Add VoiceGate workflow assets and TODO

docs/deployment-plan.md

@@ -28,8 +28,9 @@ Current local state:
 - `VoiceGate/.gitmodules` references `comfyui_voicebridge`, but the local
   `VoiceGate/comfyui_voicebridge/` directory is currently empty.
 - `VoiceGate/workflows/VoiceGate-Workflow.json` is the UI workflow.
-- `VoiceGate/workflows/VoiceGate-Workflow_api.json` exists, but currently fails
-  JSON parsing and must be re-exported or repaired before API execution.
+- `VoiceGate/workflows/VoiceGate-Workflow_api.json` exists and has been
+  confirmed as valid JSON. It still needs parameterization before Gradio can
+  submit it to ComfyUI.
 
 ## Hugging Face Space Constraints
 
@@ -179,5 +180,6 @@ Verify in this order:
 
 ## Immediate Next Step
 
-Prepare the runtime bootstrap scripts and decide whether to vendor custom nodes
-into `custom_nodes/` or download pinned commits during Space startup.
+Copy the validated API workflow into the Space `workflows/` directory, prepare
+the runtime bootstrap scripts, and decide whether to vendor custom nodes into
+`custom_nodes/` or download pinned commits during Space startup.

docs/todo.md

@@ -0,0 +1,114 @@
+# VoiceGate HF Space TODO
+
+This is the execution checklist for bringing the VoiceGate Hugging Face Space
+from scaffold to a working Gradio app.
+
+## Current Status
+
+- [x] Create the `VoiceGate-hf` Space wrapper repository.
+- [x] Configure and push to `build-small-hackathon/VoiceGate`.
+- [x] Replace the default HF template with a VoiceGate scaffold.
+- [x] Keep the upstream `VoiceGate/` checkout local-only and ignored.
+- [x] Preserve Hugging Face LFS rules in `.gitattributes`.
+- [x] Confirm `VoiceGate/workflows/VoiceGate-Workflow_api.json` is valid JSON.
+- [x] Copy the validated API workflow into `workflows/voicegate_api.json`.
+
+## Phase 1: Repository Hygiene
+
+- [x] Copy `VoiceGate/workflows/VoiceGate-Workflow_api.json` to
+  `workflows/voicegate_api.json`.
+- [x] Copy `VoiceGate/workflows/VoiceGate-Workflow.json` to
+  `workflows/voicegate_ui.json` for reference.
+- [ ] Update docs so `VoiceGate/` is clearly documented as a local upstream
+  checkout, not Space runtime content.
+- [ ] Commit and push the workflow files after confirming they contain no
+  secrets or large media.
+
+## Phase 2: Dependency Inventory
+
+- [ ] Identify exact repositories and pinned commits for every required custom
+  node package:
+  - [ ] `comfyui_voicebridge`
+  - [ ] RunningHub VoxCPM nodes
+  - [ ] MelBandRoFormer nodes
+  - [ ] rgthree nodes
+  - [ ] comfyui-easy-use
+  - [ ] Comfyroll text nodes or equivalent provider for `CR Text`
+  - [ ] `RH_LLMAPI_NODE` provider
+- [ ] Identify Python dependency constraints for ComfyUI and all custom nodes.
+- [ ] Identify system packages beyond `ffmpeg` and `git`, if any.
+- [ ] Decide whether custom nodes are vendored in `custom_nodes/` or installed
+  by pinned git URL during bootstrap.
+- [ ] Decide where model files are downloaded and cached in the Space.
+
+## Phase 3: Runtime Bootstrap
+
+- [ ] Add `scripts/bootstrap_comfy.py`.
+- [ ] Add `scripts/run_comfy.py`.
+- [ ] Add `scripts/workflow_client.py`.
+- [ ] Install or prepare ComfyUI on Space startup.
+- [ ] Install custom node dependencies.
+- [ ] Verify ComfyUI can start locally in the Space.
+- [ ] Verify ComfyUI API endpoints are reachable:
+  - [ ] `/system_stats`
+  - [ ] `/upload/image` or the audio upload equivalent used by `LoadAudio`
+  - [ ] `/prompt`
+  - [ ] `/history/{prompt_id}`
+
+## Phase 4: Workflow Parameterization
+
+- [ ] Replace node `16` `LoadAudio.inputs.audio` with the uploaded audio file.
+- [ ] Replace node `105` `api_key` with `DEEPSEEK_API_KEY` from Space secrets.
+- [ ] Parameterize node `105` `api_baseurl`.
+- [ ] Parameterize node `105` `model`.
+- [ ] Parameterize node `110` target language.
+- [ ] Add unique job-specific output prefixes for node `180` and node `214`.
+- [ ] Decide which user controls are exposed first:
+  - [ ] source language
+  - [ ] target language
+  - [ ] LLM model
+  - [ ] max input duration
+  - [ ] keep or drop background audio
+- [ ] Remove or ignore display-only `easy showAnything` nodes if they are not
+  needed for API execution.
+
+## Phase 5: Gradio Interface
+
+- [ ] Replace the placeholder `app.py` with the first VoiceGate interface.
+- [ ] Add short audio upload input.
+- [ ] Add target language input.
+- [ ] Add status/log output.
+- [ ] Add generated audio output.
+- [ ] Add generated SRT file outputs.
+- [ ] Wrap GPU-heavy execution with `@spaces.GPU(duration=...)`.
+- [ ] Add guardrails for file type and duration.
+
+## Phase 6: Minimal Runtime Tests
+
+- [ ] Start Space and confirm Gradio loads.
+- [ ] Start ComfyUI from the Space process.
+- [ ] Submit a minimal prompt to ComfyUI and receive a response.
+- [ ] Run a tiny TTS-only workflow.
+- [ ] Run a short ASR-only workflow.
+- [ ] Run SRT split -> VoxCPM -> SRT merge.
+- [ ] Run the full short-audio VoiceGate workflow.
+- [ ] Confirm output audio and SRT files are downloadable from Gradio.
+
+## Phase 7: Full VoiceGate Path
+
+- [ ] Add video input support after the audio path is stable.
+- [ ] Extract audio from video with `ffmpeg`.
+- [ ] Merge generated audio back into the original video if required.
+- [ ] Add subtitle download and optional subtitle burn-in.
+- [ ] Add examples with small sample files, if allowed by Space storage limits.
+
+## Open Questions
+
+- [ ] Which exact node repository provides `RH_LLMAPI_NODE`?
+- [ ] Which exact node repository provides `RunningHub_VoxCPM_*`?
+- [ ] Is `flash_attention_2` available and reliable in the ZeroGPU environment?
+- [ ] Does VoxCPM2 fit comfortably in ZeroGPU memory with ASR and
+  MelBandRoFormer in the same run?
+- [ ] Should the first public demo disable background separation to reduce
+  runtime and memory pressure?
+- [ ] What maximum uploaded audio/video duration should the first version allow?

workflows/voicegate_api.json

@@ -0,0 +1,375 @@
+{
+  "16": {
+    "inputs": {
+      "audio": "fcecb7fd3827851a4fb22c52886610aed0a62e5fe32ea46f3b1291ff3dd236ab.mp3",
+      "audioUI": ""
+    },
+    "class_type": "LoadAudio",
+    "_meta": {
+      "title": "输入原音频"
+    }
+  },
+  "31": {
+    "inputs": {
+      "repo_id": "Qwen/Qwen3-ASR-1.7B",
+      "precision": "bf16",
+      "attention": "flash_attention_2",
+      "max_new_tokens": 1024,
+      "forced_aligner": "Qwen/Qwen3-ForcedAligner-0.6B"
+    },
+    "class_type": "VoiceBridgeASRLoader",
+    "_meta": {
+      "title": "VoiceBridge ASR Loader"
+    }
+  },
+  "33": {
+    "inputs": {
+      "language": "auto",
+      "context": "",
+      "return_timestamps": true,
+      "model_key": [
+        "31",
+        0
+      ],
+      "audio": [
+        "99",
+        0
+      ]
+    },
+    "class_type": "VoiceBridgeASRTranscribe",
+    "_meta": {
+      "title": "VoiceBridge ASR Transcribe"
+    }
+  },
+  "36": {
+    "inputs": {
+      "text": [
+        "33",
+        1
+      ],
+      "language": [
+        "33",
+        2
+      ],
+      "save_srt": true,
+      "filename_prefix": "VoiceBridge/subtitle",
+      "forced_aligns": [
+        "33",
+        0
+      ]
+    },
+    "class_type": "GenerateSRT",
+    "_meta": {
+      "title": "Generate SRT"
+    }
+  },
+  "45": {
+    "inputs": {
+      "language": "auto",
+      "context": "",
+      "return_timestamps": false,
+      "model_key": [
+        "31",
+        0
+      ],
+      "audio": [
+        "46",
+        0
+      ]
+    },
+    "class_type": "VoiceBridgeASRTranscribe",
+    "_meta": {
+      "title": "VoiceBridge ASR Transcribe"
+    }
+  },
+  "46": {
+    "inputs": {
+      "start_index": 0,
+      "duration": 10.000000000000002,
+      "audio": [
+        "99",
+        0
+      ]
+    },
+    "class_type": "TrimAudioDuration",
+    "_meta": {
+      "title": "Trim Audio Duration"
+    }
+  },
+  "61": {
+    "inputs": {
+      "text": "1\n00:00:00,000 --> 00:00:04,000\n我相信你在各种内容下面应该都见过这样一句类似的评论\n\n2\n00:00:04,240 --> 00:00:04,960\n那几个月之前\n\n3\n00:00:04,960 --> 00:00:06,879\n其实大家把这个当成一种调侃了\n\n4\n00:00:06,879 --> 00:00:07,679\n是一种梗\n\n5\n00:00:07,679 --> 00:00:07,919\n对吧\n\n6\n00:00:08,160 --> 00:00:09,359\n你看到太离谱的东西\n\n7\n00:00:09,439 --> 00:00:10,320\n你会说\n\n8\n00:00:10,320 --> 00:00:10,800\n“哎呀\n\n9\n00:00:10,800 --> 00:00:12,560\n这是AI或者AI太好用了\n\n10\n00:00:12,560 --> 00:00:12,960\n家人\n\n11\n00:00:13,359 --> 00:00:14,240\n”但有意思的就是\n\n12\n00:00:14,240 --> 00:00:15,359\n只过了短短几个月\n\n13\n00:00:15,519 --> 00:00:19,519\n你会发现现在这句“这是AI嘛”背后的情绪它变化了\n\n14\n00:00:19,839 --> 00:00:21,280\n因为AI它又变强了\n\n",
+      "anything": [
+        "36",
+        0
+      ]
+    },
+    "class_type": "easy showAnything",
+    "_meta": {
+      "title": "Show Any"
+    }
+  },
+  "96": {
+    "inputs": {
+      "model_name": "MelBandRoFormer_comfy/MelBandRoformer_fp32.safetensors"
+    },
+    "class_type": "MelBandRoFormerModelLoader",
+    "_meta": {
+      "title": "Mel-Band RoFormer Model Loader"
+    }
+  },
+  "99": {
+    "inputs": {
+      "model": [
+        "96",
+        0
+      ],
+      "audio": [
+        "16",
+        0
+      ]
+    },
+    "class_type": "MelBandRoFormerSampler",
+    "_meta": {
+      "title": "Mel-Band RoFormer Sampler"
+    }
+  },
+  "105": {
+    "inputs": {
+      "api_baseurl": "https://api.deepseek.com",
+      "api_key": "",
+      "model": "deepseek-v4-flash",
+      "role": [
+        "109",
+        0
+      ],
+      "prompt": [
+        "36",
+        0
+      ],
+      "temperature": 0.5,
+      "seed": 1761935326
+    },
+    "class_type": "RH_LLMAPI_NODE",
+    "_meta": {
+      "title": "Runninghub LLM API Node"
+    }
+  },
+  "107": {
+    "inputs": {
+      "text": "1\n00:00:00,000 --> 00:00:04,000\nI bet you've seen a comment like this under all kinds of content.\n\n2\n00:00:04,240 --> 00:00:04,960\nA few months ago,\n\n3\n00:00:04,960 --> 00:00:06,879\neveryone used it as a joke.\n\n4\n00:00:06,879 --> 00:00:07,679\nIt was a meme.\n\n5\n00:00:07,679 --> 00:00:07,919\nRight?\n\n6\n00:00:08,160 --> 00:00:09,359\nYou see something absurd.\n\n7\n00:00:09,439 --> 00:00:10,320\nYou'd say,\n\n8\n00:00:10,320 --> 00:00:10,800\n\"Ah,\n\n9\n00:00:10,800 --> 00:00:12,560\nmust be AI, or AI is too good.\n\n10\n00:00:12,560 --> 00:00:12,960\nBro.\"\n\n11\n00:00:13,359 --> 00:00:14,240\nBut the interesting thing is,\n\n12\n00:00:14,240 --> 00:00:15,359\njust a few months later,\n\n13\n00:00:15,519 --> 00:00:19,519\nnow the emotion behind \"Is this AI?\" has changed.\n\n14\n00:00:19,839 --> 00:00:21,280\nBecause AI has gotten even stronger.",
+      "anything": [
+        "105",
+        0
+      ]
+    },
+    "class_type": "easy showAnything",
+    "_meta": {
+      "title": "Show Any"
+    }
+  },
+  "108": {
+    "inputs": {
+      "text": "You are a professional subtitle translator. Please translate all the subtitles in this SRT file into natural, spoken [Language]. Keep each timestamp exactly the same. For each subtitle line, use the shortest possible phrasing that still captures the original meaning (it doesn't have to be a direct translation, just convey the correct idea). Output the content directly in SRT format, without any explanation or descriptive text."
+    },
+    "class_type": "CR Text",
+    "_meta": {
+      "title": "🔤 CR Text"
+    }
+  },
+  "109": {
+    "inputs": {
+      "texts": [
+        "108",
+        0
+      ],
+      "find": "[Language]",
+      "replace": [
+        "110",
+        0
+      ]
+    },
+    "class_type": "ReplaceText",
+    "_meta": {
+      "title": "Replace Text (DEPRECATED)"
+    }
+  },
+  "110": {
+    "inputs": {
+      "value": "English"
+    },
+    "class_type": "easy string",
+    "_meta": {
+      "title": "输入目标语言"
+    }
+  },
+  "176": {
+    "inputs": {
+      "tempo_limit": 2,
+      "mini_gap_ms": 10,
+      "srt_items": [
+        "177",
+        1
+      ],
+      "audios": [
+        "268",
+        0
+      ]
+    },
+    "class_type": "VoiceBridgeAudioListMergerBySRT",
+    "_meta": {
+      "title": "VoiceBridge Audio List Merger by SRT"
+    }
+  },
+  "177": {
+    "inputs": {
+      "srt_string": [
+        "105",
+        0
+      ]
+    },
+    "class_type": "VoiceBridgeSRTSplitter",
+    "_meta": {
+      "title": "VoiceBridge SRT Splitter"
+    }
+  },
+  "179": {
+    "inputs": {
+      "text": "1\n00:00:00,000 --> 00:00:03,540\nI bet you've seen a comment like this under all kinds of content.\n\n2\n00:00:03,550 --> 00:00:04,681\nA few months ago,\n\n3\n00:00:04,960 --> 00:00:06,851\neveryone used it as a joke.\n\n4\n00:00:06,861 --> 00:00:07,835\nIt was a meme.\n\n5\n00:00:07,845 --> 00:00:08,400\nRight?\n\n6\n00:00:08,410 --> 00:00:09,683\nYou see something absurd.\n\n7\n00:00:09,693 --> 00:00:10,536\nYou'd say,\n\n8\n00:00:10,546 --> 00:00:10,996\n\"Ah,\n\n9\n00:00:11,006 --> 00:00:13,018\nmust be AI, or AI is too good.\n\n10\n00:00:13,028 --> 00:00:13,837\nBro.\"\n\n11\n00:00:13,847 --> 00:00:14,700\nBut the interesting thing is,\n\n12\n00:00:14,710 --> 00:00:15,946\njust a few months later,\n\n13\n00:00:15,956 --> 00:00:20,245\nnow the emotion behind \"Is this AI?\" has changed.\n\n14\n00:00:20,255 --> 00:00:23,795\nBecause AI has gotten even stronger.\n",
+      "anything": [
+        "176",
+        1
+      ]
+    },
+    "class_type": "easy showAnything",
+    "_meta": {
+      "title": "Show Any"
+    }
+  },
+  "180": {
+    "inputs": {
+      "filename_prefix": "audio/ComfyUI",
+      "quality": "V0",
+      "audioUI": "",
+      "audio": [
+        "251",
+        0
+      ]
+    },
+    "class_type": "SaveAudioMP3",
+    "_meta": {
+      "title": "Save Audio (MP3)"
+    }
+  },
+  "206": {
+    "inputs": {
+      "control_instruction": "",
+      "text": [
+        "177",
+        0
+      ],
+      "cfg_value": 2,
+      "inference_steps": 10,
+      "seed": 220751697769047,
+      "ultimate_clone": true,
+      "reference_audio_text": [
+        "264",
+        0
+      ],
+      "normalize_text": false,
+      "denoise_reference": false,
+      "max_len": 4096,
+      "retry_badcase": true,
+      "model": [
+        "207",
+        0
+      ],
+      "reference_audio": [
+        "261",
+        0
+      ]
+    },
+    "class_type": "RunningHub_VoxCPM_Generate",
+    "_meta": {
+      "title": "RunningHub VoxCPM Generate Speech"
+    }
+  },
+  "207": {
+    "inputs": {
+      "model_name": "VoxCPM2",
+      "optimize": false,
+      "lora_name": "None"
+    },
+    "class_type": "RunningHub_VoxCPM_LoadModel",
+    "_meta": {
+      "title": "RunningHub VoxCPM Load Model"
+    }
+  },
+  "214": {
+    "inputs": {
+      "srt_string": [
+        "179",
+        0
+      ],
+      "filename_prefix": "VoiceBridge/subtitle"
+    },
+    "class_type": "SaveSRTFromString",
+    "_meta": {
+      "title": "Save SRT From String"
+    }
+  },
+  "251": {
+    "inputs": {
+      "audio_1": [
+        "176",
+        0
+      ],
+      "audio_2": [
+        "99",
+        1
+      ]
+    },
+    "class_type": "MergeAudioMW",
+    "_meta": {
+      "title": "Merge Audio"
+    }
+  },
+  "261": {
+    "inputs": {
+      "any_01": [
+        "46",
+        0
+      ]
+    },
+    "class_type": "Any Switch (rgthree)",
+    "_meta": {
+      "title": "Any Switch (rgthree)"
+    }
+  },
+  "264": {
+    "inputs": {
+      "any_01": [
+        "45",
+        1
+      ]
+    },
+    "class_type": "Any Switch (rgthree)",
+    "_meta": {
+      "title": "Any Switch (rgthree)"
+    }
+  },
+  "268": {
+    "inputs": {
+      "start_index": 0.30000000000000004,
+      "duration": 100.00000000000001,
+      "audio": [
+        "206",
+        0
+      ]
+    },
+    "class_type": "TrimAudioDuration",
+    "_meta": {
+      "title": "Trim Audio Duration"
+    }
+  }
+}