TargetX

Every other KTU calculator tells you what you already scored. TargetX tells you the mark you still need.

A desktop app for the KTU 2024 scheme. It reads your marks and attendance, applies Regulations 2024 as written, and answers the question that actually matters before an exam: what do I have to score in this paper?

Download TargetX — Windows, macOS, Linux

Free for students, forever. Your record is a file on your own machine, and every figure on every screen is computed there. No account is needed to use any of it — sign-in exists only for the question box, and nothing else changes if you never do.

---

---

The thing no other tool shows you

Under Regulations 2024, attendance is not only an eligibility gate — it is a scored component of your internal marks, worth up to 5 of them in every subject, in bands (R 7.5.ii):

Attendance	CIE marks earned
85% and above	5
80–84%	4
75–79%	3
70–74%	2
60–69%	1
below 60%	0

So a student sitting comfortably above the 75% line is still forfeiting marks, in every subject, quietly, all semester. No portal shows this. No other calculator computes it. TargetX totals it across your subjects and tells you the cheapest way to win a mark back — usually a specific number of consecutive classes in one specific subject.

And below 60% the door shuts: R 6.2 lets the Principal condone attendance under 75% only as far as 60%. Below that the exam cannot be sat, and attending from there on does not change it. TargetX says so on the day it becomes true, not after results.

What else it does

• The mark you need, per subject. Both pass conditions, not one — the separate 40% ESE minimum as well as the combined total. Spreadsheets miss the first, which is how a subject that "looked passed" comes back as a supplementary. A * beside a required mark means the ESE cutoff is what is binding, not the aggregate.
• How many classes you can still miss. From real attended-over-held counts, with duty leave credited the way R 6.3.ii actually caps it — at 10% of classes held, which grows as you spend it. Answered as a number of classes.
• The cheapest route to your CGPA. Name a target; TargetX works backwards to the SGPA each remaining semester must carry, then to which subject is cheapest to push and how far — balanced so the plan is one a person can attempt, rather than four subjects at 95.
• Your marks in without typing them. Sign in to your college's etlab portal and it pulls every semester, attendance, series marks, grades and published SGPA in one pass.
• Ask it in words. Press Ctrl K and ask — "can I skip tomorrow?", "what do I need to pass CN?", "what is condonation?". Tex answers in the box rather than dropping you on a screen to work it out. Every figure in the reply is computed locally by the engine, so the answers are the same numbers the rest of the app shows, and most questions never touch the network at all.
• Updates arrive on their own. The app offers a new version a few seconds after launch. It never installs unasked, and a failed check is silent — being offline is not an error a student needs to see.

Two SGPA figures are shown on purpose. Confirmed counts only subjects with a real ESE mark. Projected assumes you hit your target where it is reachable, and the best grade still mathematically available where it is not.

Getting your data in

Four routes, strongest first. All four converge on one tolerant parser, so it is the same tested code path whichever you use.

1. Sync etlab — live login. Enter the portal URL, username and password. The client discovers rather than assumes: it probes candidate login routes, picks the form scoring highest, harvests every hidden input (so Yii1 YII_CSRF_TOKEN and Yii2 _csrf both work), and only accepts a page if course codes actually appear in it.
2. KTU grade card — past semesters into a real CGPA. KTU's own portal is captcha- and OTP-gated, so TargetX does not script a login there. It parses the card you already download, then recomputes each SGPA and flags any semester where its figure disagrees with the printed one — a mismatch means a row was misread, and you get told instead of quietly getting a wrong CGPA.
3. Public page URL — for college pages with no login.
4. Paste a table — the universal fallback. Works when a portal has a captcha, a VPN, or markup nobody predicted.

Will it work at your college? Sync has been proved against one college's etlab so far. You can check yours without giving anyone an account: save your academics page from a browser you are already signed in to, and drop the file onto Data -> "Will sync work at my college?". The page never leaves your machine; what the app offers you to send back is a separate block of table shapes with every digit blanked out.

Privacy

Your marks, attendance and CGPA never leave the machine. The app makes network requests in exactly four situations: the portal sign-in you asked for, the course-catalogue file, the update check, and — only if you sign in and only when you ask a question the app could not answer by itself — the question box. All four are named and located in the privacy statement (PRIVACY.md).

The fourth one is new in 0.3.0 and is the only one that involves a server of ours, so it is worth being exact about. What is sent is the question you typed and the code and title of the subjects you are registered for. What is not sent is every number: no marks, no attendance, no CGPA, no name, no register number. What comes back is not an answer — it is a destination, drawn from a fixed list of five screens and your own subjects, and it is rejected outright if it names anything else. Every figure you then read was computed on your machine by the same engine as the rest of the app. Sign-in gates that box and nothing else; signed out, the app is fully usable and the box still answers everything it can answer locally, which is most things.

Your portal password is used once, to sign in. It is never written to disk, never logged, never included in an export. That is enforced by a test, not by intention: credential-containment.test.ts puts a sentinel password through a whole sync and searches for it by value in every outbound request, the saved record, the export, and the text of a failed sync's error message.

Licence

Source-available under the Business Source License 1.1 (LICENSE).

• Students: free, forever. One person tracking their own coursework pays nothing and needs no permission. That grant is in the licence text, not offered as a policy that could be withdrawn.
• Institutions: commercial licence required. Any deployment by or on behalf of a college needs a separate agreement.
• It becomes Apache-2.0 on 2030-08-29. Fixed, and in the licence text.

The source is public because the central claim — that your marks never leave your machine — is one you should be able to check rather than take on trust. There are four places to check and no others: app/src-tauri/src/etlab.rs and ktu.rs carry the portal transports, app/src/sync/update.ts the update check, and app/src/state/ask.ts the question box — which is the only one that talks to a server of ours, and the only file that decides what is put in that request. The Worker on the other end is in worker/, and worker/src/schema.ts is the type its reply has to parse into before the app will act on it. That argument cannot be made from a closed repository, which is why this is source-available rather than proprietary, and why it is not permissive. Not open source under the OSI definition, and not described as such.

Third-party components ship under their own terms (THIRD-PARTY-NOTICES.md). No external pull request will be merged until a contributor licence agreement exists — merging one without it would make the project jointly owned and end the relicensing option silently.

---

---

For developers

Everything above is the product. Everything below is the repository.

Building from source

cd app
npm install
npm run tauri dev      # desktop window
npm run dev            # or just the frontend in a browser
npm run tauri build    # installers into src-tauri/target/release/bundle

The retired Python original is kept in legacy/ as the parity oracle for the calculation core. It is not part of the shipped app.

Tests

cd app && npm test        # 794
cd worker && npm test     # 75

869 tests across 71 files, counted from a run on 2026-09-02, not estimated: 257 engine, 237 UI, 180 state, 102 sync, 16 styles, 2 data, and 75 in the Worker.

engine/__tests__/parity.test.ts is the load-bearing one. It replays a frozen corpus of generated course cases and semester rollups produced by the Python original, and the corpus file has been byte-identical through eighteen refactors — so a change that alters a single computed figure anywhere in the engine fails immediately, by name. CI re-checks that the corpus itself has not been edited, because regenerating it to make a test pass is the one way this suite could be defeated without anything going red.

Other things that are checked rather than claimed: colour contrast (styles/__tests__/contrast.test.ts), keyboard reachability (npm run keyboard), layout overflow at real laptop sizes (npm run measure), and the download page at four widths in both themes (npm run site).

The 2024 rules, as implemented

Grade bands (total = CIE + ESE, out of 100):

Grade	Range	GP		Grade	Range	GP
S	90–100	10		C+	65–69	7.0
A+	85–89	9.0		C	60–64	6.5
A	80–84	8.5		D	55–59	6.0
B+	75–79	8.0		P	50–54	5.5
B	70–74	7.5		F	<50 or ESE below cutoff	0

Two independent pass conditions — both must hold:

1. CIE + ESE >= 50
2. ESE >= 40% of the ESE maximum -> 24/60 or 20/50 or 10/25

Course patterns (per subject, switchable in the grid):

Type	CIE	ESE	ESE cutoff
TH 40/60	40	60	24
TH 50/50	50	50	20
LAB 75/25	75	25	10
PRJ 100/0	100	—	n/a

SGPA = sum(credit x gradepoint) / sum(credit) · CGPA = credit-weighted across locked semesters · Percentage = 10 x CGPA.

Attendance: below 75% flags SHORTAGE; below 60% — the R 6.2 condonation floor — flags DEBARRED.

Why this exists — what the existing tools get wrong

Prior art	Failure
Entri / typical web KTU calculators	Still on the 2019 model. Some apply the legacy (10 x CGPA) - 2.5 conversion; the 2024 scheme is % = 10 x CGPA, full stop.
Old GitHub CGPA repos (2014 / 2019 schemes)	Carry the retired Low Pass (LP) grade and its grade point. Under 2024 there is no LP band.
Mobile grade apps	Ask only for the final letter grade per subject. They cannot see a semester in progress, which is the only time a prediction is worth anything.
Spreadsheets	Model the aggregate CIE + ESE >= 50 but silently ignore the separate 40% ESE minimum — the exact case where a student with 38/40 internal is told "you only need 12" and then fails on the cutoff.
All of them	No attendance-to-marks model, no target engine, no persistence.

Course catalogue

Presets load from curriculum.json — data, not code. KTU revises codes between admission batches, so correct the file once for your batch and it stays corrected. The app refreshes it from this repository on launch, which means a curriculum revision does not need a new build.

["PCCST302", "Data Structures and Algorithms", 4, "TH 40/60"]

KTU sets the first year by group, not by branch — page 1 of the 2024 curriculum sorts every B.Tech branch into four groups and prints one first-year table per group — so the catalogue is keyed the same way. 31 branches across Group A (computer and information science) and Group B (electrical science) carry S1 and S2; Computer Science and Engineering is transcribed in full and carries S1 to S8. Everything after the first year is per-branch and arrives one PDF at a time, which the picker says on screen rather than letting a student seed another branch's subjects at another branch's credits.

S1 and S2 carry KTU's first-year slots — Physics or Chemistry, Health and Wellness or Life Skills, plus a third in Group B S2 — as a choice rather than as fixed subjects, because a student takes one of each in S1 and the other in S2 in an order the institution sets. Groups C (physical science) and D (life science) are not transcribed.

Layout

Path	Role
app/src/engine/	Calculation core (evaluate, requiredEse, summarise, goals, targets)
app/src/sync/etlab.ts	Authenticated etlab client, over a Rust transport
app/src/sync/gradecard.ts	Grade-card and public-page parsers
app/src-tauri/src/etlab.rs	Cookie jar and HTTP. In memory only, never a file
app/src/data/curriculum.json	Bundled course catalogue, refreshable from the repo
app/src/state/persist.ts	Your data, saved atomically with three backups kept
app/src/sync/update.ts	The update check. Never blocks startup, never installs unasked
app/src/state/answers.ts	Tex's answers. Every figure computed here, from the engine
app/src/state/glossary.ts	What the words mean. A lookup table, not a model
worker/	The question router. Names a screen and a subject; never a number
docs/	The download page, served by GitHub Pages
legacy/	The retired Python original, kept as the parity oracle

Built on Tauri 2 with a SolidJS frontend and a TypeScript engine — a single binary with no Python at runtime. PORT_PLAN.md has the full reasoning, including the alternatives that were rejected and why.

Other documents

CHANGELOG.md is written for the student deciding whether an update is worth taking; its first section is the numbers that used to be wrong. DEPLOYMENT.md answers "what does this touch?" for a college IT department — silent-install flags, the five hosts it contacts, where the record lives, and how to take version control away from the updater. SIGNING.md covers releasing and the two signatures a build carries. ACCESSIBILITY.md is a dated self-assessment against WCAG 2.1 AA, including what is still short.