eldermoraes/quarkus-agentic-scaffolding / skills/scaffold-project
scaffold-project
VerifiedScaffold Quarkus + LangChain4j projects end-to-end and add agentic components to existing ones — AI services, tools, agents and multi-agent workflows, RAG pipelines, MCP clients and servers (Model Context Protocol), guardrails, and embedding store setups. Use this whenever the user wants a new Quarkus + LangChain4j project or module created, or an agentic component added to an existing project in this stack, including a baseline pom.xml, application.properties, project layout, or starter classes.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 20.9 KB |
| agents/openai.yaml | config | 669 B |
| templates/Agent.java.template | other | 28.4 KB |
| templates/AiService.java.template | other | 6.1 KB |
| templates/AiServiceTest.java.template | other | 6.2 KB |
| templates/ApplicationIT.java.template | other | 2.7 KB |
| templates/Guardrails.java.template | other | 4.5 KB |
| templates/McpClient.java.template | other | 2.5 KB |
| templates/McpServer.java.template | other | 2.6 KB |
| templates/RagSetup.java.template | other | 9.1 KB |
| templates/RestResource.java.template | other | 3.1 KB |
| templates/Tools.java.template | other | 2.3 KB |
| templates/application.properties.template | other | 12.6 KB |
| templates/pom.xml.template | other | 7.6 KB |
templates/application.properties.template
View on GitHub# ============================================================================
# Quarkus + LangChain4j configuration
#
# Provider: Ollama, local-first. Requires a running Ollama (`ollama serve`,
# default endpoint http://localhost:11434) and the models pulled locally:
# ollama pull qwen3:4b
# ollama pull qwen3:1.7b
# qwen3 is what the official quarkus-langchain4j Ollama guide uses, and its
# tool-calling is markedly better than llama3.2's — which is what the tool and
# agent templates here exercise. Models with >= 3B parameters follow system
# prompts more reliably, so qwen3:4b backs the default and "main" models and
# qwen3:1.7b backs the "smaller" named model below.
# ============================================================================
# --- Ollama endpoint (uncomment to override the default) ---
# quarkus.langchain4j.ollama.base-url=http://localhost:11434
# --- Default (unnamed) chat model: backs only code that names no model. ---
quarkus.langchain4j.ollama.chat-model.model-id=qwen3:4b
quarkus.langchain4j.ollama.chat-model.temperature=0.7
quarkus.langchain4j.timeout=120s
# --- Named "main" model: the primary task. Every @RegisterAiService declares a ---
# modelName (project conventions section 4); the primary services use
# @RegisterAiService(modelName = "main").
quarkus.langchain4j.ollama.main.chat-model.model-id=qwen3:4b
quarkus.langchain4j.ollama.main.chat-model.temperature=0.7
quarkus.langchain4j.ollama.main.timeout=120s
# --- Named "smaller" model: cheap/fast subtasks (classification, query rewriting). ---
# Select it with @RegisterAiService(modelName = "smaller") or @ModelName("smaller");
# an @Agent method needs @ModelName on the method (see Agent.java.template).
# Use temperature 0 for deterministic, structured results.
quarkus.langchain4j.ollama.smaller.chat-model.model-id=qwen3:1.7b
quarkus.langchain4j.ollama.smaller.chat-model.temperature=0
quarkus.langchain4j.ollama.smaller.chat-model.num-predict=1024
quarkus.langchain4j.ollama.smaller.timeout=60s
# --- Observability: log prompts and responses in dev mode only (they record user content). ---
%dev.quarkus.langchain4j.log-requests=true
%dev.quarkus.langchain4j.log-responses=true
# --- Dev Services off: a real Ollama endpoint is configured above. ---
quarkus.langchain4j.devservices.enabled=false
# ============================================================================
# Cloud models (alternative to local). Requires an Ollama cloud account.
# Privacy note: the default configuration above is local — prompts never leave
# the machine. Switching to a cloud model sends prompt content (e.g. support
# tickets) to that provider; apply your organization's consent, redaction, and
# data-retention rules first.
# quarkus.langchain4j.ollama.chat-model.model-id=deepseek-v4-pro:cloud
# quarkus.langchain4j.ollama.main.chat-model.model-id=deepseek-v4-pro:cloud
# quarkus.langchain4j.ollama.smaller.chat-model.model-id=gpt-oss:20b-cloud
# Structured output: deepseek-v4-pro:cloud and gpt-oss:20b-cloud did not enforce
# the JSON schema `format` when tested directly against Ollama (they answered in
# prose); local qwen3:1.7b did. Record-returning services then depend on their
# output guardrail (CompleteAnswerGuard in Agent.java.template), which asks again
# and fails after quarkus.langchain4j.guardrails.max-retries attempts.
# ============================================================================
# ============================================================================
# RAG (Easy RAG) — uncomment ALL FOUR keys when the quarkus-langchain4j-easy-rag
# extension and an in-process embedding model
# (langchain4j-embeddings-bge-small-en-v15-q) are on the classpath. Documents in
# this folder are ingested into an in-memory store on startup.
# Where the documents live: in src/main/resources/rag, read from the classpath
# (path-type=classpath, path=rag). The build copies them into the application, so
# dev mode, tests and the packaged application (java -jar, a container image) find
# them wherever the process starts. A filesystem path would be resolved against the
# working directory: src/main/resources/rag exists only in the source tree, and the
# packaged application fails to start elsewhere ("... is not a directory").
# Documents kept outside the application: point at an absolute folder at runtime,
# e.g. QUARKUS_LANGCHAIN4J_EASY_RAG_PATH_TYPE=filesystem and
# QUARKUS_LANGCHAIN4J_EASY_RAG_PATH=/srv/rag (environment variables override this file).
# Dev mode ingests on each restart, not when a document changes; see the RagSetup template.
# Easy RAG augments EVERY AI service by default: a service that must not see the
# retrieved documents opts out with @RegisterAiService(retrievalAugmentor =
# RegisterAiService.NoRetrievalAugmentorSupplier.class), as the templates' non-RAG
# services do.
# The embedding provider must be named: the Ollama extension ships an embedding
# model too, so two providers are present and the application refuses to start
# until one is selected.
# The %test line skips ingestion in tests, so the test application does not load
# the in-process model at startup. Continuous testing (Dev UI / the Dev MCP's
# devui-testing_runTests) runs the tests inside the dev-mode JVM, where dev mode
# has already loaded the model's native tokenizer library; a second load from the
# test classloader fails with "UnsatisfiedLinkError: Native Library ... already
# loaded in another classloader". The store stays empty in tests. A test that
# needs retrieved passages must turn ingestion back on, so it cannot run through
# continuous testing; it runs in a JVM of its own (`mvn test`, e.g. in CI).
# Native builds: Easy RAG needs the `native` profile part of pom.xml.template,
# which leaves its Tika document parser out of the native build; the native
# executable then ingests text and Markdown documents only. It cannot read them
# from the classpath (startup fails with FileSystemNotFoundException: "The Native
# Image Resource File System is not present"), so it takes the two environment
# variables above (filesystem, absolute path). That profile part sets them for
# `mvn verify -Dnative`; to run the executable yourself, ship the documents folder
# with it and set both. See RagSetup.java.template, "NATIVE BUILDS".
# ============================================================================
# quarkus.langchain4j.easy-rag.path-type=classpath
# quarkus.langchain4j.easy-rag.path=rag
# quarkus.langchain4j.embedding-model.provider=dev.langchain4j.model.embedding.onnx.bgesmallenv15q.BgeSmallEnV15QuantizedEmbeddingModel
# %test.quarkus.langchain4j.easy-rag.ingestion-strategy=off
## Alternative: embed through Ollama instead of in-process (pull an embedding model
## first and set quarkus.langchain4j.ollama.embedding-model.model-id to it):
# quarkus.langchain4j.embedding-model.provider=ollama
# ============================================================================
# MCP client — uncomment when the langchain4j-mcp extension is present. Declares
# a named client ("ops") whose tools an AI service imports with @McpToolBox("ops").
# streamable-http is the recommended transport; stdio spawns a local server process.
# Point clients only at servers you operate or explicitly trust — see the note in
# McpClient.java.template.
# ============================================================================
# quarkus.langchain4j.mcp.ops.transport-type=streamable-http
# quarkus.langchain4j.mcp.ops.url=http://localhost:8081/mcp
## stdio alternative (local subprocess). Replace BOTH placeholders below with a
## server you operate or explicitly trust, pinned to an exact version. `npm exec`
## needs the package already installed or --yes: a stdio pipe cannot answer a prompt.
# quarkus.langchain4j.mcp.ops.transport-type=stdio
# quarkus.langchain4j.mcp.ops.command=npm,exec,--yes,<your-mcp-server>@<exact-version>
# ============================================================================
# MCP server — uncomment the WHOLE block when an MCP server extension
# (mcp-server-http or mcp-server-stdio) is present. Adding the extension WITHOUT
# the support-langchain4j-annotations=false line below publishes EVERY LangChain4j
# @Tool method of the application (tools meant for your own model) to every MCP
# client: the extension's default is true. With the line, the server serves only
# methods annotated with the MCP @Tool; expose a LangChain4j tool by adding an MCP
# @Tool method that delegates to it (McpServer.java.template). The value is fixed
# at build time, so set it unscoped, not under a profile such as %dev.
# The Streamable HTTP endpoint is served at /mcp by default. Traffic logging prints
# every JSON-RPC exchange, so it is %dev.-scoped below: it records user content.
# In production, restrict /mcp to trusted, authorized clients (see the Security
# block below).
# ============================================================================
# quarkus.mcp.server.http.root-path=/mcp
# quarkus.mcp.server.support-langchain4j-annotations=false
# %dev.quarkus.mcp.server.traffic-logging.enabled=true
# %dev.quarkus.mcp.server.traffic-logging.text-limit=500
# ============================================================================
# Security (production) — the scaffolded endpoints ship open so dev mode runs
# without an identity store. Before exposing them beyond local dev or an
# internal gateway, add an auth extension (e.g. quarkus-oidc) and protect the
# HTTP surface; @Authenticated / @RolesAllowed work on WebSockets Next
# endpoints, and an HttpUpgradeCheck can gate the handshake (websockets-next
# reference, "Security"; quarkus-langchain4j secure-sql-chatbot sample).
# ============================================================================
# quarkus.oidc.auth-server-url=https://your-idp.example.com/realms/app
# quarkus.oidc.client-id=your-client-id
## A browser cannot send an Authorization header on a WebSocket upgrade, so browser
## clients need the web-app (cookie/session) type; bearer-only ("service", the
## default) suits non-browser clients, which an HttpUpgradeCheck can admit instead.
# quarkus.oidc.application-type=web-app
## Permission paths are matched exactly, with no implicit wildcard: list /mcp AND
## /mcp/*, because the MCP server also serves /mcp/sse and /mcp/messages/* under
## its root-path.
# quarkus.http.auth.permission.authenticated.paths=/ws/triage,/mcp,/mcp/*
# quarkus.http.auth.permission.authenticated.policy=authenticated
# ============================================================================
# Observability — uncomment when quarkus-micrometer-registry-prometheus and
# quarkus-opentelemetry are on the classpath. AI-service metrics (incl. GenAI
# token usage gen_ai.client.token.usage) and per-service/per-tool spans are then
# emitted automatically; metrics are scraped at /q/metrics. Keys below only tune it.
# ============================================================================
## Ship traces to an OTLP collector (the default endpoint, shown explicitly):
# quarkus.otel.exporter.otlp.traces.endpoint=http://localhost:4317
## Copy prompt/completion text onto spans (records user content) — %dev.-scoped, so
## uncommenting cannot leak it from a production profile:
# %dev.quarkus.langchain4j.tracing.include-prompt=true
# %dev.quarkus.langchain4j.tracing.include-completion=true
# ============================================================================
# Error contract (RFC 9457) — nothing to uncomment: the http-problem extension
# works as soon as it is on the classpath, turning an exception that escapes a
# resource into an application/problem+json response instead of a raw 500. The
# keys below only tune it; include-details stays false so parser and binding
# failures do not echo internal class names back to the caller.
# ============================================================================
## Return 422 instead of 400 for bean-validation failures:
# quarkus.http-problem.constraint-violation.status=422
## Copy MDC keys into the problem document (e.g. a correlation id):
# quarkus.http-problem.include-mdc-properties=uuid
## Echo raw parser messages in "detail" (leaks internal class names) — %dev.-scoped,
## so production keeps the default false:
# %dev.quarkus.http-problem.include-details=true
# ============================================================================
# Agent skills — uncomment when the langchain4j-skills extension is present.
# Each listed directory holds one subdirectory per skill, each with a SKILL.md
# whose YAML front matter declares the skill's name and description. Pair with
# @Skills on the AI service (see AiService.java.template). status:preview.
# ============================================================================
# quarkus.langchain4j.skills.directories=skills,classpath:other-skills
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |