expose seed to api

tools/server/README.md

@@ -802,6 +802,8 @@ Returns raw audio bytes (`audio/wav` by default) rather than JSON. For more info
 
 `repeat_penalty`: Repetition penalty applied to the backbone sampler over the whole generation (default `1.05`). Without this, the backbone can loop and re-generate the same utterance.
 
+`seed`: RNG seed for both the backbone sampler and the codec/vocoder (`-1` = random, the default). Set to a fixed value for reproducible output.
+
 `n_predict`: Max number of audio frames to generate. Defaults to `512`; generation normally stops earlier once the model emits an end-of-speech token.
 
 `response_format`: `wav` (default) or `pcm` (raw `float32` samples, no header).

tools/server/server-context.cpp

@@ -5241,6 +5241,8 @@ void server_routes::init_routes() {
         task.params.sampling  = params.sampling; // baseline defaults, then apply overrides below
         task.params.sampling.penalty_repeat  = json_value(body, "repeat_penalty", 1.05f);
         task.params.sampling.penalty_last_n  = -1;
+        task.params.sampling.seed = json_value(body, "seed", params.sampling.seed);
+        task.tts_inp.data.seed    = task.params.sampling.seed; // same seed for the codec/vocoder RNG
         if (task.tts_inp.data.top_k > 0) {
             task.params.sampling.top_k = task.tts_inp.data.top_k;
         }