model_specs: declare the ASR, forced-aligner and codec options the sessions read

model_specs/miocodec.json

@@ -26,6 +26,60 @@
       "speaker_reference"
     ]
   },
+  "options": {
+    "request": [],
+    "session": [
+      {
+        "name": "weight_type",
+        "type": "enum",
+        "description": "Codec weight storage type; default f32.",
+        "preset": "weight_type_full",
+        "required": false,
+        "default": "f32"
+      },
+      {
+        "name": "weight_context_mb",
+        "type": "int",
+        "description": "Codec weight descriptor context size in MiB; default 256.",
+        "required": false,
+        "min": 1,
+        "default": 256
+      },
+      {
+        "name": "constant_context_mb",
+        "type": "int",
+        "description": "Reusable constant context size in MiB; default 256.",
+        "required": false,
+        "min": 1,
+        "default": 256
+      },
+      {
+        "name": "content_graph_arena_mb",
+        "type": "int",
+        "description": "Content encoder graph arena size in MiB; default 512.",
+        "required": false,
+        "min": 1,
+        "default": 512
+      },
+      {
+        "name": "global_graph_arena_mb",
+        "type": "int",
+        "description": "Global encoder graph arena size in MiB; default 256.",
+        "required": false,
+        "min": 1,
+        "default": 256
+      },
+      {
+        "name": "wave_graph_arena_mb",
+        "type": "int",
+        "description": "Wave decoder graph arena size in MiB; default 512.",
+        "required": false,
+        "min": 1,
+        "default": 512
+      }
+    ],
+    "load": []
+  },
   "runtime": {
     "tags": [
       "gguf"

model_specs/mms_forced_aligner.json

@@ -61,7 +61,7 @@
       {
         "name": "return_timestamps",
         "type": "bool",
-        "description": "Request word timestamps in the result; set automatically by --words-out.",
+        "description": "Accepted for cross-model compatibility (--words-out sets it) and ignored: the forced aligner always returns word timestamps.",
         "required": false,
         "default": true
       }

model_specs/parakeet_tdt.json

@@ -54,6 +54,12 @@
   },
   "options": {
     "request": [
+      {
+        "name": "language",
+        "type": "string",
+        "description": "Transcription language from the OpenAI-compatible request field. Accepted for API compatibility; Parakeet TDT detects the language itself and does not use this value (see keep_language_tags).",
+        "required": false
+      },
       {
         "name": "max_tokens",
         "type": "int",

model_specs/qwen3_forced_aligner.json

@@ -37,6 +37,63 @@
         "required": false,
         "default": false
       }
+    ],
+    "session": [
+      {
+        "name": "weight_type",
+        "type": "enum",
+        "description": "Fallback weight storage type for the thinker; default native.",
+        "preset": "weight_type_full",
+        "required": false,
+        "default": "native"
+      },
+      {
+        "name": "thinker_weight_type",
+        "type": "enum",
+        "description": "Thinker matmul weight storage type; defaults to weight_type when set, otherwise native.",
+        "preset": "weight_type_full",
+        "required": false
+      },
+      {
+        "name": "audio_encoder_weight_type",
+        "type": "enum",
+        "description": "Audio encoder weight storage type; default native.",
+        "preset": "weight_type_conv",
+        "required": false,
+        "default": "native"
+      },
+      {
+        "name": "audio_encoder_graph_arena_mb",
+        "type": "int",
+        "description": "Audio encoder graph arena size in MiB; default 128.",
+        "required": false,
+        "min": 1,
+        "default": 128
+      },
+      {
+        "name": "thinker_prefill_graph_arena_mb",
+        "type": "int",
+        "description": "Thinker prefill graph arena size in MiB; default 256.",
+        "required": false,
+        "min": 1,
+        "default": 256
+      },
+      {
+        "name": "thinker_decode_graph_arena_mb",
+        "type": "int",
+        "description": "Thinker decode graph arena size in MiB; default 256.",
+        "required": false,
+        "min": 1,
+        "default": 256
+      },
+      {
+        "name": "thinker_weight_context_mb",
+        "type": "int",
+        "description": "Thinker weight descriptor context size in MiB; default 64.",
+        "required": false,
+        "min": 1,
+        "default": 64
+      }
     ]
   },
   "runtime": {

model_specs/sense_asr.json

@@ -69,10 +69,11 @@
       {
         "name": "audio_chunk_mode",
         "type": "enum",
-        "description": "Audio chunking mode: auto, fixed, or none.",
+        "description": "Audio chunking mode: auto, fixed, vad, or none.",
         "values": [
           "auto",
           "fixed",
+          "vad",
           "none"
         ],
         "required": false,