Spaces:
Paused
Paused
Update utils/planner.py
Browse files- utils/planner.py +14 -57
utils/planner.py
CHANGED
|
@@ -1,64 +1,20 @@
|
|
| 1 |
# utils/planner.py
|
| 2 |
|
| 3 |
-
import openai
|
| 4 |
import os
|
| 5 |
-
from openai import OpenAI
|
| 6 |
-
from dotenv import load_dotenv
|
| 7 |
import json
|
| 8 |
-
load_dotenv()
|
| 9 |
-
|
| 10 |
-
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
|
| 11 |
-
|
| 12 |
-
SYSTEM_INSTRUCTIONS = """
|
| 13 |
-
You are a scene planning assistant for an AI image generation system.
|
| 14 |
-
Your job is to take the user's prompt and return a structured JSON with:
|
| 15 |
-
- scene (environment, setting)
|
| 16 |
-
- subject (main actor)
|
| 17 |
-
- objects (main product or items)
|
| 18 |
-
- layout (foreground/background elements and their placement)
|
| 19 |
-
- rules (validation rules to ensure visual correctness)
|
| 20 |
-
Respond ONLY in raw JSON format. Do NOT include explanations.
|
| 21 |
-
"""
|
| 22 |
-
|
| 23 |
-
def extract_scene_plan(prompt: str) -> dict:
|
| 24 |
-
try:
|
| 25 |
-
response = client.chat.completions.create(
|
| 26 |
-
model="gpt-4o-mini-2024-07-18",
|
| 27 |
-
messages=[
|
| 28 |
-
{"role": "system", "content": SYSTEM_INSTRUCTIONS},
|
| 29 |
-
{"role": "user", "content": prompt}
|
| 30 |
-
],
|
| 31 |
-
temperature=0.3,
|
| 32 |
-
max_tokens=500
|
| 33 |
-
)
|
| 34 |
-
json_output = response.choices[0].message.content
|
| 35 |
-
print("🧠 Scene Plan (Raw):", json_output)
|
| 36 |
-
return json.loads(json_output) # Be cautious: Use `json.loads()` if possible
|
| 37 |
-
|
| 38 |
-
except Exception as e:
|
| 39 |
-
print("❌ extract_scene_plan() Error:", e)
|
| 40 |
-
return {
|
| 41 |
-
"scene": None,
|
| 42 |
-
"subject": None,
|
| 43 |
-
"objects": [],
|
| 44 |
-
"layout": {},
|
| 45 |
-
"rules": {}
|
| 46 |
-
}
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
# utils/planner.py
|
| 50 |
-
|
| 51 |
-
import openai
|
| 52 |
-
import os
|
| 53 |
-
from openai import OpenAI
|
| 54 |
from dotenv import load_dotenv
|
| 55 |
-
import
|
| 56 |
-
load_dotenv()
|
| 57 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
|
| 59 |
|
|
|
|
| 60 |
# 🧠 Scene Plan Extractor
|
| 61 |
-
|
|
|
|
| 62 |
You are a scene planning assistant for an AI image generation system.
|
| 63 |
Your job is to take the user's prompt and return a structured JSON with:
|
| 64 |
- scene (environment, setting)
|
|
@@ -74,7 +30,7 @@ def extract_scene_plan(prompt: str) -> dict:
|
|
| 74 |
response = client.chat.completions.create(
|
| 75 |
model="gpt-4o-mini-2024-07-18",
|
| 76 |
messages=[
|
| 77 |
-
{"role": "system", "content":
|
| 78 |
{"role": "user", "content": prompt}
|
| 79 |
],
|
| 80 |
temperature=0.3,
|
|
@@ -94,19 +50,19 @@ def extract_scene_plan(prompt: str) -> dict:
|
|
| 94 |
"rules": {}
|
| 95 |
}
|
| 96 |
|
| 97 |
-
#
|
|
|
|
|
|
|
| 98 |
def generate_prompt_variations_from_scene(scene_plan: dict, base_prompt: str, n: int = 3) -> list:
|
| 99 |
try:
|
| 100 |
system_msg = f"""
|
| 101 |
You are a creative prompt variation generator for an AI image generation system.
|
| 102 |
-
|
| 103 |
Given a base user prompt and its structured scene plan, generate {n} diverse image generation prompts.
|
| 104 |
Each prompt should:
|
| 105 |
- Be visually rich and descriptive
|
| 106 |
- Include stylistic or contextual variation
|
| 107 |
- Reference the same product and environment
|
| 108 |
- Stay faithful to the base prompt and extracted plan
|
| 109 |
-
|
| 110 |
Respond ONLY with a JSON array of strings. No explanations.
|
| 111 |
"""
|
| 112 |
|
|
@@ -129,4 +85,5 @@ Respond ONLY with a JSON array of strings. No explanations.
|
|
| 129 |
|
| 130 |
except Exception as e:
|
| 131 |
print("❌ generate_prompt_variations_from_scene() Error:", e)
|
| 132 |
-
return [base_prompt] # fallback to original
|
|
|
|
|
|
| 1 |
# utils/planner.py
|
| 2 |
|
|
|
|
| 3 |
import os
|
|
|
|
|
|
|
| 4 |
import json
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
from dotenv import load_dotenv
|
| 6 |
+
from openai import OpenAI
|
|
|
|
| 7 |
|
| 8 |
+
# ----------------------------
|
| 9 |
+
# 🔐 Load Environment & Client
|
| 10 |
+
# ----------------------------
|
| 11 |
+
load_dotenv()
|
| 12 |
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
|
| 13 |
|
| 14 |
+
# ----------------------------
|
| 15 |
# 🧠 Scene Plan Extractor
|
| 16 |
+
# ----------------------------
|
| 17 |
+
SCENE_SYSTEM_INSTRUCTIONS = """
|
| 18 |
You are a scene planning assistant for an AI image generation system.
|
| 19 |
Your job is to take the user's prompt and return a structured JSON with:
|
| 20 |
- scene (environment, setting)
|
|
|
|
| 30 |
response = client.chat.completions.create(
|
| 31 |
model="gpt-4o-mini-2024-07-18",
|
| 32 |
messages=[
|
| 33 |
+
{"role": "system", "content": SCENE_SYSTEM_INSTRUCTIONS},
|
| 34 |
{"role": "user", "content": prompt}
|
| 35 |
],
|
| 36 |
temperature=0.3,
|
|
|
|
| 50 |
"rules": {}
|
| 51 |
}
|
| 52 |
|
| 53 |
+
# ----------------------------
|
| 54 |
+
# 🧠 Prompt Variation Generator
|
| 55 |
+
# ----------------------------
|
| 56 |
def generate_prompt_variations_from_scene(scene_plan: dict, base_prompt: str, n: int = 3) -> list:
|
| 57 |
try:
|
| 58 |
system_msg = f"""
|
| 59 |
You are a creative prompt variation generator for an AI image generation system.
|
|
|
|
| 60 |
Given a base user prompt and its structured scene plan, generate {n} diverse image generation prompts.
|
| 61 |
Each prompt should:
|
| 62 |
- Be visually rich and descriptive
|
| 63 |
- Include stylistic or contextual variation
|
| 64 |
- Reference the same product and environment
|
| 65 |
- Stay faithful to the base prompt and extracted plan
|
|
|
|
| 66 |
Respond ONLY with a JSON array of strings. No explanations.
|
| 67 |
"""
|
| 68 |
|
|
|
|
| 85 |
|
| 86 |
except Exception as e:
|
| 87 |
print("❌ generate_prompt_variations_from_scene() Error:", e)
|
| 88 |
+
return [base_prompt] # fallback to original
|
| 89 |
+
|