final
Browse files- server/app.py +55 -23
server/app.py
CHANGED
|
@@ -1,19 +1,35 @@
|
|
|
|
|
|
|
|
| 1 |
from fastapi import FastAPI
|
| 2 |
from pydantic import BaseModel
|
| 3 |
-
|
|
|
|
| 4 |
from environment import SupportEnv
|
| 5 |
-
from dataset import easy_cases, medium_cases, hard_cases
|
| 6 |
|
| 7 |
app = FastAPI()
|
| 8 |
|
| 9 |
-
|
|
|
|
| 10 |
class Action(BaseModel):
|
| 11 |
message: str
|
| 12 |
|
|
|
|
| 13 |
class ResetRequest(BaseModel):
|
| 14 |
task: str = "easy"
|
| 15 |
|
| 16 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
def parse_action(message):
|
| 18 |
message = message.lower()
|
| 19 |
|
|
@@ -28,12 +44,9 @@ def parse_action(message):
|
|
| 28 |
else:
|
| 29 |
return "reject"
|
| 30 |
|
| 31 |
-
|
| 32 |
-
TASKS
|
| 33 |
-
|
| 34 |
-
"medium": medium_cases,
|
| 35 |
-
"hard": hard_cases
|
| 36 |
-
}
|
| 37 |
|
| 38 |
# AUTO TASK CYCLING
|
| 39 |
TASK_SEQUENCE = ["easy", "medium", "hard"]
|
|
@@ -41,7 +54,8 @@ current_task_index = 0
|
|
| 41 |
|
| 42 |
env = None
|
| 43 |
|
| 44 |
-
|
|
|
|
| 45 |
@app.post("/reset")
|
| 46 |
def reset(req: Optional[ResetRequest] = None):
|
| 47 |
global env, current_task_index
|
|
@@ -55,34 +69,52 @@ def reset(req: Optional[ResetRequest] = None):
|
|
| 55 |
|
| 56 |
obs = env.reset()
|
| 57 |
|
| 58 |
-
return {
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
|
| 64 |
# STEP
|
| 65 |
-
@app.post("/step")
|
| 66 |
def step(action: Action):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 67 |
parsed_action = parse_action(action.message)
|
| 68 |
|
| 69 |
obs, reward, done = env.step(parsed_action)
|
| 70 |
|
| 71 |
-
response = {
|
| 72 |
-
"observation": {"echoed_message": obs},
|
| 73 |
-
"reward": reward,
|
| 74 |
-
"done": done
|
| 75 |
-
}
|
| 76 |
|
| 77 |
-
#
|
| 78 |
if done:
|
| 79 |
response["score"] = env.get_score()
|
| 80 |
|
| 81 |
return response
|
| 82 |
|
|
|
|
| 83 |
def main():
|
| 84 |
import uvicorn
|
|
|
|
| 85 |
uvicorn.run(app, host="0.0.0.0", port=7860)
|
| 86 |
|
|
|
|
|
|
|
|
|
|
| 87 |
if __name__ == "__main__":
|
| 88 |
main()
|
|
|
|
| 1 |
+
from typing import Optional
|
| 2 |
+
|
| 3 |
from fastapi import FastAPI
|
| 4 |
from pydantic import BaseModel
|
| 5 |
+
|
| 6 |
+
from dataset import easy_cases, hard_cases, medium_cases
|
| 7 |
from environment import SupportEnv
|
|
|
|
| 8 |
|
| 9 |
app = FastAPI()
|
| 10 |
|
| 11 |
+
|
| 12 |
+
# Models
|
| 13 |
class Action(BaseModel):
|
| 14 |
message: str
|
| 15 |
|
| 16 |
+
|
| 17 |
class ResetRequest(BaseModel):
|
| 18 |
task: str = "easy"
|
| 19 |
|
| 20 |
+
|
| 21 |
+
class Observation(BaseModel):
|
| 22 |
+
echoed_message: str
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
class StepResponse(BaseModel):
|
| 26 |
+
observation: Observation
|
| 27 |
+
reward: float
|
| 28 |
+
done: bool
|
| 29 |
+
score: Optional[float] = None
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
# Parser
|
| 33 |
def parse_action(message):
|
| 34 |
message = message.lower()
|
| 35 |
|
|
|
|
| 44 |
else:
|
| 45 |
return "reject"
|
| 46 |
|
| 47 |
+
|
| 48 |
+
# TASKS
|
| 49 |
+
TASKS = {"easy": easy_cases, "medium": medium_cases, "hard": hard_cases}
|
|
|
|
|
|
|
|
|
|
| 50 |
|
| 51 |
# AUTO TASK CYCLING
|
| 52 |
TASK_SEQUENCE = ["easy", "medium", "hard"]
|
|
|
|
| 54 |
|
| 55 |
env = None
|
| 56 |
|
| 57 |
+
|
| 58 |
+
# RESET
|
| 59 |
@app.post("/reset")
|
| 60 |
def reset(req: Optional[ResetRequest] = None):
|
| 61 |
global env, current_task_index
|
|
|
|
| 69 |
|
| 70 |
obs = env.reset()
|
| 71 |
|
| 72 |
+
return {"observation": {"echoed_message": obs}, "reward": 0.0, "done": False}
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
# STATE
|
| 76 |
+
@app.get("/state")
|
| 77 |
+
def get_state():
|
| 78 |
+
global env
|
| 79 |
+
if env is None:
|
| 80 |
+
return {"error": "Environment not initialized. Call /reset first."}
|
| 81 |
+
|
| 82 |
+
return {"observation": {"echoed_message": env._get_obs()}}
|
| 83 |
+
|
| 84 |
|
| 85 |
# STEP
|
| 86 |
+
@app.post("/step", response_model=StepResponse)
|
| 87 |
def step(action: Action):
|
| 88 |
+
global env
|
| 89 |
+
|
| 90 |
+
# Safety check if step is called before reset
|
| 91 |
+
if env is None:
|
| 92 |
+
return {
|
| 93 |
+
"observation": {"echoed_message": "Error: Environment not initialized."},
|
| 94 |
+
"reward": 0.0,
|
| 95 |
+
"done": True,
|
| 96 |
+
}
|
| 97 |
+
|
| 98 |
parsed_action = parse_action(action.message)
|
| 99 |
|
| 100 |
obs, reward, done = env.step(parsed_action)
|
| 101 |
|
| 102 |
+
response = {"observation": {"echoed_message": obs}, "reward": reward, "done": done}
|
|
|
|
|
|
|
|
|
|
|
|
|
| 103 |
|
| 104 |
+
# RETURN SCORE PER TASK
|
| 105 |
if done:
|
| 106 |
response["score"] = env.get_score()
|
| 107 |
|
| 108 |
return response
|
| 109 |
|
| 110 |
+
|
| 111 |
def main():
|
| 112 |
import uvicorn
|
| 113 |
+
|
| 114 |
uvicorn.run(app, host="0.0.0.0", port=7860)
|
| 115 |
|
| 116 |
+
|
| 117 |
+
if __name__ == "__main__":
|
| 118 |
+
main()
|
| 119 |
if __name__ == "__main__":
|
| 120 |
main()
|