generates Pydantic models for request/response validation based on the IPC schema.
• add_resource(channel, path_template, tags, method): Mounts all discovered events of a channel based on a template (e.g., /{event}).

---

System Reference: CommData (Message Object)

CommData models all messages and metadata.

Fields

• sender_id (str): Sender identifier.
• server_id (str): Routing server identifier.
• channel (str): Channel name.
• event (str): Event name or subscription ID.
• data (Any): Message content.
• timestamp (int): Creation timestamp.
• metadata (Dict): Additional data.
• request_id (str): Correlation ID for calls.
• target_id (str): Recipient identifier.
• path (List[str]): Chain of routing servers.
• is_stream (bool): Stream flag.
• is_final (bool): End-of-stream flag.
• signature (str): Message integrity signature.
• origin_server_id (str): First server in the chain.
• sub_name (str): Subscription identifier.

---

System Reference: CommIPCBridge

CommIPCBridge synchronizes two separate hubs.

Constructor: CommIPCBridge

• bridge_id (str): Bridge identifier.
• socket_path1, socket_path2 (str): Local socket paths.
• ssl_context1, ssl_context2: SSL contexts.
• allowed_channels (List[str]): List of channels to sync. Default: all.

Methods

• await connect(target1_params, target2_params): Connect two targets.
• await stop(): Stop the bridge.

from comm_ipc.bridge import CommIPCBridge

bridge = CommIPCBridge(bridge_id="bridge-1")
await bridge.connect(
    target1_params={"socket_path": "/tmp/s1.sock"},
    target2_params={"socket_path": "/tmp/s2.sock"}
)

---

System Reference: System Channels

The server provides read-only channels for monitoring:

• __comm_ipc_logs: Server log broadcast.
• __comm_ipc_errors: Global error broadcast.
• __comm_ipc_system: System event broadcast (e.g., new_registration).

Running Examples & Demos

The repository includes a comprehensive examples/ directory demonstrating various communication patterns.

Quick Start: All Demos

To run the full integration suite and verify your installation:

chmod +x examples/run_demos.sh
./examples/run_demos.sh

Manual Examples

You can run individual examples by starting the server first:

1. Start Server: python examples/server.py
2. Start Provider: python examples/rpc_prov.py
3. Run Client: python examples/rpc_client.py

Other examples include:

• decorator_prov.py / decorator_client.py: Modern decoupled API.
• pub_prov.py / sub_client.py: Publish/Subscribe pattern.
• stream_prov.py / stream_client.py: Async streaming.
• decoupled_demo.py: "FastAPI-style" top-level application structure.
• fastapi_integration.py: Full REST/SSE bridge using CommAPI.

Resilience

Clients recover state automatically if auto_reconnect is True. On reconnection, the client:

1. Re-identifies with the same ID.
2. Re-opens and authenticates active channels.
3. Re-registers all handlers and endpoints.

License

Licensed under the LGPLv3 License.