Mohd-Shujath commited on
Commit
2419d6a
·
1 Parent(s): 2c0a1bd
Files changed (1) hide show
  1. app.py +27 -23
app.py CHANGED
@@ -44,7 +44,7 @@ except (ImportError, RuntimeError, OSError):
44
  import spaces
45
  from typing import List, Literal, Optional
46
  from pydantic import BaseModel, Field
47
- from fastapi import HTTPException, Header
48
  import gradio as gr
49
 
50
  # =============================================================
@@ -69,7 +69,6 @@ SYSTEM_PROMPT = (
69
  "earnings, and low traffic delays."
70
  )
71
 
72
- # ZeroGPU decorator satisfies HF's startup check
73
  @spaces.GPU
74
  def run_inference(prompt: str, history: List[dict] = None) -> str:
75
  messages = [{"role": "system", "content": SYSTEM_PROMPT}]
@@ -83,26 +82,10 @@ def run_inference(prompt: str, history: List[dict] = None) -> str:
83
  return response["choices"][0]["message"]["content"]
84
 
85
  # =============================================================
86
- # 3. Gradio UI
87
  # =============================================================
88
- def gradio_chat(user_message, history):
89
- hist_formatted = []
90
- if history:
91
- for user_msg, bot_msg in history:
92
- hist_formatted.append({"role": "user", "content": user_msg})
93
- if bot_msg:
94
- hist_formatted.append({"role": "assistant", "content": bot_msg})
95
- return run_inference(user_message, hist_formatted)
96
 
97
- demo = gr.ChatInterface(
98
- fn=gradio_chat,
99
- title="🧭 CommitCon Sherpa Driver Copilot",
100
- description="Fine-tuned Qwen 0.8B Q4_K_M — llama-cpp-python inference"
101
- )
102
-
103
- # =============================================================
104
- # 4. Attach FastAPI Endpoints to demo.app
105
- # =============================================================
106
  class HistoryMessage(BaseModel):
107
  role: Literal["user", "assistant"]
108
  content: str = Field(min_length=1, max_length=4000)
@@ -113,7 +96,7 @@ class ChatRequest(BaseModel):
113
  current_zone_id: Optional[str] = None
114
  history: List[HistoryMessage] = Field(default_factory=list)
115
 
116
- @demo.app.post("/chat")
117
  def chat_endpoint(payload: ChatRequest, x_api_key: Optional[str] = Header(None)):
118
  if API_KEY and x_api_key != API_KEY:
119
  raise HTTPException(status_code=401, detail="Unauthorized: Invalid API Key")
@@ -127,7 +110,28 @@ def chat_endpoint(payload: ChatRequest, x_api_key: Optional[str] = Header(None))
127
  }
128
 
129
  # =============================================================
130
- # 5. Launch
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
131
  # =============================================================
132
  if __name__ == "__main__":
133
- demo.launch()
 
44
  import spaces
45
  from typing import List, Literal, Optional
46
  from pydantic import BaseModel, Field
47
+ from fastapi import FastAPI, HTTPException, Header
48
  import gradio as gr
49
 
50
  # =============================================================
 
69
  "earnings, and low traffic delays."
70
  )
71
 
 
72
  @spaces.GPU
73
  def run_inference(prompt: str, history: List[dict] = None) -> str:
74
  messages = [{"role": "system", "content": SYSTEM_PROMPT}]
 
82
  return response["choices"][0]["message"]["content"]
83
 
84
  # =============================================================
85
+ # 3. Create FastAPI app & define POST /chat endpoint
86
  # =============================================================
87
+ app = FastAPI(title="CommitCon Sherpa AI Copilot")
 
 
 
 
 
 
 
88
 
 
 
 
 
 
 
 
 
 
89
  class HistoryMessage(BaseModel):
90
  role: Literal["user", "assistant"]
91
  content: str = Field(min_length=1, max_length=4000)
 
96
  current_zone_id: Optional[str] = None
97
  history: List[HistoryMessage] = Field(default_factory=list)
98
 
99
+ @app.post("/chat")
100
  def chat_endpoint(payload: ChatRequest, x_api_key: Optional[str] = Header(None)):
101
  if API_KEY and x_api_key != API_KEY:
102
  raise HTTPException(status_code=401, detail="Unauthorized: Invalid API Key")
 
110
  }
111
 
112
  # =============================================================
113
+ # 4. Gradio Chat UI mounted at /ui (avoiding root conflict)
114
+ # =============================================================
115
+ def gradio_chat(user_message, history):
116
+ hist_formatted = []
117
+ if history:
118
+ for user_msg, bot_msg in history:
119
+ hist_formatted.append({"role": "user", "content": user_msg})
120
+ if bot_msg:
121
+ hist_formatted.append({"role": "assistant", "content": bot_msg})
122
+ return run_inference(user_message, hist_formatted)
123
+
124
+ demo = gr.ChatInterface(
125
+ fn=gradio_chat,
126
+ title="🧭 CommitCon Sherpa Driver Copilot",
127
+ description="Fine-tuned Qwen 0.8B Q4_K_M — llama-cpp-python inference"
128
+ )
129
+
130
+ # Mount Gradio UI cleanly without hijacking API endpoints
131
+ app = gr.mount_gradio_app(app, demo, path="/")
132
+
133
+ # =============================================================
134
+ # 5. Native Launch using demo.launch(app=...)
135
  # =============================================================
136
  if __name__ == "__main__":
137
+ demo.launch(app=app)