Ada turns spreadsheets into auditable BI dashboards

Ada is an open-source business intelligence tool built for a simple promise: upload a CSV, XLSX, or XLSM file and get a working analysis without setting up a traditional BI stack. The project, shared on GitHub by saineshnakra, generates an interactive Plotly dashboard, detects business schema, flags anomalies, produces a baseline forecast, and answers plain-English questions with the underlying calculation shown for each response.

Unlike many chat-with-CSV tools, Ada is designed to keep its reasoning visible. Its core analysis is deterministic and local, using pandas rather than hidden model output. The project separates its system into four layers: calculation, conversation, interpretation, and an optional AI layer. According to the repository, uploaded rows are never sent to a model prompt; when AI is enabled, the model only sees schema metadata or computed evidence.

The app supports:

Recommended reading

MacWhisper 14 adds a new editor and faster transcription

• Zero-configuration analytics for CSV and Excel files
• Local question answering through Ask ADA
• Anomaly detection over time-series data
• A guarded baseline forecast with visible backtested error
• Drill-down analysis by segment
• Downloadable Markdown briefs and cleaned CSV exports

Ada also tries to infer a business dataset automatically, looking for a primary metric such as revenue, sales, profit, cost, amount, or units, along with a date field and useful segment dimensions like product, category, channel, region, or customer.

Privacy model and optional OpenAI usage

The software runs fully without an API key. In that mode, no model calls are made, including for questions the rule-based parser can answer on its own. If a user adds a key, Ada can make two limited calls through the OpenAI Responses API:

• A query planner for questions the deterministic parser cannot handle
• A strategic read that summarizes computed evidence

The repository says both are typed, button-triggered, and cached, with storage disabled for requests and a hashed anonymous session identifier used for safety controls. The default model is listed as gpt-5.6-luna with low reasoning, while gpt-5.6-terra with medium reasoning is available for more ambiguous decisions.

How to run it

Ada is a standard Streamlit app. To run it locally:

• git clone https://github.com/saineshnakra/automated-data-analyst.git
• cd automated-data-analyst
• python -m venv .venv
• Activate the environment
• python -m pip install -r requirements.txt
• streamlit run app.py

No secret is required unless you want the optional AI features. The project is released under the MIT license, with tests, linting, and bytecode compilation run through GitHub Actions on every push and pull request.