Created an annotation using instructor

examples/annotation/Readme.md

@@ -0,0 +1,7 @@
+# Introduction
+
+This showcases a simple streamlit module which can be used to do data annotation for files in a specific table.
+
+Make sure to install the dependencies first with `uv pip install -r requirements.txt`
+
+To populate the table, run `main.py`. This should generate ~20 different todos insert it into the table and mark it as unannotated. Once you've done so, you can then boot up the `annotate.py` file using the command `streamlit run annotate.py`

examples/annotation/annotate.py

@@ -0,0 +1,72 @@
+import streamlit as st
+import sqlite3
+
+
+def fetch_unannotated_todos():
+    with sqlite3.connect("tutorial.db") as con:
+        cur = con.cursor()
+        cur.execute(
+            "SELECT title, description, annotated,id FROM todos WHERE annotated = FALSE"
+        )
+        todos = cur.fetchall()
+
+    return [
+        {"title": title, "description": description, "annotated": annotated, "id": id}
+        for title, description, annotated, id in todos
+    ]
+
+
+def display_todos(todos):
+    st.write("### Unannotated Todos")
+    for todo in todos:
+        st.write(f'({todo["id"]}) {todo["title"]}')
+        if st.button(f"Select {todo['id']}"):
+            st.session_state.curr_selected_todo = todo["id"]
+
+
+st.title("Todo Annotation")
+
+# Initialize session state
+if "curr_selected_todo" not in st.session_state:
+    st.session_state.curr_selected_todo = None
+
+
+def render_selected_todo():
+    if st.session_state.curr_selected_todo is not None:
+        with sqlite3.connect("tutorial.db") as con:
+            cur = con.cursor()
+            cur.execute(
+                "SELECT original_prompt,title, description FROM todos WHERE id = ?",
+                (st.session_state.curr_selected_todo,),
+            )
+            todo_data = cur.fetchone()
+            if todo_data:
+                st.write("Original Prompt: " + todo_data[0])
+                new_title = st.text_input("Title", value=todo_data[1])
+                new_description = st.text_area("Description", value=todo_data[2])
+                if st.button("Update"):
+                    with sqlite3.connect("tutorial.db") as con:
+                        cur = con.cursor()
+                        cur.execute(
+                            "UPDATE todos SET title = ?, description = ?, annotated = ? WHERE id = ?",
+                            (
+                                new_title,
+                                new_description,
+                                True,
+                                st.session_state.curr_selected_todo,
+                            ),
+                        )
+                        con.commit()
+                        st.success("Todo updated successfully!")
+            else:
+                st.write("Selected todo not found.")
+    else:
+        st.write("No todo selected.")
+
+
+render_selected_todo()
+unannotated_todos = fetch_unannotated_todos()
+if unannotated_todos:
+    display_todos(unannotated_todos)
+else:
+    st.write("No unannotated todos found.")

examples/annotation/main.py

@@ -0,0 +1,70 @@
+import instructor
+from typing import List
+from openai import AsyncOpenAI
+from asyncio import run
+from tqdm.asyncio import tqdm_asyncio as asyncio
+from pydantic import BaseModel, Field
+import sqlite3
+
+
+client = instructor.from_openai(AsyncOpenAI())
+
+
+class TodoItem(BaseModel):
+    """
+    This is a schema that represents an actionable item which the user needs to consider
+    """
+
+    title: str = Field(..., description="This is a title for the todo item")
+    description: str = Field(
+        ...,
+        description="This is a description that explains a plan of action for the todo",
+    )
+
+
+async def extract_todo(user_query: str):
+    res = await client.chat.completions.create(
+        model="gpt-3.5-turbo",
+        messages=[
+            {
+                "role": "system",
+                "content": "You are a world class system that excels at extracting todo items from a user query",
+            },
+            {"role": "user", "content": user_query},
+        ],
+        response_model=List[TodoItem],
+    )
+    return [(item, user_query) for item in res]
+
+
+async def process_todos(items):
+    coros = [extract_todo(item) for item in items]
+    results = await asyncio.gather(*coros)
+    return [item for sublist in results for item in sublist]
+
+
+if __name__ == "__main__":
+    con = sqlite3.connect("tutorial.db")
+    cur = con.cursor()
+    cur.execute(
+        "CREATE TABLE IF NOT EXISTS todos(id INTEGER PRIMARY KEY AUTOINCREMENT, annotated BOOLEAN DEFAULT FALSE, title TEXT, description TEXT, original_prompt TEXT)"
+    )
+
+    data = [
+        "This week I need to finalize the project report, schedule a meeting with the team, prepare the presentation slides, submit the budget review, and send the client update emails.",
+        "Next week I must organize the department outing, update the project timeline, review the new intern applications, and coordinate the quarterly webinars.",
+        "Tomorrow I should finalize the contract details, call the supplier for an update, draft the monthly newsletter, and check the inventory status.",
+        "By the end of this month, I need to complete the performance reviews, plan the training sessions, archive old project files, and renew the software licenses.",
+        "This Friday I have to prepare the weekly sales report, confirm the client appointments, oversee the network upgrade, and document the audit findings.",
+    ]
+
+    todos: List[TodoItem] = run(process_todos(data))
+
+    with sqlite3.connect("tutorial.db") as con:
+        cur = con.cursor()
+        for todo, original_query in todos:
+            cur.execute(
+                "INSERT INTO todos (title, description,original_prompt) VALUES (?, ?,?)",
+                (todo.title, todo.description, original_query),
+            )
+        con.commit()

examples/annotation/requirements.txt

@@ -0,0 +1,5 @@
+openai==1.23.6
+instructor==1.2.3
+pydantic==2.7.0
+typer==0.12.3
+streamlit==1.33.0