Expert guidance for developing with the tinystruct Java framework. Use when working on the tinystruct codebase or any project built on tinystruct — including generating the bin/dispatcher and bin/dispatcher.cmd launcher scripts when a project lacks them, creating Application classes, @Action-mapped routes, unit tests, ActionRegistry, HTTP/CLI dual-mode handling, the built-in HTTP server, the event system, JSON with Builder/Builders, database persistence with AbstractData, POJO generation, Server-Sent Events (SSE), file uploads, and outbound HTTP networking.
85%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Passed
No findings from the security scan
Architecture and implementation patterns for building modules with the tinystruct Java framework – a lightweight, high-performance framework that treats CLI and HTTP as equal citizens, requiring no main() method and minimal configuration.
CLI and HTTP are equal citizens. A transport-neutral @Action uses Mode.DEFAULT (or omits mode) and has no HTTP-only dependencies. Such an action can run from both a terminal and a web browser. Mode.DEFAULT removes the transport restriction; it does not make an HTTP-dependent method CLI-compatible. Request and Response parameters come from HTTP context attributes and remain null when those attributes are absent, as in an ordinary CLI invocation. Keep shared logic independent of HTTP objects, sessions, and uploads, and call it from transport-specific entry points when needed. This "dual-mode" capability is the core design philosophy of tinystruct. An action with a transport-specific mode (Mode.CLI, Mode.HTTP_GET, Mode.HTTP_POST, ...) is deliberately bound to that transport: the framework rejects it from the other one with "The action is not allowed to be executed."
bin/dispatcherbin/dispatcher is the default, highest-priority tool for developing, running, testing, and debugging a tinystruct application — reach for it before an IDE run configuration, a hand-written main(), curl, or a browser. Because transport-neutral @Action methods are dual-mode by design, bin/dispatcher lets you exercise routing, argument binding, and business logic directly from the terminal, with the fastest possible feedback loop and no server/browser required. Actions bound to an HTTP method (Mode.HTTP_POST, Mode.HTTP_GET, ...) cannot be reached from bin/dispatcher; test those through an HTTP client against the running server.
# Run a CLI-compatible @Action directly to verify a new action
# (the class must be imported; see --import or default.import.applications in application.properties)
bin/dispatcher greet/James --import com.example.MyService
bin/dispatcher echo --words "Praise the Lord"
# Start the HTTP server when you need the web-facing counterpart
bin/dispatcher start --import org.tinystruct.system.HttpServer
# Import additional Application/MCP classes for the current run
# (one --import per class; a comma-separated list fails with ClassNotFoundException)
bin/dispatcher start --import org.tinystruct.system.HttpServer --import com.example.MyService
# Discover what's available
bin/dispatcher --help
bin/dispatcher --versionDefault to this workflow: implement a transport-neutral @Action, run it immediately via bin/dispatcher <action> to confirm it behaves correctly in CLI mode, and only start the HTTP server (also via bin/dispatcher start ...) once you need to verify the web-facing path (e.g. mode = Mode.HTTP_POST, sessions, file uploads). For HTTP-only actions skip the CLI step and test them over HTTP; for Mode.CLI actions no server is needed. Never hardcode a main(String[] args) as an app's entry point — bin/dispatcher (or bin/dispatcher.cmd on Windows) is the one entry point for every module.
bin/dispatcher and bin/dispatcher.cmdEvery tinystruct project needs its launcher in bin/. The framework generates it — never write it by hand, and never copy it from another project or substitute a main(). ApplicationManager.init() writes the launcher for the current operating system into bin/ under the working directory whenever it is missing, from templates inside the tinystruct jar, with that jar's version filled in. Line endings and the executable bit are already right.
1. Find the tinystruct version the project builds against: the <tinystruct.version> property or the org.tinystruct:tinystruct dependency in pom.xml. The jar must be reachable, in ~/.m2/repository/org/tinystruct/tinystruct/<version>/ (run mvn dependency:resolve if it is not) or in the project's lib/.
2. From the project root, run the framework once:
java -cp ~/.m2/repository/org/tinystruct/tinystruct/<version>/tinystruct-<version>.jar \
org.tinystruct.system.Dispatcher --versionReplace <version> with the version found in step 1. On Windows use %USERPROFILE%\.m2\repository\org\tinystruct\tinystruct\<version>\tinystruct-<version>.jar and the same class. This creates only the script for the OS you ran it on, and only if it is not already there:
| Run on | Creates |
|---|---|
| Linux, macOS | bin/dispatcher (executable, LF) |
| Windows | bin\dispatcher.cmd (CRLF) |
3. Verify. From the project root run bin/dispatcher --version (Windows: bin\dispatcher.cmd --version). It prints Dispatcher (cli) (built on tinystruct-<version>), and the VERSION inside the script equals the jar's version. Then bin/dispatcher --help lists the commands and every imported action. Show the user the file you created.
4. Get both scripts by running step 2 on each OS, or let a teammate or CI on the other OS run it and commit the result. Commit both, so a checkout on either OS works. Do not try to fake the other OS with -Dos.name=…: on Windows it fails (UnsupportedOperationException from setPosixFilePermissions) and leaves an empty bin/dispatcher. Recommended .gitattributes, so a checkout does not corrupt the line endings:
bin/dispatcher text eol=lf
bin/dispatcher.cmd text eol=crlfUpgrading tinystruct: change the version in pom.xml, delete the old script, and repeat step 2 with the new jar. Alternatively bin/dispatcher update checks Maven Central for the latest release, upgrades the project's dependency and regenerates the script with force; it needs the network, edits the pom.xml, and goes to the latest version, not a chosen one.
It also happens implicitly. Any code that reaches ApplicationManager.init() (the dispatcher itself, or ApplicationManager.install(app, config) in a unit test) creates the script in the current directory if it is missing. If unit tests scatter bin/dispatcher.cmd into module directories, run them from the build directory: in the surefire configuration set <workingDirectory>${project.build.directory}</workingDirectory>.
What the scripts do, which explains most problems:
.cmd takes the parent of bin/.target/classes, lib/*.jar, WEB-INF/lib/*, WEB-INF/classes and the tinystruct jar (from lib/, else ~/.m2), then they run org.tinystruct.system.Dispatcher with your arguments.mvnw / mvnw.cmd is missing, they extract the Maven Wrapper into the project root (unzip on Unix, PowerShell on Windows). Tell the user this happens; the files can be committed or ignored.-D… and -X… arguments as JVM options.Gotchas:
bin\dispatcher.cmd. The shell script joins the classpath with :, so under Git Bash/MSYS it fails with ClassNotFoundException: org.tinystruct.system.Dispatcher. Use WSL, or the .cmd.lib/: mvn dependency:copy-dependencies -DoutputDirectory=lib (and git-ignore lib/). In a multi-module build, keep the launcher in the directory that holds target/classes and lib/, and copy the sibling modules' jars into that lib/; running it from a module whose siblings are not on its classpath fails with NoClassDefFoundError.--import or configuration. One --import per class (--import a.A,b.B fails; write --import a.A --import b.B), or list them once in application.properties: default.import.applications=a.A;b.B (;-separated).JAVA_HOME must be set for the .cmd; it stops with an error otherwise.bin/ sits directly under the directory that holds target/classes and lib/. The .cmd treats the parent of bin/ as the root, so a bin/ nested in the wrong directory puts the wrong classpath in front of Dispatcher.Mode.CLI @Action via bin/dispatcher — this is the default way to work with a tinystruct app, before reaching for HTTP or an IDE run configuration. HTTP-only actions are tested over HTTP instead.bin/dispatcher / bin/dispatcher.cmd — have the framework generate it (see "Generating bin/dispatcher and bin/dispatcher.cmd").Application modules by extending AbstractApplication.@Action.Context.Builder and Builders components.AbstractData POJOs.generate command, with an XML mapping file or annotations (--mapping annotation).URLRequest and HTTPHandler.application.properties.The tinystruct framework treats any method annotated with @Action as a routable endpoint. Without a mode, dispatch is not restricted to one transport, but its parameters and implementation still need to support that environment. Ordinary CLI calls do not supply HTTP Request or Response objects. With mode = Mode.CLI or an HTTP mode such as Mode.HTTP_POST, dispatch is restricted to that transport. Applications are created by extending AbstractApplication, which provides core lifecycle hooks like init() and access to the request Context.
Routing is handled by the ActionRegistry, which automatically maps path segments to method arguments and injects dependencies. For data-only services, the native Builder and Builders components should be used for JSON serialization to maintain a zero-dependency footprint. The database layer uses AbstractData POJOs mapped to tables with @Table/@Column annotations or XML mapping files for CRUD operations without external ORM libraries. Missing tables can be created automatically with database.autocreate=true.
public class MyService extends AbstractApplication {
@Override
public void init() {
this.setTemplateRequired(false); // Disable .view lookup for data/API apps
}
@Override public String version() { return "1.0.0"; }
@Action("greet")
public String greet() {
return "Hello from tinystruct!";
}
// Path parameter: GET /?q=greet/James OR bin/dispatcher greet/James
@Action("greet")
public String greet(String name) {
return "Hello, " + name + "!";
}
}The limits below are example application policy measured in UTF-16 code units, not tinystruct defaults. Align them with your authentication service. Reject blank credentials without trimming or otherwise changing the values passed to authentication.
@Action(value = "login", mode = Mode.HTTP_POST)
public String doLogin(Request<?, ?> request) throws ApplicationException {
String username = request.getParameter("username");
String password = request.getParameter("password");
final int maxUsernameLength = 256;
final int maxPasswordLength = 1024;
if (username == null || password == null
|| username.length() > maxUsernameLength || password.length() > maxPasswordLength
|| username.isBlank() || password.isBlank()) {
throw new ApplicationException("Invalid credentials");
}
// AuthService is your application's own class, not part of tinystruct
String userId = AuthService.getInstance().authenticate(username, password); // null on failure
if (userId == null) {
throw new ApplicationException("Invalid credentials");
}
// Only touch the session after a successful authentication
request.getSession().setAttribute("userId", userId);
return "Logged in";
}import org.tinystruct.data.component.Builder;
import org.tinystruct.data.component.Builders;
@Action("api/data")
public String getData() throws ApplicationException {
Builders dataList = new Builders();
Builder item = new Builder();
item.put("id", 1);
item.put("name", "James");
dataList.add(item);
Builder response = new Builder();
response.put("status", "success");
response.put("data", dataList);
return response.toString(); // {"status":"success","data":[{"id":1,"name":"James"}]}
}import org.tinystruct.http.SSEPushManager;
@Action("sse/connect")
public String connect() {
return "{\"type\":\"connect\",\"message\":\"Connected to SSE\"}";
}
// Push to a specific client
String sessionId = getContext().getId();
Builder msg = new Builder();
msg.put("text", "Hello, user!");
SSEPushManager.getInstance().push(sessionId, msg);
// Broadcast to all
// Broadcast to all
SSEPushManager.getInstance().broadcast(msg);import org.tinystruct.data.FileEntity;
@Action(value = "upload", mode = Mode.HTTP_POST)
public String upload(Request<?, ?> request) throws ApplicationException {
List<FileEntity> files = request.getAttachments();
if (files != null) {
for (FileEntity file : files) {
System.out.println("Uploaded: " + file.getFilename());
}
}
return "Upload OK";
}DatabaseOperator provides direct access to connection-level catalog, schema, and metadata:
import org.tinystruct.data.DatabaseOperator;
DatabaseOperator operator = new DatabaseOperator();
try {
String catalog = operator.getCatalog();
String schema = operator.getSchema();
java.sql.DatabaseMetaData metaData = operator.getMetaData();
System.out.println("Using catalog: " + catalog + ", schema: " + schema);
} finally {
operator.close();
}tinystruct provides native support for the Model Context Protocol (MCP) starting with SDK version 1.7.0.
The MCP APIs (e.g., org.tinystruct.mcp.MCPTool, org.tinystruct.mcp.MCPServer, org.tinystruct.mcp.MCPException) are included directly in the core dependency:
<dependency>
<groupId>org.tinystruct</groupId>
<artifactId>tinystruct</artifactId>
<version>1.7.37</version>
</dependency>SECURITY WARNING (Prompt Injection): Tool return values are fed directly back into the AI model's context window. You MUST validate and sanitize all caller-supplied arguments before including them in the tool's return string. Failure to sanitize inputs can allow an attacker to inject adversarial instructions (Prompt Injection) that override the model's behavior. Always validate length, character sets, and nullity.
To create an MCP Tool:
org.tinystruct.mcp.MCPTool.@Action and declare parameters using @Argument within the arguments array.@Argument. (Do not use getContext().getAttribute(...) for tool arguments).import org.tinystruct.mcp.MCPTool;
import org.tinystruct.mcp.MCPException;
import org.tinystruct.system.annotation.Action;
import org.tinystruct.system.annotation.Argument;
public class MyCustomTool extends MCPTool {
public MyCustomTool() {
super("custom", "A custom tool for demonstrating MCP");
}
@Action(
value = "custom/hello",
description = "Say hello to someone",
arguments = {
@Argument(key = "name", description = "The name to greet", type = "string", optional = false)
}
)
public String hello(String name) throws MCPException {
// SECURITY: Validate/sanitize tool inputs before returning to the model
// to prevent prompt injection vulnerabilities.
if (name == null || name.length() > 50 || !name.matches("^[a-zA-Z0-9 ]+$")) {
throw new MCPException("Invalid name provided");
}
return "Hello, " + name + "!";
}
}To deploy an MCP Server:
org.tinystruct.mcp.MCPServer.init() and register your tools using this.registerTool(). The framework automatically scans and maps the @Action methods.import org.tinystruct.mcp.MCPServer;
public class MyMCPServer extends MCPServer {
@Override
public void init() {
super.init();
this.registerTool(new MyCustomTool());
}
@Override
public String version() {
return "1.0.0";
}
}Run the server via the dispatcher:
bin/dispatcher start --import org.tinystruct.system.HttpServer --import com.example.MyMCPServertinystruct supports overloaded methods for the same tool name in MCPTool (sharing the same tool name but accepting different parameter signatures).
Settings are managed in src/main/resources/application.properties.
# Database
driver=org.h2.Driver
database.url=jdbc:h2:~/mydb
database.user=sa
database.password=
# Optional: create missing tables from the class mapping on first use (off by default)
# database.autocreate=true
# Server
default.home.page=hello
server.port=8080
default.server.open_browser=true
# Static files are served only from these directories (comma- or semicolon-separated),
# Use real public directories without symlinked root components.
# Do not expose the working directory or its ancestors. Default: static
# server.static.dirs=static,public
# Locale
default.language=en_US
# Session (Redis for clustered environments)
# default.session.repository=org.tinystruct.http.RedisSessionRepository
# redis.host=127.0.0.1
# redis.port=6379
# Programmatic Logging Configuration
logging.enabled=true
logging.level=INFO
org.tinystruct.level=FINE
# Handlers to create (comma- or whitespace-separated). Default: ConsoleHandler only
handlers=java.util.logging.ConsoleHandler
# To also write rotating log files, append the FileHandler:
# handlers=java.util.logging.ConsoleHandler, java.util.logging.FileHandler
# java.util.logging.FileHandler.pattern=logs/tinystruct-%g.log
# java.util.logging.FileHandler.limit=10485760
# java.util.logging.FileHandler.count=5
# java.util.logging.FileHandler.level=INFO
# java.util.logging.FileHandler.formatter=org.tinystruct.system.logging.LogFormatter
# java.util.logging.FileHandler.encoding=UTF-8Access config values in your application:
String port = this.getConfiguration("server.port");The framework includes a programmatic wrapper around java.util.logging (JUL) that provides beautiful console output and advanced caller tracing out of the box.
StackWalker API to trace the call stack at runtime. It identifies and logs the exact class, method, filename, and line number of the caller that initiated the log (bypassing internal utility and logging layers).application.properties (e.g., org.tinystruct.level=FINE).logging.enabled=false turns all logging off. logging.level sets the root logger level and accepts OFF, SEVERE/ERROR, WARNING/WARN, INFO, CONFIG, FINE/DEBUG, FINER, FINEST/TRACE and ALL (default INFO).handlers property lists the handlers to create (comma- or whitespace-separated); it defaults to java.util.logging.ConsoleHandler. Each handler reads its own <handler>.level (default ALL), <handler>.formatter and <handler>.encoding. Any java.util.logging.Handler or Formatter class with a public no-argument constructor can be named; an unknown formatter falls back to the built-in LogFormatter, and a handler that cannot be created is reported on stderr while the others still start.java.util.logging.FileHandler.pattern is optional (default %h/java%u.log, in the user's home directory) and its parent directory is created if missing. A relative pattern such as logs/tinystruct-%g.log resolves against the directory the application is launched from (the project root when you run bin/dispatcher there). limit (default 0, unlimited; otherwise the maximum bytes per file), count (default 1, the number of files to keep) and append (default true) control rotation. %g in the pattern is the rotation number, so count=5 produces tinystruct-0.log through tinystruct-4.log. To enable it: add java.util.logging.FileHandler to handlers, set the properties above, and restart. ANSI colors are applied to console output only, never to log files.FINE shows nothing until the logger allows it too, for example logging.level=FINE or a package override such as org.tinystruct.level=FINE.application.properties when ApplicationManager initializes; restart the application after changing it.The built-in HttpServer serves static files from the roots named by server.static.dirs (comma- or semicolon-separated) or, failing that, server.static.dir; the default is static. In 1.7.37, the working-directory/ancestor check compares normalized path names before resolving symlinks. It rejects ordinary roots such as . or .., but a configured root that symlinks to the project directory or an ancestor can pass that check. Sibling directories outside the working directory are also allowed; a descendants-only policy is not enforced.
Use a real, dedicated public-assets directory such as static/, containing only files intended for public access. Check that the configured root and its parent path components do not symlink to a broader or private directory. Requests containing . or .. segments, hidden names such as .git or .env, or names ending in . or a space are rejected. Candidate files must remain within the configured root's resolved path. That containment does not prevent project-root exposure when the configured root itself resolves there: the handler serves a matching static file before the later isProjectRootFile check. Do not rely on a blanket project-root-file denial with symlinked configured roots. See the pinned v1.7.37 implementation.
| Symptom | Correct Pattern |
|---|---|
Importing com.google.gson or com.fasterxml.jackson | Use org.tinystruct.data.component.Builder / Builders. |
Using List<Builder> for JSON arrays | Use Builders to avoid generic type erasure issues. |
ApplicationRuntimeException: template not found | Call setTemplateRequired(false) in init() for API-only apps. |
Annotating private methods with @Action | Actions must be public to be registered by the framework. |
Hardcoding main(String[] args) in apps, or testing only via curl/browser/IDE run configs | Use bin/dispatcher as the entry point for all modules and the default dev/test tool for transport-neutral and CLI actions (HTTP-only actions are tested over HTTP against the server it starts). |
bin/dispatcher is missing, or a hand-written or copied launcher script | Let the framework generate it: run org.tinystruct.system.Dispatcher --version once from the project root (creates the current OS's script). |
bin/dispatcher fails with ClassNotFoundException: org.tinystruct.system.Dispatcher on Windows | The shell script cannot run under Git Bash/MSYS; use bin\dispatcher.cmd or WSL. |
ClassNotFoundException for --import a.A,b.B | One --import per class: --import a.A --import b.B. |
A shell-script launcher that fails with \r: command not found | The file has CRLF endings; convert bin/dispatcher to LF. |
Manual ActionRegistry registration | Prefer the @Action annotation for automatic discovery. |
| Action not found at runtime | Ensure class is imported via --import or listed in application.properties. |
| CLI arg not visible | Pass with --key value; access via getContext().getAttribute("--key"). |
| Two methods same path, wrong one fires | Set explicit mode (e.g., HTTP_GET vs HTTP_POST) to disambiguate. |
bin/dispatcher fails with "The action is not allowed to be executed." | The action has an HTTP-specific mode; it is not reachable from the CLI. Test it over HTTP. To support both transports, first separate HTTP-only dependencies from shared logic, then use Mode.DEFAULT for a CLI-compatible entry point. |
| A static file returns 404 though it exists | Check server.static.dirs / server.static.dir (default static/), forbidden request names, and resolved-root containment. Use a real public directory; symlinked configured roots can expose project or ancestor files before the project-root check. |
| No log file is written | FileHandler is opt-in: add java.util.logging.FileHandler to handlers. Set java.util.logging.FileHandler.pattern or the file lands in %h/java%u.log (your home directory). |
FINE / debug messages never appear | Both gates must allow them: set logging.level=FINE (or a package override like org.tinystruct.level=FINE) and make sure the handler's .level is not stricter. |
bin/dispatcher first: For transport-neutral and Mode.CLI actions, develop and verify against bin/dispatcher before anything else — run the action from the CLI, confirm behavior, then layer on HTTP/mode concerns. Actions with an HTTP mode (Mode.HTTP_POST, ...) are rejected from the CLI, so start the server and test them over HTTP instead. It's the fastest inner-loop and the one tool guaranteed to match how the framework actually routes and binds arguments.init(): Leverage init() for setup (config, DB) rather than the constructor. Do NOT call setAction() — use @Action annotation.Mode parameter in @Action to restrict sensitive operations to CLI only or specific HTTP methods.getContext().getAttribute("--flag") rather than adding parameters to the method signature.CompletableFuture.runAsync() inside the event handler.Detailed guides are available in the references/ directory:
src/main/java/org/tinystruct/AbstractApplication.java — Core base class with lifecycle hookssrc/main/java/org/tinystruct/system/annotation/Action.java — Annotation & Modessrc/main/java/org/tinystruct/application/ActionRegistry.java — Routing Enginesrc/main/java/org/tinystruct/data/component/Builder.java — JSON object serializersrc/main/java/org/tinystruct/data/component/Builders.java — JSON array serializersrc/main/java/org/tinystruct/data/component/AbstractData.java — Base POJO class with CRUDsrc/main/java/org/tinystruct/data/Mapping.java — Mapping metadata (annotations or XML), cached per classsrc/main/java/org/tinystruct/data/annotation/Table.java — @Table, with @Id and @Column alongside itsrc/main/java/org/tinystruct/data/tools/TableCreator.java — Creates missing tables (database.autocreate)src/main/java/org/tinystruct/data/tools/MySQLGenerator.java — POJO generator reference (MappingMode selects XML or annotations)src/main/java/org/tinystruct/data/component/FieldType.java — SQL-to-Java type mappingssrc/main/java/org/tinystruct/data/component/Condition.java — Fluent SQL query buildersrc/main/java/org/tinystruct/http/SSEPushManager.java — SSE connection managementsrc/main/java/org/tinystruct/system/logging/LogFormatter.java — Custom log formatter with ANSI console colors and StackWalker-based caller tracingsrc/main/java/org/tinystruct/system/logging/LoggerConfigurer.java — Programmatic logging configurator from application propertiessrc/test/java/org/tinystruct/application/ActionRegistryTest.java — Registry test examplessrc/test/java/org/tinystruct/system/HttpServerHttpModeTest.java — HTTP integration test patternsaffaan-m/ECC
3f14a10
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.