Miqyas Beta
Log in

Changelog

A running record of what has changed in Miqyas.

Version 0.13.0

August 10, 2026
  • Ask your own AI assistant about your survey data. Connect Claude Desktop — or any assistant that speaks the Model Context Protocol — to Miqyas, and ask in plain language: “what are the top three countries where average tree circumference is highest?” The assistant works out which questions to use, runs the query through the same analytics engine that draws your charts, and answers with the same numbers. Every answer comes with a link back into Miqyas so you can see the result for yourself.
  • It sees exactly what you see, and it cannot write anything. A connection carries no permissions of its own — it presents your access, re-checked on every request, so it covers the same surveys under the same regional rules and stops working by itself the moment you leave the organization or lose access to analytics. It reads aggregated results only: never an individual response, never a free-text answer. It cannot create, edit, or delete anything, including dashboards — saving a chart stays something a person does in Miqyas.
  • Off until your organization turns it on, and off again in one click. Assistant access starts disabled for every organization; an owner or administrator switches it on, and switching it back off refuses every connection at once without revoking any of them or losing any of your settings. Surveys that collect health or other sensitive information stay out of reach entirely unless an owner or administrator allows each one individually — and if such a survey later changes what it collects, that permission lapses on its own and has to be given again.
  • You can see everything an assistant read. Each connection is named, shows when it was last used, and can be revoked instantly. An access record lists every request — who, when, which survey, which questions — alongside the decisions behind it, so “who opened a door” reads differently from “who walked through it”. The values you filtered by are deliberately never recorded, because a filter value can itself be someone’s answer.
  • The record has its own page, and it is built to be searched rather than scrolled. Look for a person, a connection, a survey, a question, or the action itself in your own language; narrow it to a single day; sort by any column; and page through it however long it grows. What you narrowed to lives in the address bar, so you can send the view to a colleague or come back to it later. Your own history sits on your profile in the same shape, showing only what concerns you.
  • Miqyas runs no AI model of its own. Your assistant runs on your side and asks us; we only answer. Nothing is sent to an AI company by us, and choosing which assistant to trust — and what it may do with what it receives — stays yours.

Version 0.12.0

August 5, 2026
  • Responses that don’t look like they came from a real participant are now marked for your attention. Miqyas watches for signs that a response wasn’t filled in by a person working through the form at a human pace, and labels the ones it finds doubtful as “suspicious” — nothing is discarded and nothing is corrected on your behalf. A badge appears next to the response in the submissions list, a “Suspicious only” filter lets you review them as one batch, the submissions summary tells you how many there are, and each response’s detail page explains in plain language what looked off about it.
  • It stays advisory, and it costs your participants nothing. A suspicious label never changes a response’s quality score, its review state, or whether it gets exported — the judgement stays entirely yours. It also works quietly in the background: the survey-filling experience is unchanged, a genuine participant is never blocked or asked to prove anything, and nothing new about them — no personal details, no device fingerprint, no tracking — is collected to make it work.
  • Narrow any chart to just the responses you care about. The chart editor now has a filter builder: add conditions on any answer, or on a response’s own details (review state, country, the collector who gathered it), and combine them into “All conditions” / “Any condition” groups that nest — so “women, living in either Aleppo or Idlib” is one chart instead of three. Text answers can be matched by what they contain, start with, or end with; dates by a range or a before/after comparison; any question at all by whether it was answered. Each chart prints its own conditions on the card, so anyone reading the dashboard can see what a chart leaves out without opening the editor.
  • Ask about the same person, not just the same household. When a survey repeats a section — one entry per child, per well, per visit — a filter can now be scoped to that repeated section, and every condition inside it is checked against one entry at a time. “Any child who is under 5 and female” now means exactly that, where two ordinary conditions would also have matched a household with a young boy and an older girl who are different children. You choose whether any, every, or no entry has to match, and the editor tells you up front how a household with no children at all is treated — instead of leaving you to discover it in the numbers.
  • Hide the groups that are too small to read, and sort the ones that remain. A chart can leave out groups with fewer than a number of responses you choose, or groups that don’t meet a threshold on the value being measured (“districts averaging above 500”), and it always says how many it left out and why. You can also sort a chart by any of its groupings or measurements, several keys deep, and an unchanged chart always comes back in the same order. Leaving out small groups is a readability tool: it applies to that one chart, and never changes what an export contains.
  • The question picker now tells you what each question actually is. It shows the hint text you wrote in the form itself, so two similarly named questions — “Received vaccine” and “Vaccine received (specify)” — are no longer guesswork, and you don’t have to reopen the form design in another tab to tell them apart.
  • “Never asked” and “asked but left blank” are finally two different things. When a question is only shown to some respondents — asked only of households that keep livestock, say — Miqyas can now tell you how many people were never shown it at all, separately from how many saw it and skipped it. Where the skip rule is more involved than it can read with confidence, it says nothing rather than guessing at a split it can’t stand behind.
  • Chart a rating scale in its real order, and get a median for it. Mark a question as an ordered scale — Miqyas suggests it for you when the answers look like one, so it’s usually a single click rather than a form to fill in — and its answers stop coming back in whatever order the database happened to return. “Strongly disagree” through “Strongly agree” charts in that order, on every chart built from it, and a new Median measure becomes available that reports the middle answer by name, not as a number.
  • Know a question is a poor fit for a chart before you build it. The picker now shows roughly how many different answers a question actually received and how often it went unanswered, drawn from your real responses. Pick a free-text question as the grouping for a pie chart and you’re warned first, instead of discovering it as three hundred one-off slices. A survey with no responses yet simply shows nothing here — it isn’t penalised for being new.

Version 0.11.0

July 19, 2026
  • See where a survey’s data came from, on a map. A new map chart plots every matching response’s location — a point, a path, or an area — on an interactive map, so you can spot coverage gaps or suspicious clusters at a glance. Points group together at low zoom so even a large survey stays readable; you can color locations by another answer (a category with a legend, or a number as a gradient), switch to a heatmap for dense point sets, and click any location to open that response. Like every chart, the map follows the dashboard’s shared date range and filters.
  • Group and map your data by administrative region. Miqyas can now work out which area each response was collected in — a whole country, or one of Syria’s 14 governorates — and aggregate by it: shade a region map by value (with a clear legend and a distinct shade for areas with no data), or group any bar, pie, or line chart by region. It works with any measurement, not just a count of responses (say, the average household size per governorate), and always shows an explicit “outside / no location” group for responses that fall in no area. Region names appear in your language and switch instantly when you change the dashboard’s display language.

Version 0.10.1

July 17, 2026
  • Give each public share link its own accent color. In a link’s branding step you can now pick a color, and the whole survey page a respondent sees — buttons, the progress bar, the answer selection controls, sliders and rating stars — is recolored to match, so a link shared under a partner’s brand looks the part from end to end.
  • Your custom thank-you message is now shown. The message you set on a link appears on the confirmation screen after someone submits, instead of the generic default.
  • The survey overview now shows real recent activity — when the survey was created, each revision you published, and each public link you created, each with its real time — in place of the earlier placeholder, and the header shows the true created and last-edited dates.
  • Countries read naturally in the submissions list — a response’s country now shows as its full name in your language (e.g. “Germany” / “ألمانيا”) rather than a two-letter code, and the technical “source” column is hidden by default (you can bring it back from the column menu).

Version 0.10.0

July 16, 2026
  • Collect responses in the field with ODK Collect. Alongside shareable web links, a study can now gather data through ODK Collect — the offline-first mobile app — so a field team keeps working where there’s no signal and syncs when the connection returns. A platform administrator creates an account for each field collector; the collector signs in to their own portal, accepts the surveys they’ve been invited to, and connects their phone by scanning a single QR code — no server addresses or passwords to type. On the device they see exactly the surveys they’ve been engaged on, each under its Miqyas name.
  • Every response is attributed and located. A response gathered through ODK Collect records who collected it — shown as “collected by” on the submission and now filterable in the submissions list — and is stamped with that collector’s assigned country at the moment it’s submitted, so cross-border collection stays clear even if the collector’s details change later. Fill duration is now captured reliably for every response, from the web or from a phone, so the duration-plausibility quality check works the same across both — and the timing fields stay hidden metadata, never shown among the respondent’s own answers.
  • Set a target and a fee for each collector, and watch progress in real time. When engaging a collector on a survey you can set how many responses you need, a fee per approved response in the currency of your choice, and an optional deadline. Both you and the collector then see live progress — approved responses against the target, fees earned so far, total expected on completion, and an overdue flag — counting only responses you’ve approved, so payment always tracks accepted work. Each survey also has its own Collectors tab with a per-currency total-fees summary.
  • Your exports stay tied to work you’ve approved — and every row is traceable. A survey’s data export, whether a spreadsheet or the attachments, now includes only the responses you’ve approved, so unreviewed or rejected data never leaves the platform and a collector’s fees always match what’s actually exported. Each exported row also gains a source identifier that traces it back to the field collector who submitted it or the exact share link it came through.

Version 0.9.2

July 13, 2026
  • When a share link is set to accept one response per person, that limit is now actually enforced. After someone submits through the link, that browser is gently stopped from sending a second response — reopening the link shows a friendly “you’ve already responded” note, and the confirmation screen no longer offers to submit again. It stays deliberately best-effort and private: the check lives entirely in the respondent’s own browser, so nothing about who they are is ever stored or sent to us.

Version 0.9.1

July 13, 2026
  • Your progress is now saved as you fill out a survey. If your connection drops, your phone locks, or you close the page by accident, your answers — including repeated sections and uploaded photos — are kept safely on your own device. Reopen the same link and Miqyas offers to continue exactly where you left off (or to delete the saved answers and start over), and a small indicator shows at a glance whether your latest changes are saved. The saved draft never leaves your device, is removed the moment you submit, and expires on its own after 7 days.

Version 0.9.0

July 12, 2026
  • Registries — reusable lists of records (households, trees, patients, clinics, …) shared across a study’s surveys. Miqyas now recognizes automatically when a form builds up such a list or looks records up from one: a new Registries tab on each study shows every registry, the forms that feed and use it, and how many records it holds, and the live list reaches respondents as up-to-date choices — so those forms no longer look like they’re missing a file upload.
  • Browse and manage a registry’s records without leaving Miqyas — search and sort the table, and open any record to see its current values (locations on a map, attached files) and its full history: who, or which submission, changed what and when. Add or edit records by hand, import a batch from a spreadsheet with a preview before anything is saved, and delete or restore records.
  • Survey answers that point at a record are now clickable links to it, and each record shows how many submissions reference it — so you can trace a value back and forth between a response and the shared list it came from.

Version 0.8.0

July 9, 2026
  • Location answers are now readable. A GPS point, trace, or area answer shows a compact control — copy the raw coordinates, download it as GeoJSON, or open it on an interactive map with point count, length, and enclosed area — instead of an unreadable string of numbers.
  • Maps — the location picker while answering a survey, and the map view of a submitted answer — are served by MapTiler, an EU/Switzerland-hosted map provider under a data processing agreement, now disclosed in our Privacy Policy.

Version 0.7.2

July 7, 2026
  • A more realistic estimated fill time for surveys — it now accounts for question types and skip logic (counting only the questions a respondent actually reaches, not every branch of the form) and refines toward the real median time as responses arrive, instead of a flat minute-per-question guess. It also appears on the survey details, and display-only note text no longer counts as a question anywhere.
  • The response-quality duration check now shares that same baseline, so a respondent who legitimately takes a short path through the form is no longer flagged as suspiciously fast.

Version 0.7.1

July 6, 2026
  • An optional participant-count tracking target for surveys, set during creation — see progress toward it right in the submissions summary and the study’s survey list.

Version 0.7.0

July 5, 2026
  • Dashboards now live on the server and are shared across your organization — build several named dashboards and open them from any device or teammate, instead of one that lived only in your browser.
  • Two new chart types — scatter plots and heatmaps for comparing two categories at once — plus grouping a measurement by category.
  • A more accurate progress indicator while filling a survey — it counts the questions you’ve actually answered against the ones that currently apply (updating as follow-up questions appear), shows each section’s own completion, and now appears on surveys with no sections too.

Version 0.6.0

July 3, 2026
  • Persisted XLSForm source with a safe draft-revision workflow — upload a new form version without touching the live survey.
  • A full analytics dashboard: configurable, drag-and-drop charts (pie, bar, line, scatter, heatmap, or a single number), time series, cross-tabs, shared and per-chart date-range filters, CSV and image export.
  • Attachment storage and retrieval for survey media.
  • Submission export with filters and multi-select.

Version 0.5.0

June 28, 2026
  • Automated submission quality scoring — completeness, fill-duration plausibility, and straight-lining detection — with configurable weights and a live preview.
  • Shorter, cleaner links for studies and surveys.
  • Collapsible survey settings with clearer summaries.

Version 0.4.0

June 23, 2026
  • The submission detail page now remembers how you like to view answers — by question wording or by raw field name.
  • Pinned and recently-visited surveys in the sidebar.
  • The survey-filling experience translated into Arabic.
  • Clearer limits and handling for oversized attachment uploads.
  • A redesigned control for marking a submission’s review status.

Version 0.3.0

June 20, 2026
  • Clearer answer details on the submission page, plus the ability to leave comments.
  • Every submission now records which language it was filled in, along with its consent details.

Version 0.2.0

June 19, 2026
  • A light/dark theme toggle on the survey-filling experience.
  • Improved file and photo upload controls on public surveys.