wehub-resource-sync
483 lines
17 KiB
Python
483 lines
17 KiB
Python
import base64
|
||
import json
|
||
import os
|
||
|
||
import streamlit as st
|
||
from dotenv import load_dotenv
|
||
from memory_utils import MemoriManager # type: ignore[unresolved-import]
|
||
from study_graph import ( # type: ignore[unresolved-import]
|
||
LearnerProfile,
|
||
StudyLog,
|
||
run_full_evaluation,
|
||
run_initial_verification,
|
||
)
|
||
|
||
load_dotenv()
|
||
|
||
|
||
st.set_page_config(
|
||
page_title="AI Study Coach with Memori",
|
||
layout="wide",
|
||
)
|
||
|
||
|
||
def _load_inline_image(path: str, height_px: int) -> str:
|
||
"""Return an inline <img> tag for a local PNG, or empty string on failure."""
|
||
try:
|
||
with open(path, "rb") as f:
|
||
encoded = base64.b64encode(f.read()).decode()
|
||
return (
|
||
f"<img src='data:image/png;base64,{encoded}' "
|
||
f"style='height:{height_px}px; width:auto; display:inline-block; "
|
||
f"vertical-align:middle; margin:0 8px;' alt='Logo'>"
|
||
)
|
||
except Exception:
|
||
return ""
|
||
|
||
|
||
# Branded title with Memori logo
|
||
memori_img_inline = _load_inline_image(
|
||
"assets/Memori_Logo.png",
|
||
height_px=85,
|
||
)
|
||
title_html = f"""
|
||
<div style='display:flex; align-items:center; width:120%; padding:8px 0;'>
|
||
<h1 style='margin:0; padding:0; font-size:2.2rem; font-weight:800; display:flex; align-items:center; gap:10px;'>
|
||
<span>Study Coach Agent with</span>
|
||
{memori_img_inline}
|
||
</h1>
|
||
</div>
|
||
"""
|
||
st.markdown(title_html, unsafe_allow_html=True)
|
||
|
||
|
||
@st.cache_resource
|
||
def get_memori_manager(openai_key: str, db_url: str | None) -> MemoriManager:
|
||
return MemoriManager(
|
||
openai_api_key=openai_key,
|
||
db_url=db_url,
|
||
)
|
||
|
||
|
||
def _ensure_state():
|
||
if "learner_profile" not in st.session_state:
|
||
st.session_state.learner_profile = None
|
||
if "quiz" not in st.session_state:
|
||
st.session_state.quiz = []
|
||
if "explanation_prompt" not in st.session_state:
|
||
st.session_state.explanation_prompt = ""
|
||
if "answers" not in st.session_state:
|
||
st.session_state.answers = []
|
||
if "explanation" not in st.session_state:
|
||
st.session_state.explanation = ""
|
||
if "last_result" not in st.session_state:
|
||
st.session_state.last_result = None
|
||
if "progress_messages" not in st.session_state:
|
||
st.session_state.progress_messages = []
|
||
|
||
|
||
def _maybe_restore_profile_from_memori(memori_mgr: MemoriManager) -> None:
|
||
"""
|
||
On fresh app loads (after refresh), try to reconstruct the learner profile
|
||
from Memori so the user doesn't have to re-enter it.
|
||
"""
|
||
if st.session_state.learner_profile is not None:
|
||
return
|
||
|
||
try:
|
||
system_prompt = (
|
||
"You are an AI study coach with access to a long-term memory store "
|
||
"about a single learner. Using that memory, reconstruct the most "
|
||
"recent learner profile that was described.\n\n"
|
||
"Respond ONLY with a JSON object with the following keys:\n"
|
||
' "name": string,\n'
|
||
' "main_goal": string,\n'
|
||
' "timeframe": string,\n'
|
||
' "subjects": list of strings,\n'
|
||
' "weekly_hours": integer,\n'
|
||
' "preferred_formats": list of strings.\n\n'
|
||
"If you truly have no stored information about the learner, respond "
|
||
"with an empty JSON object: {}"
|
||
)
|
||
response = memori_mgr.openai_client.chat.completions.create(
|
||
model="gpt-4o-mini",
|
||
response_format={"type": "json_object"},
|
||
messages=[
|
||
{"role": "system", "content": system_prompt},
|
||
{
|
||
"role": "user",
|
||
"content": "Return the last known learner profile now.",
|
||
},
|
||
],
|
||
)
|
||
raw = response.choices[0].message.content or "{}"
|
||
|
||
data = json.loads(raw)
|
||
if not isinstance(data, dict) or not data:
|
||
return
|
||
|
||
# Build LearnerProfile; if it fails, just ignore and keep requiring manual entry
|
||
profile = LearnerProfile(
|
||
name=str(data.get("name", "")).strip(),
|
||
main_goal=str(data.get("main_goal", "")).strip(),
|
||
timeframe=str(data.get("timeframe", "")).strip(),
|
||
subjects=[
|
||
str(s).strip() for s in data.get("subjects", []) if str(s).strip()
|
||
],
|
||
weekly_hours=int(data.get("weekly_hours", 1) or 1),
|
||
preferred_formats=[
|
||
str(f).strip()
|
||
for f in data.get("preferred_formats", [])
|
||
if str(f).strip()
|
||
],
|
||
)
|
||
# Basic sanity check: require at least a name and goal
|
||
if profile.name and profile.main_goal:
|
||
st.session_state.learner_profile = profile
|
||
except Exception:
|
||
# Fail silently; user can always re-enter profile if needed.
|
||
return
|
||
|
||
|
||
def sidebar_keys():
|
||
with st.sidebar:
|
||
st.subheader("🔑 API Keys")
|
||
openai_api_key_input = st.text_input(
|
||
"OpenAI API Key",
|
||
value=os.getenv("OPENAI_API_KEY", ""),
|
||
type="password",
|
||
)
|
||
memori_api_key_input = st.text_input(
|
||
"Memori API Key (optional)",
|
||
value=os.getenv("MEMORI_API_KEY", ""),
|
||
type="password",
|
||
help="Used for Memori Advanced Augmentation and higher quotas.",
|
||
)
|
||
db_url_input = st.text_input(
|
||
"CockroachDB URL",
|
||
value=os.getenv("MEMORI_DB_URL", ""),
|
||
help=(
|
||
"CockroachDB connection string using the Postgres+psycopg driver, e.g. "
|
||
"postgresql+psycopg://user:password@host:26257/database"
|
||
),
|
||
)
|
||
|
||
if st.button("Save Settings"):
|
||
if openai_api_key_input:
|
||
os.environ["OPENAI_API_KEY"] = openai_api_key_input
|
||
if memori_api_key_input:
|
||
os.environ["MEMORI_API_KEY"] = memori_api_key_input
|
||
if db_url_input:
|
||
os.environ["MEMORI_DB_URL"] = db_url_input
|
||
|
||
if openai_api_key_input or memori_api_key_input or db_url_input:
|
||
st.success("✅ Settings saved for this session")
|
||
else:
|
||
st.warning(
|
||
"Please enter at least an OpenAI API key and CockroachDB URL"
|
||
)
|
||
|
||
st.markdown("---")
|
||
st.markdown("### ℹ️ About")
|
||
st.markdown(
|
||
"""
|
||
This is an **AI Study Coach** demo built for Memori:
|
||
|
||
- Plans and tracks study sessions.
|
||
- Uses **LangGraph** to verify understanding with quizzes + explanations.
|
||
- Uses **Memori v3** as long-term learning memory.
|
||
"""
|
||
)
|
||
|
||
|
||
def study_plan_tab(memori_mgr: MemoriManager):
|
||
st.markdown("#### 🧭 Study Plan & Learner Profile")
|
||
|
||
with st.form("profile_form"):
|
||
col1, col2 = st.columns(2)
|
||
with col1:
|
||
name = st.text_input("Name or handle", placeholder="e.g. 3rdSon")
|
||
main_goal = st.text_input(
|
||
"Main goal",
|
||
placeholder="e.g. Pass AWS SAA, master LangGraph, finish CS50",
|
||
)
|
||
timeframe = st.text_input(
|
||
"Timeframe",
|
||
placeholder="e.g. 3 months, 6 weeks",
|
||
)
|
||
with col2:
|
||
weekly_hours = st.number_input(
|
||
"Planned study hours per week", min_value=1, max_value=80, value=7
|
||
)
|
||
subjects = st.text_input(
|
||
"Subjects / topics (comma-separated)",
|
||
placeholder="e.g. LangGraph, Memori, algorithms",
|
||
)
|
||
preferred_formats = st.multiselect(
|
||
"Preferred learning formats",
|
||
options=[
|
||
"videos",
|
||
"docs",
|
||
"practice problems",
|
||
"flashcards",
|
||
"projects",
|
||
],
|
||
)
|
||
|
||
submitted = st.form_submit_button("Save Profile")
|
||
|
||
if submitted:
|
||
if not name or not main_goal or not timeframe:
|
||
st.error("Please fill in at least name, main goal, and timeframe.")
|
||
return
|
||
profile = LearnerProfile(
|
||
name=name.strip(),
|
||
main_goal=main_goal.strip(),
|
||
timeframe=timeframe.strip(),
|
||
subjects=[s.strip() for s in subjects.split(",") if s.strip()],
|
||
weekly_hours=weekly_hours,
|
||
preferred_formats=preferred_formats,
|
||
)
|
||
st.session_state.learner_profile = profile
|
||
|
||
# Log structured profile into Memori so it can be recalled later
|
||
try:
|
||
memori_mgr.log_learner_profile(profile.model_dump())
|
||
st.success("✅ Profile saved and stored in Memori.")
|
||
except Exception as e:
|
||
st.warning(f"Profile saved in session, but Memori logging failed: {e}")
|
||
|
||
if st.session_state.learner_profile:
|
||
p: LearnerProfile = st.session_state.learner_profile
|
||
st.markdown("##### Current Profile")
|
||
st.write(
|
||
{
|
||
"name": p.name,
|
||
"goal": p.main_goal,
|
||
"timeframe": p.timeframe,
|
||
"subjects": p.subjects,
|
||
"weekly_hours": p.weekly_hours,
|
||
"preferred_formats": p.preferred_formats,
|
||
}
|
||
)
|
||
|
||
|
||
def today_session_tab(memori_mgr: MemoriManager):
|
||
st.markdown("#### 📅 Today’s Study Session")
|
||
|
||
profile: LearnerProfile | None = st.session_state.learner_profile
|
||
if not profile:
|
||
st.info("Set up your study profile first in the **Study Plan** tab.")
|
||
return
|
||
|
||
col1, col2 = st.columns(2)
|
||
with col1:
|
||
topic = st.text_input(
|
||
"What did you study today?", placeholder="e.g. LangGraph basics"
|
||
)
|
||
duration = st.number_input(
|
||
"How many minutes did you study?",
|
||
min_value=5,
|
||
max_value=600,
|
||
value=45,
|
||
)
|
||
resource_type = st.selectbox(
|
||
"Main resource type",
|
||
options=["video", "article", "course", "problems", "other"],
|
||
)
|
||
with col2:
|
||
perceived_difficulty = st.selectbox(
|
||
"How difficult was it?",
|
||
options=["easy", "medium", "hard"],
|
||
)
|
||
mood = st.text_input(
|
||
"How did you feel?",
|
||
placeholder="e.g. focused, tired, motivated, frustrated",
|
||
)
|
||
notes = st.text_area("Any additional notes?", height=80)
|
||
|
||
if st.button("Generate quiz & explanation check", type="primary"):
|
||
if not topic:
|
||
st.error("Please enter what you studied today.")
|
||
return
|
||
|
||
log = StudyLog(
|
||
topic=topic.strip(),
|
||
duration_minutes=duration,
|
||
resource_type=resource_type,
|
||
perceived_difficulty=perceived_difficulty,
|
||
mood=mood.strip() or None,
|
||
free_notes=notes.strip() or None,
|
||
)
|
||
try:
|
||
mgr = memori_mgr # alias
|
||
initial = run_initial_verification(
|
||
profile=profile, log=log, llm_client=mgr.openai_client
|
||
)
|
||
st.session_state.quiz = initial.quiz
|
||
st.session_state.explanation_prompt = initial.explanation_prompt
|
||
st.session_state.answers = ["" for _ in initial.quiz]
|
||
st.session_state.explanation = ""
|
||
st.session_state.current_log = log
|
||
except Exception as e:
|
||
st.error(f"Failed to generate quiz: {e}")
|
||
|
||
# Show quiz if available
|
||
quiz = st.session_state.quiz
|
||
if quiz:
|
||
st.markdown("##### 🧪 Quick Understanding Check")
|
||
new_answers: list[str] = []
|
||
for i, q in enumerate(quiz):
|
||
ans = st.text_area(
|
||
f"Q{i + 1}. {q.question}",
|
||
value=(
|
||
st.session_state.answers[i]
|
||
if i < len(st.session_state.answers)
|
||
else ""
|
||
),
|
||
height=80,
|
||
)
|
||
new_answers.append(ans)
|
||
st.session_state.answers = new_answers
|
||
|
||
st.markdown("##### ✍️ Explain in your own words")
|
||
st.session_state.explanation = st.text_area(
|
||
"Explanation",
|
||
value=st.session_state.explanation,
|
||
placeholder=st.session_state.explanation_prompt,
|
||
height=160,
|
||
)
|
||
|
||
if st.button("Evaluate my understanding", type="secondary"):
|
||
try:
|
||
mgr = memori_mgr
|
||
log: StudyLog = st.session_state.current_log
|
||
result = run_full_evaluation(
|
||
profile=profile,
|
||
log=log,
|
||
user_quiz_answers=st.session_state.answers,
|
||
user_explanation=st.session_state.explanation,
|
||
llm_client=mgr.openai_client,
|
||
)
|
||
st.session_state.last_result = result
|
||
|
||
# Log study session into Memori
|
||
summary = (
|
||
f"Study session summary:\n"
|
||
f"- Topic: {log.topic}\n"
|
||
f"- Duration (min): {log.duration_minutes}\n"
|
||
f"- Resource: {log.resource_type}\n"
|
||
f"- Difficulty: {log.perceived_difficulty}\n"
|
||
f"- Mood: {log.mood or 'N/A'}\n"
|
||
f"- Score: {result.score}\n"
|
||
f"- Feedback: {result.feedback or ''}\n"
|
||
f"- Next step: {result.next_step_recommendation or ''}"
|
||
)
|
||
mgr.log_study_session(summary)
|
||
|
||
except Exception as e:
|
||
st.error(f"Failed to evaluate and log session: {e}")
|
||
|
||
# Show last result
|
||
if st.session_state.last_result:
|
||
r = st.session_state.last_result
|
||
st.markdown("##### 🎯 Result")
|
||
if r.score is not None:
|
||
st.metric("Understanding score", f"{r.score}/100")
|
||
if r.feedback:
|
||
st.markdown("**Feedback**")
|
||
st.write(r.feedback)
|
||
if r.next_step_recommendation:
|
||
st.markdown("**Recommended next step**")
|
||
st.write(r.next_step_recommendation)
|
||
|
||
|
||
def progress_tab(memori_mgr: MemoriManager):
|
||
st.markdown("#### 📈 Progress & Memory (Memori-powered)")
|
||
st.markdown(
|
||
"Ask questions about your learning history, weak/strong topics, or patterns.\n\n"
|
||
"Examples:\n"
|
||
"- *What are my weakest topics right now?*\n"
|
||
"- *When do I usually perform best?*\n"
|
||
"- *Do I learn better from videos or practice problems?*"
|
||
)
|
||
|
||
# Display chat history
|
||
for message in st.session_state.progress_messages:
|
||
with st.chat_message(message["role"]):
|
||
st.markdown(message["content"])
|
||
|
||
# Chat input
|
||
prompt = st.chat_input("Ask about your learning progress…")
|
||
if prompt:
|
||
# User message
|
||
st.session_state.progress_messages.append({"role": "user", "content": prompt})
|
||
with st.chat_message("user"):
|
||
st.markdown(prompt)
|
||
|
||
# Assistant response via Memori
|
||
with st.chat_message("assistant"):
|
||
with st.spinner("🔍 Checking your study memories…"):
|
||
try:
|
||
answer = memori_mgr.summarize_progress(prompt)
|
||
st.session_state.progress_messages.append(
|
||
{"role": "assistant", "content": answer}
|
||
)
|
||
st.markdown(answer)
|
||
except Exception as e:
|
||
err = f"❌ Failed to query Memori: {e}"
|
||
st.session_state.progress_messages.append(
|
||
{"role": "assistant", "content": err}
|
||
)
|
||
st.error(err)
|
||
|
||
|
||
def main():
|
||
sidebar_keys()
|
||
_ensure_state()
|
||
|
||
try:
|
||
db_url = os.getenv("MEMORI_DB_URL", "") or None
|
||
openai_key = os.getenv("OPENAI_API_KEY", "")
|
||
memori_mgr = get_memori_manager(openai_key, db_url)
|
||
except Exception as e:
|
||
st.error(
|
||
f"Failed to initialize Memori / OpenAI. "
|
||
f"Check your OPENAI_API_KEY and DB settings. Details: {e}"
|
||
)
|
||
return
|
||
|
||
# After Memori is ready, try to restore learner profile from Memori on fresh loads.
|
||
# This lets the app remember your profile across refreshes and new runs.
|
||
if st.session_state.learner_profile is None:
|
||
profile_dict: dict | None = None
|
||
try:
|
||
profile_dict = memori_mgr.get_latest_learner_profile()
|
||
except Exception:
|
||
profile_dict = None
|
||
|
||
if profile_dict:
|
||
try:
|
||
st.session_state.learner_profile = LearnerProfile(**profile_dict)
|
||
except Exception:
|
||
# Fall back to LLM-based reconstruction if structured load fails.
|
||
_maybe_restore_profile_from_memori(memori_mgr)
|
||
else:
|
||
# No structured profile found – try to reconstruct via Memori + LLM.
|
||
_maybe_restore_profile_from_memori(memori_mgr)
|
||
|
||
tab1, tab2, tab3 = st.tabs(
|
||
["🧭 Study Plan", "📅 Today’s Session", "📈 Progress & Memory"]
|
||
)
|
||
|
||
with tab1:
|
||
study_plan_tab(memori_mgr)
|
||
with tab2:
|
||
today_session_tab(memori_mgr)
|
||
with tab3:
|
||
progress_tab(memori_mgr)
|
||
|
||
|
||
if __name__ == "__main__":
|
||
main()
|