Evidrilo

From a research task to a claim whose evidence and limits stay in view.

Evidrilo is a student research workspace. It connects a project brief, sources, evidence notes, findings, claims, limitations, and next steps so students can inspect the reasoning behind their work.

Students can organize their own local project or follow three separate bundled practice cases (the first Free, two additional cases with verified Pro). These cases are not project templates and do not fill in a student's research.

Version 1.0.0: Android APK and iOS Simulator build use hosted Staging and RevenueCat Test Store. The simulator package is not an iPhone IPA. Current acceptance and remaining work →

Features · Landing page · Landing guide · Quick start · Architecture · System flows · Workflows · Documentation · Contributing · Code of Conduct

Why Evidrilo

Academic work is spread across assignment instructions, source notes, analysis, and the final conclusion. Evidrilo keeps the path visible: what the task asks, which material supports a finding, how far a claim can go, what remains uncertain, and what to do next.

Students choose and assess their own sources and write their own conclusions. Evidrilo is not a source-discovery service, citation verifier, plagiarism checker, academic grader, or scientific-truth oracle.

Features

Student Project Workspace

Choose a project family from the catalog, then use an editable offline starter or begin blank. Work through focused sections for your brief, question, sources, evidence notes, findings, claims, limits, and next steps. Project data stays local; signing in does not upload or merge it.

Project workflow and lifecycle →

Guided Case Practice

Three bundled cases explore observations, source comparison, and reconsidering data. The first is Free; verified Pro opens the other two. Their synthetic material and deterministic feedback support practice separately from your own projects. They do not grade academic work or automatically populate a project.

Inside the Current App

Android 1.0.0 / build 2, captured 1 October 2026. These nine screens use synthetic demo data and Test Store plans; they illustrate the interface, not real payments or a completed research study. Select a screen to view the full image.

Your next step Home keeps your projects and learning routes within reach.	Choose a starting point Five illustrated families offer editable offline structures.	Keep work organized A focused list brings active and completed projects together.
Move through your project A section drawer makes the research path easier to navigate.	Record your sources Keep source details and evidence notes alongside your own work.	Follow the Practice trail Three separate cases make evidence reasoning concrete.
Start an investigation Short missions introduce the task before case-bounded feedback.	Ask for assistance Optional signed-in AI chat keeps the conversation focused.	Explore Evidrilo Pro Clear plan comparisons and RevenueCat Test Store checkout.

Project Files and Portability

Generate Markdown, PDF, DOCX, and structured CSV reports, or exchange a versioned .evproj archive with validation and preview before writing. Source and test coverage exists; device save/reopen and picker acceptance remain open.

Project and file workflows →

Project Lifecycle and History

Autosave and revision checkpoints help preserve a student's work while they develop it. A single project list shows active and completed work, with catalog-first creation and export in the project menu. Permanent deletion requires confirmation. Free supports five projects total per installation; verified Pro supports 50. Existing legacy archive/Trash records remain readable for compatibility.

Project lifecycle and history →

Optional Project Deadline

Set or clear a calendar date in Project Basics. It is saved with the local project, revision history, and .evproj export; setting a date does not schedule a notification.

Project workflow →

Project Guides and Reviewed Templates

Five offline family starters can create editable local project structures for lab experiments, observational or survey work, literature synthesis, qualitative field studies, and design/engineering work. They add prompts and sections only; they are not human-reviewed method templates or evaluators. The separate server catalog supports draft, review, approval, and publication, and currently has no reviewed student-selectable template.

Catalog workstream →

Bounded AI Assistance

Signed-in students can chat with AI or select a local project for contextual help. Messages may include recent conversation and the chosen project snapshot; chat history stays local and server activity is metadata-only. Proposed edits show a preview and require confirmation before being applied. General Chat was verified on Android through hosted Staging using Experiential Luna 6. Reviewed method-specific scaffolding still requires an approved, published template; AI does not replace deterministic case feedback or scientific review.

AI assistance boundary →

Accounts and Evidrilo Pro

Local project work does not require sign-in and is not uploaded when a user signs in. Google sign-in and Android session restoration have been observed. RevenueCat Test Store monthly/annual purchase, checkout cancellation/failure recovery, restore, relaunch, and sandbox expiry have evidence; real-store billing and iOS transaction acceptance remain open.

RevenueCat architecture →

API and Background Worker

The ASP.NET Core API, PostgreSQL migrations, and bounded .NET worker provide a hosted Staging foundation on Render with Supabase Auth/PostgreSQL. Authenticated credits, entitlements, progress, notifications, and sync returned successful responses; AI chat was verified from Android. Staging uses server-verified Test Store reconciliation; Production webhook billing is a separate release gate.

Platform architecture → · API docs →

Static Landing Page

Responsive vanilla HTML/CSS/JavaScript site with an animated project-workspace phone preview, a clearly gated Project AI story, an interactive synthetic-case walkthrough, and a capability FAQ. Open the live landing page → · Landing guide

The production site is hosted on Cloudflare Pages and connected to this GitHub repository. Changes under apps/landing/ on main are configured to trigger a deployment. Preview with python3 -m http.server 8000 --directory apps/landing.

Quick start

Requirements

• JDK 21 with java and javac;
• Android SDK for Android builds;
• .NET 10 SDK and Node.js for API, worker, and contract checks;
• Docker Compose for the local platform stack;
• macOS and Xcode for running the iOS host.

The Gradle Wrapper is included; a global Gradle installation is not required.

Mobile and shared tests

git clone https://github.com/AndroLay/evidrilo.git
cd evidrilo
cp local.properties.example local.properties
# Set sdk.dir in local.properties to the Android SDK location.

./gradlew :composeApp:jvmTest :composeApp:compileKotlinJvm
./gradlew :androidApp:assembleDebug

API, worker, and contracts

dotnet test platform/api.Tests/Evidrilo.Api.Tests.csproj
dotnet test platform/worker.Tests/Evidrilo.Worker.Tests.csproj --configuration Release
node --test contracts/contracts.test.mjs
node --test platform/database/migrations/migrations.test.mjs

Local platform stack

bash scripts/verification/check-deployment.sh .
docker compose -f infra/environments/local/docker-compose.yml up --build -d
curl --fail http://127.0.0.1:5080/health/live
docker compose -f infra/environments/local/docker-compose.yml down

The local PostgreSQL trust mode is development-only. Keep SDKs, caches, NuGet packages, generated output, and private media outside a source-only checkout. For iOS, open apps/ios/iosApp.xcodeproj in Xcode on macOS; Linux compilation does not prove simulator/device behavior. See the development guide for overrides and more checks.

Verification: what each check proves

Check	Proves	Does not prove
Common/JVM tests	Behavior covered by the tests that ran	Android/iOS rendering or device behavior
Android build	Android sources compile and package	Visual quality, accessibility, signing, or store upload
.NET, contract, database tests	API/worker/migration contracts in tested environments	Managed provider or production traffic
Architecture and package checks	Public boundary, dependency rules, and artifact protection	Human review or final publication
Emulator/simulator and accessibility review	Runtime on the tested targets	Untested devices and provider configurations
RevenueCat Test Store and staging	Behavior observed in those environments	Long-term revenue, retention, or scale
Human validation	Learner understanding of the flow and feedback	Guaranteed learning outcomes or competition results

Use focused checks during development and run the full local verification script for a release candidate:

bash scripts/ci/verify-local.sh

Tech stack

The mobile app is built with Kotlin Multiplatform and Compose Multiplatform; the optional platform API uses ASP.NET Core/.NET 10 with PostgreSQL.

Layer	Technology	Responsibility
Shared mobile	Kotlin Multiplatform 2.4.20	Domain, evaluator, state, persistence contracts, and network boundary
Shared UI	Compose Multiplatform 1.12.1	Android/iOS presentation and JVM walkthrough
Native hosts	Android compileSdk 37 / targetSdk 35; SwiftUI/Xcode	Lifecycle, secure storage, audio, HTTP, and packaging
Billing	RevenueCat KMP 3.10.1	Offerings, purchase/restore boundary, entitlements, and paywall integration
API	ASP.NET Core / .NET 10	Versioned REST, verified-auth policy, rate limits, and health checks
Data	PostgreSQL 16 / Npgsql 10.0.3	Content, accounts, projections, and audit persistence
Worker	.NET 10 hosted worker	Bounded background processing, retries, and projections
Local operations	Docker Compose and migration ledger	Reproducible local API/worker/PostgreSQL stack
Landing page	Vanilla HTML/CSS/JavaScript	Responsive static site and interactive synthetic-case walkthrough

Repository map

Path	Responsibility
apps/mobile-shared	Compose screens, coordinator, content/audio/billing adapters, shared tests
apps/android	Android host, manifest, and platform configuration
apps/ios	Xcode host, SwiftUI entry point, and iOS configuration
apps/landing	Static landing page and curated synthetic-case screenshots
modules/core	Cross-cutting shared primitives
modules/domain	Framework-neutral case/project models, evaluator, reducers, feedback
modules/application	Account/session, consent, sync, analytics, recommendation, AI orchestration
modules/data	Persistence contracts, codecs, recovery, storage adapters
modules/features	Feature presentation contracts and user-facing state
modules/design-system	Visual tokens, components, icons, fonts, reviewed assets
contracts	Versioned JSON schemas, route manifests, synthetic fixtures
platform/api	ASP.NET Core API, authentication, authorization, server operations
platform/worker	Bounded .NET background processing
platform/database	PostgreSQL migrations and local database integration
platform/integration	Local API/worker integration harness
infra	Docker Compose, Dockerfiles, deployment handoff boundaries
scripts	Verification, safety, packaging, asset, architecture checks
tests / tooling	Cross-boundary test assets and repository tools/configuration
docs	Product, workflow, architecture, development, testing, release guides
examples	Synthetic examples only
internal / audit	Owner-local research and evidence; private and excluded from public packages

Contracts and safety rules

• A session evaluates one identified CaseVersion; published versions are immutable.
• Feedback anchors active-case facts or rules; stale or ambiguous inputs reject or return CANNOT_ASSESS.
• The same case version and input produce the same result. AI never decides truth; M0 allows one revision.
• Verified provider claims establish API identity; client-supplied account IDs do not.
• Sync requires consent, bounded commands, idempotency, and a cursor; sign-in alone does not upload projects.
• Keep credentials, provider payloads, private research/reviewer data, and generated output out of the public repo.

See the product contract, architecture, and decisions for the detailed rules.

Public repository boundary

The public package contains source, contracts, synthetic fixtures, tests, scripts, and public documentation. It excludes private research/reviewer data, design files, credentials, database secrets, signing material, generated output, caches, and private media.

bash scripts/github/export-public-package.sh . /path/to/empty-candidate
bash scripts/verification/check-public-package.sh /path/to/empty-candidate
bash scripts/security/check-github-safety.sh .

A clean workspace does not prove that a public package, provider transaction, or production deployment has been verified.

Documentation, contribution, and license

• Product workflows and diagrams
• Architecture and repository structure
• System execution flows and diagrams
• RevenueCat architecture
• AI assistance architecture
• Development · Testing · Release
• Roadmap · Decisions · API docs

See CONTRIBUTING.md to contribute. For public review, publish this as a public open-source repository with a license; this source uses the MIT License, and third-party notices are in THIRD_PARTY_NOTICES.md.