diff --git a/gamutrf/grinference2mqtt.py b/gamutrf/grinference2mqtt.py
index 82943b23..4990ed23 100644
--- a/gamutrf/grinference2mqtt.py
+++ b/gamutrf/grinference2mqtt.py
@@ -34,7 +34,7 @@ def __init__(
         external_gps_server_port,
         log_path,
     ):
-        self.yaml_buffer = ""
+        self.json_buffer = ""
         self.mqtt_reporter = None
         self.q = queue.Queue()
         self.mqtt_reporter_thread = threading.Thread(
@@ -94,14 +94,14 @@ def work(self, input_items, output_items):
         for input_item in input_items:
             raw_input_item = input_item.tobytes().decode("utf8")
             n += len(raw_input_item)
-            self.yaml_buffer += raw_input_item
+            self.json_buffer += raw_input_item
         while True:
-            delim_pos = self.yaml_buffer.find(DELIM)
+            delim_pos = self.json_buffer.find(DELIM)
             if delim_pos == -1:
                 break
-            raw_item = self.yaml_buffer[:delim_pos]
+            raw_item = self.json_buffer[:delim_pos]
             item = json.loads(raw_item)
-            self.yaml_buffer = self.yaml_buffer[delim_pos + len(DELIM) :]
+            self.json_buffer = self.json_buffer[delim_pos + len(DELIM) :]
             self.process_item(item)
         return n
 
diff --git a/gamutrf/grscan.py b/gamutrf/grscan.py
index ec6d6bd6..d9e7f22a 100644
--- a/gamutrf/grscan.py
+++ b/gamutrf/grscan.py
@@ -41,6 +41,8 @@ def __init__(
         freq_start=100e6,
         gps_server="",
         igain=0,
+        inference_addr="0.0.0.0",  # nosec
+        inference_port=8002,
         inference_min_confidence=0.5,
         inference_min_db=-200,
         inference_model_server="",
@@ -204,9 +206,12 @@ def __init__(
             not pretune and low_power_hold_down,
         )
         self.fft_blocks.append(retune_fft)
-        zmq_addr = f"tcp://{logaddr}:{logport}"
-        logging.info("serving FFT on %s", zmq_addr)
-        self.fft_blocks.append((zeromq.pub_sink(1, 1, zmq_addr, 100, False, 65536, "")))
+        fft_zmq_addr = f"tcp://{logaddr}:{logport}"
+        inference_zmq_addr = f"tcp://{inference_addr}:{inference_port}"
+        logging.info("serving FFT on %s", fft_zmq_addr)
+        self.fft_blocks.append(
+            (zeromq.pub_sink(1, 1, fft_zmq_addr, 100, False, 65536, ""))
+        )
 
         self.inference_blocks = []
         if inference_output_dir:
@@ -255,20 +260,23 @@ def __init__(
                         )
                     ]
                 )
-            else:
-                self.inference_blocks.extend([blocks.null_sink(1)])
-        if not self.inference_blocks:
-            self.inference_blocks = [blocks.null_sink(gr.sizeof_float * nfft)]
-
         if pretune:
             self.msg_connect((self.retune_pre_fft, "tune"), (self.sources[0], cmd_port))
             self.msg_connect((self.retune_pre_fft, "tune"), (retune_fft, "cmd"))
         else:
             self.msg_connect((retune_fft, "tune"), (self.sources[0], cmd_port))
         self.connect_blocks(self.sources[0], self.sources[1:])
-        self.connect((retune_fft, 1), (self.inference_blocks[0], 0))
 
-        self.connect_blocks(self.inference_blocks[0], self.inference_blocks[1:])
+        if self.inference_blocks:
+            self.connect((retune_fft, 1), (self.inference_blocks[0], 0))
+            self.connect_blocks(self.inference_blocks[0], self.inference_blocks[1:])
+            self.connect_blocks(
+                self.inference_blocks[0],
+                [zeromq.pub_sink(1, 1, inference_zmq_addr, 100, False, 65536, "")],
+            )
+        else:
+            self.connect((retune_fft, 1), (blocks.null_sink(gr.sizeof_float * nfft)))
+
         for pipeline_blocks in (
             self.fft_blocks,
             self.samples_blocks,
diff --git a/gamutrf/scan.py b/gamutrf/scan.py
index c64cf468..96d88450 100644
--- a/gamutrf/scan.py
+++ b/gamutrf/scan.py
@@ -134,6 +134,20 @@ def argument_parser():
         default=8001,
         help="Log FFT results to this port",
     )
+    parser.add_argument(
+        "--inference_addr",
+        dest="inference_addr",
+        type=str,
+        default="0.0.0.0",  # nosec
+        help="Log inference results to this address",
+    )
+    parser.add_argument(
+        "--inference_port",
+        dest="inference_port",
+        type=int,
+        default=8002,
+        help="Log inference results to this port",
+    )
     parser.add_argument(
         "--promport",
         dest="promport",
diff --git a/orchestrator.yml b/orchestrator.yml
index 053f5e4a..6a8e1048 100644
--- a/orchestrator.yml
+++ b/orchestrator.yml
@@ -34,6 +34,7 @@ services:
     ports:
       - '9001:9000'
       - '10000:10000'
+      - '10001:10001'
     cap_add:
       - SYS_NICE
       - SYS_RAWIO
@@ -73,6 +74,8 @@ services:
       - --no-compass
       - --use_external_gps
       - --use_external_heading
+      - --inference_addr=0.0.0.0
+      - --inference_port=10001
       - --inference_min_confidence=0.25
       - --inference_min_db=-80
       - --inference_model_name=mini2_snr