fix: call Modal directly from the canonical Space

README.md

@@ -113,11 +113,10 @@ hackathon's 32B total limit.
 - **Training:** the MiniCPM and FLUX adapters, validation runs, and deployment
   experiments used Modal.
 
-Text and image inference scale independently. The canonical hackathon Space
-serves the custom interface and streams NDJSON progress. Because organization
-members cannot manage that Space's secrets, it forwards generation requests to
-the owner-controlled CPU orchestrator, which HMAC-signs calls to the two Modal
-services. No credential is stored in the public repository.
+Text and image inference scale independently. The canonical hackathon Space is
+a CPU orchestrator: it serves the custom interface, validates requests, streams
+NDJSON progress, and HMAC-signs calls directly to the two Modal services. No
+credential is stored in the public repository.
 
 ## Published Artifacts
 

docs/build-small-hackathon-article.md

@@ -133,10 +133,9 @@ animals, so the visual variety felt smaller than the style menu suggested.
 
 ## Modal as the GPU Layer
 
-The canonical organization Space serves the custom interface and streams the
-API response. Since organization members cannot manage its secrets, it forwards
-generation requests to an owner-controlled CPU Space that holds the HMAC
-credential. Text and image workloads then run on separate Modal applications:
+The canonical organization Space serves the custom interface, validates
+requests, and streams the API response. It holds an HMAC credential as a
+private Space secret and calls two separate Modal applications directly:
 
 - MiniCPM4.1-8B on an A100 40GB endpoint
 - FLUX.1-schnell plus the four style adapters on an A100 80GB endpoint
@@ -147,9 +146,8 @@ model from evicting the other and lets each service scale to zero
 independently. The public repository contains no credentials.
 
 Modal was also used for adapter training, validation grids, GGUF smoke tests,
-and deployment experiments. The runtime bridge signs requests with HMAC, and
-the organization Space preserves the NDJSON stream so a long generation
-remains visibly alive through both Hugging Face hops.
+and deployment experiments. The organization Space signs requests with HMAC
+and preserves the NDJSON stream so a long generation remains visibly alive.
 
 ## Codex as an Engineering Partner
 

src/compliment_forest/config.py

@@ -64,11 +64,6 @@ class AppConfig(BaseModel):
     def from_env(cls) -> AppConfig:
         trace_path = os.getenv("CF_TRACE_PATH")
         hosted_space = bool(os.getenv("SPACE_ID"))
-        submission_upstream = (
-            "https://thangvip-compliment-forest.hf.space"
-            if os.getenv("SPACE_ID") == "build-small-hackathon/compliment-forest"
-            else None
-        )
         default_text_backend = "transformers" if hosted_space else "demo"
         default_image_backend = "zerogpu" if hosted_space else "demo"
         return cls(
@@ -104,7 +99,7 @@ class AppConfig(BaseModel):
             modal_image_endpoint=os.getenv("CF_MODAL_IMAGE_ENDPOINT"),
             modal_music_endpoint=os.getenv("CF_MODAL_MUSIC_ENDPOINT"),
             modal_signing_key=(os.getenv("CF_MODAL_SIGNING_KEY") or os.getenv("HF_TOKEN")),
-            upstream_space_url=os.getenv("CF_UPSTREAM_SPACE_URL") or submission_upstream,
+            upstream_space_url=os.getenv("CF_UPSTREAM_SPACE_URL"),
             local_files_only=os.getenv("CF_LOCAL_FILES_ONLY", "0") == "1",
             default_seed=int(os.getenv("CF_DEFAULT_SEED", "3407")),
             default_style=os.getenv("CF_DEFAULT_STYLE", "surprise"),