mirror of
https://github.com/dataforcanada/d4c-service-geo-assistant.git
synced 2026-06-13 22:41:01 +02:00
Soumya Ranjan Mohanty
be8affaa6c
* Add docs to download overture places data & ollama model * Hit local overture parquet files * Add osx gitignore * Add .env.example * Make overture data source selectable using .env * Add pytest marker to set right ENV vars during CI --------- Co-authored-by: Daniel Wiesmann <yellowcap@users.noreply.github.com>
103 lines
3.0 KiB
Python
103 lines
3.0 KiB
Python
"""Tools for summarizing satellite images using LLM-based analysis."""
|
|
|
|
import os
|
|
from typing import Annotated, Optional
|
|
import dspy
|
|
from langchain_core.tools import tool
|
|
from langgraph.types import Command
|
|
from langchain_core.messages import ToolMessage
|
|
from langchain_core.tools.base import InjectedToolCallId
|
|
|
|
import dotenv
|
|
|
|
dotenv.load_dotenv()
|
|
|
|
|
|
class SatImgSummary(dspy.Signature):
|
|
"Describe things you see in the satellite image."
|
|
|
|
img: dspy.Image = dspy.InputField(desc="A satellite image")
|
|
answer: str = dspy.OutputField(desc="Description of the image")
|
|
|
|
|
|
class SatImgSummaryAgent(dspy.Module):
|
|
"""Agent for generating summaries of satellite images using an LLM."""
|
|
|
|
def __init__(
|
|
self,
|
|
model: str = os.environ.get("OLLAMA_IMAGE_MODEL", "ministral-3:14b-cloud"),
|
|
api_base: str = os.environ.get("OLLAMA_BASE_URL", "http://localhost:11434"),
|
|
temperature: float = 0.5,
|
|
max_tokens: int = 4_096,
|
|
) -> None:
|
|
"""Initialize the satellite image summary agent.
|
|
|
|
Args:
|
|
model: The Ollama model to use for summarization
|
|
api_base: Base URL for the Ollama API
|
|
temperature: Sampling temperature (0-1)
|
|
max_tokens: Maximum tokens to generate
|
|
"""
|
|
super().__init__()
|
|
self.ollama_model = dspy.LM(
|
|
model=f"ollama/{model}",
|
|
api_base=api_base,
|
|
api_key="",
|
|
temperature=temperature,
|
|
max_tokens=max_tokens,
|
|
)
|
|
dspy.configure(lm=self.ollama_model)
|
|
self.summarizer = dspy.Predict(SatImgSummary)
|
|
|
|
def forward(self, img_url: str) -> dspy.Prediction:
|
|
"""Generate a summary for the given image URL.
|
|
|
|
Args:
|
|
img_url: URL of the image to summarize
|
|
|
|
Returns:
|
|
dspy.Prediction containing the image summary
|
|
"""
|
|
return self.summarizer(img=dspy.Image(img_url))
|
|
|
|
|
|
# Singleton instance to avoid repeated initialization
|
|
_SUMMARIZER_AGENT = SatImgSummaryAgent()
|
|
|
|
|
|
@tool
|
|
def summarize_sat_img(
|
|
img_url: str,
|
|
tool_call_id: Annotated[Optional[str], InjectedToolCallId] = None,
|
|
) -> Command:
|
|
"""Summarize the contents of a satellite image using an LLM.
|
|
|
|
Args:
|
|
img_url: URL of the satellite image to analyze
|
|
tool_call_id: Optional ID for tracking the tool call
|
|
|
|
Returns:
|
|
Command containing the image summary and metadata
|
|
|
|
Raises:
|
|
ValueError: If the image URL is invalid or the image cannot be processed
|
|
"""
|
|
if not img_url or not isinstance(img_url, str):
|
|
raise ValueError("img_url must be a non-empty string")
|
|
|
|
summary = _SUMMARIZER_AGENT(img_url)
|
|
message_content = summary.answer
|
|
artifact = {"img_url": img_url}
|
|
|
|
return Command(
|
|
update={
|
|
"messages": [
|
|
ToolMessage(
|
|
content=message_content,
|
|
artifact=artifact,
|
|
tool_call_id=tool_call_id,
|
|
)
|
|
]
|
|
}
|
|
)
|