Anaplan-inspired AI analyst for Filipino SMEs. Turns messy F&B sales data into decision-ready insights, structured as Headline / Findings / Concerns / Actions.
Filipino F&B SMEs — restaurants, bars, cafés, small retail — typically can't justify a full-time analyst or enterprise tools like Anaplan, but their decisions still benefit from real analysis. Sariling AI is a Streamlit-based AI analyst built for that gap: takes messy real-world business data (CSV, Excel, pasted tables) and produces structured analytical outputs aimed at owner-operators who don't have a finance team.
The output structure is opinionated. Every analysis returns:
- Headline — the single most important finding, in plain language
- Findings — supporting evidence broken out by metric
- Concerns — things the owner should be worried about
- Actions — concrete next steps, ranked by priority
This mirrors how a senior business analyst delivers a brief, not how an LLM rambles.
Four top-level modes with 18 specialized sub-modes:
- Finance — P&L analysis, margin investigation, break-even modeling
- Sales — period comparisons, hourly patterns, mix shifts
- Operations — labor optimization, inventory turnover, supplier review
- General Business — strategic overviews, what-if scenario modeling
Each mode is backed by a structured prompt designed for that analytical domain. Prompts encode Filipino business context: peso amounts, payday-cycle effects on sales, suki (regular customer) dynamics, fiesta and holiday seasonality, and Tagalog phrasing where natural.
A complete analysis walkthrough — from data input through structured output:
Single-file Streamlit entrypoint routes to seven pages:
app.py # entry + auth gate + sidebar router
├── app_pages/
│ ├── home.py # landing
│ ├── dashboard.py # activity stats + Plotly charts
│ ├── analyze.py # main flow: data → AI → charts → export
│ ├── compare.py # side-by-side period comparison
│ ├── scenarios.py # what-if scenario modeling
│ ├── history.py # browse / re-export past analyses
│ └── about.py # static info
├── utils/ # AI client, parsing, charts, exports, history, styling
├── prompts/ # mode-specific system prompts
├── data/ # SQLite history (gitignored)
└── .streamlit/ # theme config
The AI layer (utils/ai_client.py) is provider-swappable — wraps both Anthropic Claude and Groq behind a single query_ai() function, so the app can be pointed at either backend.
History is local SQLite (data/history.db, gitignored), so every analysis persists across sessions without leaving the user's machine.
- Python 3.10+
- Streamlit — UI
- Anthropic Claude — primary analyst LLM
- Groq — fallback / fast inference
- Plotly — auto-generated charts
- ReportLab — PDF export with embedded charts
- pandas + openpyxl — CSV/Excel parsing
- SQLite (stdlib) — local history persistence
git clone https://github.com/Jenkikan01/sariling-analyst.git
cd sariling-analyst
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env # then edit with your API keys
streamlit run app.pyYou'll need API keys from Anthropic and Groq, plus an APP_PASSWORD of your choice for the auth gate. See .env.example for the variable names.
- ✅ All four analyst modes and 18 sub-modes
- ✅ Compare and Scenarios features
- ✅ PDF / Markdown / Text export with embedded charts
- ✅ Filipino context (peso, payday cycles, fiesta seasonality, suki dynamics, Tagalog where natural)
- ✅ Deployed on Streamlit Community Cloud (currently access-restricted; demo on request)
- 🔲 Multi-user / role-based access
- 🔲 POS / accounting system integrations
- 🔲 White-label option
This is a working solo-developer project, deployed and being demoed to F&B SME prospects in the Philippines.
Proprietary. See LICENSE. Code is published for portfolio review and evaluation; commercial use requires written permission.
Joshua Jen Robiano Pujante — BSAIS student at Saint Paul School of Professional Studies, Tacloban City, Philippines.