Spaces:
Sleeping
Sleeping
File size: 6,966 Bytes
c8f3cbb |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 |
import os
import base64
import json
from datetime import datetime
import traceback
import gradio as gr
from huggingface_hub import HfApi, InferenceClient
from pydantic import BaseModel, Field
# -------------------------------
# Environment variables / Constants
# -------------------------------
HF_DATASET_REPO = os.environ.get("HF_DATASET_REPO", "OppaAI/Robot_MCP")
HF_VLM_MODEL = os.environ.get("HF_VLM_MODEL", "Qwen/Qwen2.5-VL-7B-Instruct")
# -------------------------------
# Pydantic schema for the tool payload
# -------------------------------
class RobotWatchPayload(BaseModel):
"""
Defines the expected input structure for the robot VLM analysis tool.
Attributes:
hf_token (str): Your Hugging Face API token.
robot_id (str): Identifier for the robot (default "unknown").
image_b64 (str): Base64 encoded image string to analyze.
"""
hf_token: str = Field(description="Your Hugging Face API token.")
robot_id: str = Field(description="Robot identifier.", default="unknown")
image_b64: str = Field(description="Base64 encoded image data.")
# -------------------------------
# Helper function: Upload image to Hugging Face dataset
# -------------------------------
def upload_image(image_b64: str, hf_token: str):
"""
Decodes a base64 image string, saves it locally, and uploads to Hugging Face dataset.
Args:
image_b64 (str): Base64 encoded image data.
hf_token (str): Hugging Face API token.
Returns:
tuple: (local_path, hf_url, filename, size_bytes)
"""
try:
image_bytes = base64.b64decode(image_b64)
os.makedirs("/tmp", exist_ok=True)
# Generate unique timestamped filename
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S_%f")
local_path = f"/tmp/robot_img_{timestamp}.jpg"
# Save locally
with open(local_path, "wb") as f:
f.write(image_bytes)
filename = f"robot_{timestamp}.jpg"
# Upload to Hugging Face dataset
api = HfApi()
api.upload_file(
path_or_fileobj=local_path,
path_in_repo=f"tmp/{filename}",
repo_id=HF_DATASET_REPO,
repo_type="dataset",
token=hf_token
)
hf_url = f"https://huggingface.co/datasets/{HF_DATASET_REPO}/resolve/main/tmp/{filename}"
return local_path, hf_url, filename, len(image_bytes)
except Exception:
traceback.print_exc()
return None, None, None, 0
# -------------------------------
# Helper function: Parse JSON safely
# -------------------------------
def safe_parse_json_from_text(text: str):
"""
Attempts to parse JSON from text returned by the VLM model.
Strips any leading/trailing characters and handles malformed responses.
Args:
text (str): Raw text output from the model.
Returns:
dict or None: Parsed JSON dictionary, or None if parsing fails.
"""
if not text:
return None
try:
return json.loads(text)
except:
pass
cleaned = text.strip().strip("`").strip()
if cleaned.lower().startswith("json"):
cleaned = cleaned[4:].strip()
try:
start = cleaned.find("{")
end = cleaned.rfind("}")
return json.loads(cleaned[start:end + 1])
except:
return None
# -------------------------------
# Core VLM analysis function
# -------------------------------
def run_vlm_analysis(payload: RobotWatchPayload):
"""
Main logic for analyzing an image using Hugging Face VLM model.
Args:
payload (RobotWatchPayload): Validated payload containing token, robot_id, and image.
Returns:
dict: Analysis result including description, objects, and raw VLM output.
"""
hf_token = payload.hf_token
image_b64 = payload.image_b64
robot_id = payload.robot_id
# Upload the image to Hugging Face dataset
_, hf_url, _, size_bytes = upload_image(image_b64, hf_token)
if not hf_url:
return {"error": "Image upload failed"}
# System prompt instructs VLM to return strict JSON
system_prompt = """
Respond in STRICT JSON ONLY. Put more details in Description. Ensure all the fields are never empty; list general items if specific ones are not clear.
{
"description": "...",
"environment": "...",
"indoor_or_outdoor": "...",
"lighting_condition": "..."
"human": "...",
"animals": "...",
"objects": [],
"hazards": "...",
}
"""
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": [
{"type": "text", "text": "Analyze the image."},
{"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{image_b64}"}}
]}
]
client = InferenceClient(token=hf_token)
try:
resp = client.chat.completions.create(
model=HF_VLM_MODEL,
messages=messages,
max_tokens=500,
temperature=0.1
)
except Exception as e:
return {"status": "error", "message": str(e)}
vlm_output = resp.choices[0].message.content.strip()
parsed = safe_parse_json_from_text(vlm_output) or {}
return {
"status": "success",
"robot_id": robot_id,
"file_size_bytes": size_bytes,
"image_url": hf_url,
"result": parsed,
"vlm_raw": vlm_output
}
# -------------------------------
# Gradio interface function
# -------------------------------
def robot_watch(
hf_token_input: str,
robot_id_input: str,
image_b64_input: str
):
"""
Gradio wrapper for run_vlm_analysis.
Converts individual fields into Pydantic model and calls core logic.
Args:
hf_token_input (str): Hugging Face API token input from UI.
robot_id_input (str): Robot ID input from UI.
image_b64_input (str): Base64 image input from UI.
Returns:
dict: Result from run_vlm_analysis.
"""
if not image_b64_input:
return {"error": "Base64 image string is empty."}
# Create the payload instance
payload_instance = RobotWatchPayload(
hf_token=hf_token_input,
robot_id=robot_id_input,
image_b64=image_b64_input
)
# Run core analysis
result = run_vlm_analysis(payload_instance)
return result
# -------------------------------
# Gradio App
# -------------------------------
app = gr.Interface(
fn=robot_watch,
inputs=[
gr.Textbox(label="Hugging Face Token", lines=1),
gr.Textbox(label="Robot ID", lines=1, value="unknown"),
gr.Textbox(label="Image Base64 String", lines=5)
],
outputs=gr.Json(label="Tool Output"),
title="Robot CV MCP Server",
description="Interface for robot VLM analysis using individual fields, including base64 image string.",
api_name="predict"
)
if __name__ == "__main__":
# Launch Gradio app with MCP server enabled
app.launch(mcp_server=True) |