EOD Reporter

An automated End of Day report generator for interns. Log work activities throughout the day, then let AI compile everything into polished EOD reports -- posted to Microsoft Teams and auto-filled on the Internity (aufccs.org) internship platform.

How It Works

1. Log activities throughout the day (CLI or web UI) -- each entry is timestamped and auto-categorized into morning, afternoon, or evening
2. At the scheduled time (or manually), a LangGraph agent generates a narrative EOD report, self-reviews it, and revises if needed
3. Preview and edit the generated report in the web UI
4. Post to Teams with one click, or let auto-post handle it
5. Submit to Internity -- a browser opens, fills the aufccs.org form automatically, and waits for you to review and click Submit

Tech Stack

Component	Technology
LLM	Gemini 2.5 Flash via langchain-google-genai
Agent Framework	LangGraph (stateful graph with self-review loop)
Web Framework	FastAPI + Jinja2 + HTMX
CLI	Typer + Rich
Database	SQLite + SQLAlchemy
Scheduler	APScheduler
Teams Integration	Power Automate HTTP trigger
Internity Automation	Playwright (browser automation)
Package Manager	uv

Prerequisites

• Python 3.12+
• uv package manager
• Google AI Studio API key (get one here)
• Power Automate flow with HTTP trigger for Teams posting
• Playwright browsers: uv run playwright install chromium

Setup

# Clone and install (Python 3.12 required — 3.13/3.14 will cause build errors)
git clone https://github.com/jansencruz23/internship-eod-automation.git
cd internship-eod-automation
uv python install 3.12
uv venv --python 3.12
uv sync

# Install Playwright browser
uv run playwright install chromium

# Configure environment
cp .env.example .env
# Edit .env with your credentials

Environment Variables

Variable	Description	Default
GOOGLE_API_KEY	Gemini API key from AI Studio	required
POWER_AUTOMATE_URL	Power Automate HTTP trigger URL	required
INTERNITY_USERNAME	aufccs.org login email	required for Internity
INTERNITY_PASSWORD	aufccs.org login password	required for Internity
INTERNITY_FORM_URL	Full URL to the Internity EOD form	required for Internity
EOD_SCHEDULE_TIME	Daily generation time (HH:MM)	17:00
MODEL_NAME	Gemini model to use	gemini-2.5-flash

Usage

Web Interface

uv run uvicorn app.main:app --reload

Open http://localhost:8000. The dashboard lets you:

• Log activities with the quick-entry form
• View activities grouped by time period (morning, afternoon, evening)
• Generate EOD reports on demand
• Preview, edit, and post reports to Teams
• Submit to Internity -- opens a visible browser, fills the form, you click Submit
• Configure scheduler time, Teams auto-post, and Internity auto-fill toggles

CLI

# Log an activity
uv run eod log "Had team huddle for Project 2"

# Log with time period override
uv run eod log "Reviewed PRs" --time morning

# List today's activities
uv run eod list

# Generate an EOD report
uv run eod compile

# Post the report to Teams
uv run eod post

# Submit to Internity (opens browser, fills form, you click Submit)
uv run eod internity

# Submit to Internity automatically (no user interaction)
uv run eod internity --auto-submit

# Test Internity login
uv run eod test-internity

# Debug Internity selectors (opens visible browser with Inspector)
uv run eod test-internity --headed

# Test the Power Automate connection
uv run eod test-webhook

Scheduler

The built-in scheduler (APScheduler) automatically generates EOD reports at the configured time. Adjust the time and toggles from the web UI -- no restart required.

• Auto-post to Teams OFF (default): Generates a draft. Visit /reports/preview to review and post.
• Auto-post to Teams ON: Generates and immediately posts to Teams.
• Auto-fill Internity ON: Opens a browser with the Internity form pre-filled for manual review and submission.

Project Structure

app/
  core/              Config and database setup
  models/            SQLAlchemy ORM models (Activity, EODReport, AppSettings)
  schemas/           Pydantic models (InternityEOD, InternityTask)
  repositories/      Data access layer (generic base + specific repos)
  services/teams/    TeamsPoster (Power Automate HTTP post)
  services/internity/ InternityPoster (Playwright browser automation)
  agent/teams/       LangGraph agent for Teams EOD (nodes, graph, prompts, state)
  agent/internity/   Internity structured output (nodes, prompts)
  agent/llm.py       Shared Gemini LLM singleton
  api/v1/            FastAPI endpoints and page routes
  templates/         Jinja2 HTML templates
  static/            CSS ("Warm Terminal" design system)
  cli.py             Typer CLI entry point
  main.py            FastAPI app with scheduler lifespan
docs/
  IMPLEMENTATION_GUIDE.md       Teams integration reference
  INTERNITY_INTEGRATION.md      Internity integration reference

Agent Architecture

Teams EOD (Narrative Report)

The LangGraph agent follows a generate-review-revise pattern:

fetch_activities -> generate_draft -> self_review -> [revise or finalize] -> END

• fetch_activities: Pulls activities from the database, groups by time period
• generate_draft: Uses few-shot examples and activities to generate a narrative EOD
• self_review: Structured output (ReviewResult) evaluates the draft against quality criteria
• revise_draft: If review fails, revises based on feedback (max 2 revision cycles)
• finalize: Copies the approved draft to the final output

Internity EOD (Structured Form)

Uses .with_structured_output() to generate InternityEOD directly:

• Tasks: Description + hours + minutes for each task
• Key Successes: Accomplishments for the day
• Main Challenges: Difficulties encountered
• Plans for Tomorrow: Next steps

Power Automate Setup

Since group chats don't support incoming webhooks, the app posts through a Power Automate flow:

1. Create a new flow with "When an HTTP request is received" trigger
2. Add "Post message in a chat or channel" action (Microsoft Teams connector)
   • Post as: User
   • Post in: Your group chat
   • Message: Use dynamic content (date and message fields)
3. Copy the trigger URL to .env as POWER_AUTOMATE_URL

License

Private project.