Is this a feature or a BUG fix? BUG fix.

Related issue: Fixes #946

Summary

POST /v1/tts returns an HTTP 500 with a JSON error body when text is empty.
A client that streams the response then writes that JSON error into a .wav file.

Root cause is a schema asymmetry in fish_speech/utils/schema.py:

• ServeTTSRequest.text was an unconstrained str, so an empty string passes
  validation and flows into synthesis.
• Empty text yields zero audio segments, which TTSInferenceEngine.inference
  reports as an error and tools/server/inference.py re-raises as
  HTTPException(INTERNAL_SERVER_ERROR) → the 500 reported in baize.exceptions.HTTPException: (500, "'No audio generated, please check the input text.'") #946
  ("No audio generated, please check the input text.").
• The sibling model AddReferenceRequest.text already constrains
  min_length=1, so the asymmetry is the bug.

The fix

Constrain ServeTTSRequest.text the same way, so empty input is rejected with a
clean 4xx validation error at the request boundary — before any model work:

# fish_speech/utils/schema.py
class ServeTTSRequest(BaseModel):
-    text: str
+    text: str = Field(..., min_length=1)

One line, no behavior change for valid input. Matches the existing min_length=1
on AddReferenceRequest.text and the Field(...) constraint style already used on
ServeTTSRequest (top_p, temperature, repetition_penalty).

Validation

The change is pure pydantic validation (no model weights needed). Importing the
model and constructing it:

Scope

This fixes the empty-string trigger at the schema boundary. The broader
"no segments generated → 500" path (e.g. other degenerate inputs) is a separate,
larger change in tools/server/inference.py and is intentionally out of scope here.

Disclosure: this fix was drafted with AI assistance (Claude) and reviewed, tested,
and submitted by me. The commit carries a Co-Authored-By trailer.