See what's in your data, without needing the internet.
Níyì turns spreadsheets into charts, checks and plain-language reports right on your device, without ever connecting to the internet. Use it online, or download it and turn your internet off. It works the same either way. Built for peacebuilders, NGOs and journalists.
Works offlineNo internet needed, even to open it.
Keeps data on your deviceNothing is uploaded, sent or saved.
Small on purposeRuns on older laptops. Best with a few related files.
1. Add your data
Drop in one or more CSV or Excel files. They can have different columns and will be merged. This works best when the files share a column, like a date or location.
Live log — exactly what Níyì is doing to your data, step by step, so nothing happens where you can't see it.
2. Your compiled model & report
What would you like to do with this data?
Pick one to start — you can switch any time.
Narrow it down
Only look at part of your data — one location, one date range, whatever you pick. Everything below updates to match.
What this is: an optional way to save your filter choices as a small file, so you — or a colleague —
can reload them later instead of setting filters by hand again. You don't need this unless you want it;
nothing else here depends on it.
Look at one column
Pick a column to see what's in it — how complete it is, the most common answers, anything unusual.
See how two things connect
Pick two columns to see how they relate — for example, "in County X, 7 out of 10 records are Type A." Simple counting, not a statistics score.
Sources
How complete is this?
How much is filled in, and whether anything looks like it was entered twice.
Possible mistakes
Things that look like typos or entry slips — different spellings of the same word, mixed date formats, repeated ID numbers. Worth a look, not necessarily wrong.
Activity over time
How many records happened in each period. The highlighted bars were unusually busy — worth a look, not necessarily a problem.
Click a bar to see the records behind it.
Every record, one by one
Every single dated record, laid out in order — helpful for spotting clusters the chart above might hide.
Click a dot to see that record.
Compare two periods
Pick two date ranges and see what changed — how many records, and which categories became more or less common.
Most common answers
For columns like location or type, the answers that come up most, and how often.
Do these numbers move together?
When one number goes up, does the other? Closer to +1 means yes, together. Closer to −1 means the opposite. Near 0 means no clear link.
Print makes a document you can read or hand off. Download saves the same numbers as a data file
(JSON — plain structured text any program can open) so you have a permanent record, since nothing here is
saved automatically once you close or refresh the tab.
About Níyì
What this is
A single HTML file that runs entirely inside your browser. Drop in CSV or Excel data from the field, and it
puts together a real analysis (trends, anything unusual, connections between numbers, data quality, a
timeline of events) into a readable report you can filter, print, save, or hand off. Nothing to install
beyond opening the file.
A few things worth knowing about how it handles multiple files and lets you narrow things down. Column
names that only differ by capitalization or spacing (like "Date" and "date") are combined automatically.
Anything more different than that is only ever suggested, never merged without you seeing it. The filter
panel in the report lets you narrow to a date range, a set of categories, or a number range, just point
and click, no spreadsheet formulas needed, and every chart and number updates to match.
How this was built
Níyì was built by the PeaceTech Alliance with no funding. It is meant to be free, independent and
answerable to nobody. It is free to use and share, and you will never be asked for money
to use it.
Taking it into the field
We recommend downloading Níyì and keeping a copy on a USB stick. It is a single file, so you can put it on
any computer before you travel, or when you are somewhere with no internet, and still analyse your data.
Nothing else needs installing.
It works in a browser tab, never connects to anything, and is small on purpose so it runs on older
laptops. It works best with a handful of related files, and large exports can slow it down.
Why it exists
Field practitioners — NGOs, conflict monitors, peacebuilders — often need to make sense of incident logs,
survey exports, or monitoring data, but the usual paths both carry a cost: sending data to a major AI model
means uploading it somewhere you may not trust and depending on outside infrastructure; traditional data
science means needing scikit-learn and Jupyter expertise most teams don't have on hand. Níyì exists to
remove both barriers — drop in data, get a real answer, entirely offline.
If you work in a sensitive area, handling data from vulnerable groups, uploading it to a big AI tool can put you
and the people you work with at risk. Níyì is here so you do not have to.
Who it's for, and how
Practitioner (peacebuilder, conflict monitor, protection): compile incident or monitoring logs from
multiple sites or enumerators into one dataset, catch data-entry inconsistencies — misspelled locations,
mixed date formats, duplicate case IDs — before they skew a report, spot unusual spikes with the timeline and
trend charts, and compare two reporting periods to see what's shifted. All without the data leaving your
machine.
Journalist: run a data-driven first pass on a dataset before writing about it — check it for internal
inconsistencies that can point to errors or manipulation, find outliers worth investigating further, and
compare two time periods around an event or policy change. If a source handed you the file, Níyì also
strips hidden Excel metadata (author, company) before you publish or pass it on, so you're not accidentally
exposing who it came from.
NGO worker: turn program or beneficiary data into something you can actually report on — data quality
checks before compiling a donor report, a codebook documenting every column for handoff to a new team member,
and period-over-period comparisons for standard M&E reporting — without uploading anything containing
personal data about the people you work with.
Built on what practitioners actually said
This was shaped by "PeaceTech in Practice"
(Holzinger & Coyle, 2026, Harvard Dataverse, doi.org/10.7910/DVN/7VAN5E — not a clickable link,
by design, since this page never connects to anything), a survey of 211 peacebuilding practitioners on how
they use technology and what they need from it. Offline access came up again and again:
"Internet access is a huge barrier in remote areas. We need tools that work offline or with very low bandwidth."— Project Manager, East Africa
"Many of our beneficiaries have no internet or very limited data. Offline functionality is critical for inclusion."— Program Coordinator, West Africa
"Training and support for digital tools often assumes stable internet, which isn't our reality. Offline resources would be a game-changer."— M&E Officer, South Asia
Based on qualitative responses from 211 practitioners via the PeaceTech Alliance network. As an English-language survey, this is not a globally representative sample. To find the source data, search for PeaceTech in Practice: Survey of Peacebuilding Practitioners on Technology Use, Risks, and Design Needs in Fragile and Conflict-Affected Settings on Harvard Dataverse.
How this works
What "model" means here. This isn't a black-box predictor guessing at answers, it's a statistical profile built directly from the numbers in your files: what's typical for each column, which columns tend to move together, and how activity changes over time. Every sentence in the report below is filled in from a fixed template using those real, computed values, not free-form AI writing. You can read it now, print it, or export the underlying profile as a JSON file to reuse later.
The analysis engine is plain code built into this file, running directly in your browser tab — no Python,
no external runtime, nothing fetched from anywhere while you use it. Merging your files, detecting column
types, computing statistics, and finding trends and correlations all happen with ordinary code your browser
already knows how to run. The narrative you read is built from templates filled in with those real, computed
numbers — not AI-generated writing. The live log above the drop zone shows each of these steps as they
happen, so you can see exactly what it's doing to your data.
Reading Excel files (.xlsx/.xls) uses one exception to "hand-written code": a well-known open-source library
called SheetJS that understands Excel's file format. Rather than fetching it from anywhere, its entire source
is bundled directly inside this HTML file, so it works the same as everything else here — no connection, ever.
CSV files don't need it at all; those are read with hand-written code you can read top to bottom.
Because there's nothing to load, there's no connection to plan around. This page makes no network request
at any point — not when it opens, not while you use it, not ever. You can confirm this yourself: open your
browser's developer tools, click the Network tab, then reload the page and build a report — you'll see
nothing go out.
What it can do
Multi-file merge with column matching — why it matters: combine incident logs or survey exports from
multiple sites or enumerators into one dataset, without hours of manually reconciling column names — and
without ever silently merging columns that don't actually match.
Data quality checks (missing data, duplicates) — why it matters: catches double-entered records and
incomplete fields before they quietly skew a report.
Data Consistency Checker — why it matters: flags spelling/casing variants of the same value, mixed
date formats, out-of-range coordinates, negative counts, and duplicate IDs — the kind of errors that come from
multiple people entering data differently.
Trend & spike detection — why it matters: flags periods of unusually high activity automatically,
no statistics background needed.
Event timeline — why it matters: shows every individual dated record on its own timeline, useful for
spotting clustering the bucketed trend chart can smooth over.
Category breakdown — why it matters: shows what's most common in a field like location or incident
type at a glance, without building a pivot table.
Correlation — why it matters: shows which numeric columns tend to move together, a first signal for
deeper questions worth asking.
Filters — why it matters: narrows the whole report to a date range, category, or number range,
point-and-click, no spreadsheet formulas needed.
Record drill-down — why it matters: every chart and finding can be clicked through to the actual rows
behind it, so nothing in the report is a black-box number.
Variable Explorer — why it matters: lets you inspect any single column's completeness, most common
values, and outliers on demand.
Change Detection — why it matters: compares two time periods directly — record counts, and which
categories became more or less common — useful for before/after or period-over-period reporting.
Codebook export — why it matters: documents every column (type, completeness, range, examples) as a
plain-text file, generated directly from the data, for handoff or donor reporting.
Excel support (including multi-sheet workbooks) — why it matters: reads .xlsx/.xls/.xlsb the same as
CSV, since most field data isn't already in CSV form.
Relationship Explorer — why it matters: shows how two category columns relate — e.g. "of records in
County X, 70% are Type A" — a well-defined breakdown, not a correlation guess, for the more-common case where
the columns you care about aren't numeric.
Analysis Recipes — why it matters: saves your current filters as a small file you can reuse or hand to
a colleague, so they can reproduce your exact slice of the data without you describing it in words.
What it's not
Níyì is built small on purpose, so it can run on an old laptop with no install and no connection. That
trade-off comes with real limits, worth knowing before you rely on it.
Not built for huge datasets. Everything runs in one browser tab, held in memory, with no server or
database behind it. A handful of files and a few thousand rows works well. Tens of thousands of rows, or a
large batch of big files, can get slow or unresponsive, especially on older hardware.
Not a statistics or machine learning tool. It computes counts, completeness, correlations, and trends
using plain arithmetic, nothing predictive and nothing modeled. If you need regression, significance testing,
or clustering, this isn't that tool.
Its checks are educated guesses, not certainties. Consistency checks and sensitive-pattern flags look
at column names and the shape of the data, not at what's actually true. They can miss real problems and
occasionally flag something harmless. Treat every flag as a reason to look closer, not a final answer.
Single-user, single-session. There's no collaboration, no saved history, and no version tracking.
Once you close the tab, the analysis is gone unless you printed or downloaded it first.
CSV and Excel only. No databases, PDFs, photos, or free-text documents, and the interface itself is
English only.
Privacy
Níyì keeps your data on your device, but it cannot protect the device itself. If your computer is already
compromised, anything on it is at risk whatever you use, so keep it updated and protected.
Your data is never uploaded, transmitted, or seen by anyone but you. Files you drop in are read straight off
your disk into the browser's own memory and processed there — never sent anywhere.
Nothing is kept, either. Everything lives only in this tab's memory for as long as it's open. Close the tab
or refresh, and both your files and the report are gone — nothing is written to disk automatically. Use
Print / Save as PDF or Download model (JSON) first if you want to keep a copy; otherwise, treat this as a
one-time analysis that disappears the moment you do.
Excel files carry a second, hidden layer of data most people never see: author name, company, manager, and
similar fields stored in the file itself, separate from anything in the actual spreadsheet cells. Níyì
checks every Excel file for this on load, tells you exactly what it found, and can hand you back a cleaned
copy with that layer removed — so a file can be fully anonymized in what it shows without still carrying who
made it underneath. The same check also flags whether a file contains macros (a well-known way spreadsheets
carry malware); Níyì never runs them, but a cleaned copy strips them out too. This has no equivalent for
CSV files, which don't have this hidden layer to begin with.
Once your data is loaded, Níyì also scans column names and value patterns for things that can be
sensitive on their own — GPS coordinates, phone numbers, email addresses, columns that look like they hold
passport or national ID numbers. This is a pattern check, not a guarantee: it can miss things, and it can
occasionally flag something harmless. Treat it as a second opinion, not a verdict.
"Clear everything now," in the header, wipes every file, filter, and report Níyì is currently holding in
this tab's memory. That's the honest limit of what any tool like this can promise, though — it clears this tab's own
memory, not anything your operating system might retain elsewhere. If you need real device-level protection
against seizure or forensic recovery, that's a job for full-disk encryption and your operating system, not for
a page running inside a browser tab.