<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Finastra AI COE Technology Radar</title>
    <link>https://finastra-ai.github.io/architecture-tech-radar</link>
    <description>Technology choices and recommendations from the Finastra AI Centre of Excellence. Subscribe to get notified when technologies are added or change ring levels.</description>
    <language>en</language>
    <lastBuildDate>Wed, 22 Jul 2026 14:16:57 GMT</lastBuildDate>
    <atom:link href="https://finastra-ai.github.io/architecture-tech-radar/feed.xml" rel="self" type="application/rss+xml"/>
    <item>
      <title><![CDATA[Async/Await Architecture — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/async-await-architecture/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[End-to-end asynchronous programming model adopted throughout Python AI microservices. Using Python&#39;s native async/await with asyncio, FastAPI, and async-compatible database/HTTP clients eliminates blocking I/O and enables high-concurrency request handling under the Uvicorn ASGI server without mu]]></description>
      <guid isPermaLink="false">async-await-architecture-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[Azure AI Foundry — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/platforms-and-operations/azure-ai-foundry/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Enterprise AI operations platform from Microsoft for managing the lifecycle of AI models and applications. Azure AI Foundry provides a unified hub for model catalogues, fine-tuning, prompt flow authoring, deployment management, and responsible AI evaluations. Adopted as the MLOps and AI project mana]]></description>
      <guid isPermaLink="false">azure-ai-foundry-2026-06-01</guid>
      <category>platforms and operations</category>
    </item>
    <item>
      <title><![CDATA[Azure AI Search — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/platforms-and-operations/azure-ai-search/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Fully managed cloud search service used for indexing and querying structured and unstructured content in AI-powered applications. Azure AI Search supports hybrid retrieval (keyword + vector), semantic ranking, and integrated vectorisation pipelines connected to Azure OpenAI embeddings. Adopted in th]]></description>
      <guid isPermaLink="false">azure-ai-search-2026-06-01</guid>
      <category>platforms and operations</category>
    </item>
    <item>
      <title><![CDATA[Azure OpenAI (GPT-4.1) — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/platforms-and-operations/azure-openai/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Production LLM service providing GPT-4.1 model access via the Azure OpenAI Service. Azure OpenAI delivers enterprise-grade SLAs, data residency guarantees, private endpoint support, and compliance controls required for financial services workloads. GPT-4.1 is the primary model for all natural-langua]]></description>
      <guid isPermaLink="false">azure-openai-2026-06-01</guid>
      <category>platforms and operations</category>
    </item>
    <item>
      <title><![CDATA[Black + Ruff + MyPy — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/black-ruff-mypy/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Python code quality toolchain adopted across all AI COE Python services. Black provides opinionated, deterministic code formatting eliminating style debates; Ruff delivers fast linting covering flake8, isort, and pyupgrade rules in a single tool; MyPy enforces static type checking against PEP-484 ty]]></description>
      <guid isPermaLink="false">black-ruff-mypy-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Checkmarx — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/checkmarx/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Static Application Security Testing (SAST) tool integrated into the CI pipeline as a mandatory security gate for all AI COE services. Checkmarx scans Python and Java source code for vulnerabilities including injection flaws, insecure deserialization, and hardcoded credentials. Required by Finastra s]]></description>
      <guid isPermaLink="false">checkmarx-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Circuit Breaker Pattern — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/circuit-breaker-pattern/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Fault-tolerance pattern applied to calls against external LLM providers, vector stores, and downstream backend APIs. When failure thresholds are exceeded the circuit opens, failing fast rather than exhausting thread pools or cascading latency. Implemented via Resilience4j in Java services and a comp]]></description>
      <guid isPermaLink="false">circuit-breaker-pattern-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[Cypress — Assess]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/cypress/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Alternative UI automation option with strong developer experience, but not the default for enterprise smoke strategy.
]]></description>
      <guid isPermaLink="false">cypress-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[DeepEval — Trial]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/deepeval/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Open-source LLM evaluation framework for measuring the quality of AI-generated outputs. DeepEval provides metrics such as answer relevancy, faithfulness, contextual precision, and hallucination detection, enabling automated regression testing of LLM pipelines. On trial to establish a repeatable eval]]></description>
      <guid isPermaLink="false">deepeval-2026-06-01</guid>
      <category>languages and frameworks</category>
    </item>
    <item>
      <title><![CDATA[Docker — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/platforms-and-operations/docker/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Standard containerisation technology for packaging all AI microservices and their dependencies into reproducible, portable images. All services in the AI COE stack are containerised with multi-stage Dockerfiles that produce minimal runtime images, enabling consistent behaviour across local developme]]></description>
      <guid isPermaLink="false">docker-2026-06-01</guid>
      <category>platforms and operations</category>
    </item>
    <item>
      <title><![CDATA[Entitlement-based Query Authorization — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/entitlement-based-query-authorization/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Dynamic SQL rewriting pattern that appends entitlement predicates to queries at runtime based on claims extracted from the authenticated user&#39;s JWT. This ensures that all database queries are transparently scoped to the data the caller is entitled to see without requiring application-layer filte]]></description>
      <guid isPermaLink="false">entitlement-based-query-authorization-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[FastAPI — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/fastapi/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Async REST framework of choice for Python-based AI microservices. FastAPI provides automatic OpenAPI schema generation, request validation via Pydantic, and native async/await support through ASGI. Its performance characteristics and developer ergonomics make it well suited for high-throughput LLM p]]></description>
      <guid isPermaLink="false">fastapi-2026-06-01</guid>
      <category>languages and frameworks</category>
    </item>
    <item>
      <title><![CDATA[FastMCP — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/fastmcp/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Python framework for building Model Context Protocol servers. FastMCP provides a high-level decorator-based API that simplifies exposing tools, resources, and prompts over the MCP protocol, reducing boilerplate and ensuring spec compliance. Adopted as the standard way to create Python-based MCP serv]]></description>
      <guid isPermaLink="false">fastmcp-2026-06-01</guid>
      <category>languages and frameworks</category>
    </item>
    <item>
      <title><![CDATA[Gatling — Assess]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/gatling/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Performance testing option for JVM-heavy teams with strong reporting, while k6 remains the default cross-team backend load testing standard.
]]></description>
      <guid isPermaLink="false">gatling-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Google Java Format — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/google-java-format/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Opinionated Java source code formatter that enforces Google&#39;s Java Style Guide, eliminating formatting discussions in code reviews. Integrated as a Maven/Gradle plugin and pre-commit hook, Google Java Format produces deterministic diffs that make code review focus on logic rather than style. Ado]]></description>
      <guid isPermaLink="false">google-java-format-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[HNSW Vector Index — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/hnsw-vector-index/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Hierarchical Navigable Small World (HNSW) is the approximate nearest-neighbour index algorithm adopted for all vector similarity search workloads. HNSW provides sub-linear query time with high recall, making it suitable for real-time semantic search over large embedding corpora. Used within Redis St]]></description>
      <guid isPermaLink="false">hnsw-vector-index-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[Hybrid Vector Search — Trial]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/hybrid-vector-search/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Search strategy that combines traditional keyword (BM25/TF-IDF) scores with dense vector similarity scores via a configurable weighting function, delivering better retrieval quality than either approach alone. On trial for payment message search use cases where domain-specific terminology requires e]]></description>
      <guid isPermaLink="false">hybrid-vector-search-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[Java — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/java/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Primary language for MCP server implementations in the GPP AI service stack. The current supported LTS versions are Java 17 and Java 21. Java 17 is the minimum baseline required by Spring Boot 3.x, providing sealed classes, pattern matching for instanceof, records, and text blocks. Java 21 adds virt]]></description>
      <guid isPermaLink="false">java-2026-06-01</guid>
      <category>languages and frameworks</category>
    </item>
    <item>
      <title><![CDATA[JMeter — Hold]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/jmeter/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Not preferred for new performance pipelines due to heavier maintenance and GUI-centric workflows when compared with script-first CI tooling.
]]></description>
      <guid isPermaLink="false">jmeter-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[JSqlParser — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/jsqlparser/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Java SQL parsing library used to validate and safely transform SQL queries as part of the entitlement-based query authorisation pattern. JSqlParser parses SQL strings into an abstract syntax tree, enabling the service layer to inspect, rewrite, and inject entitlement predicates without string concat]]></description>
      <guid isPermaLink="false">jsqlparser-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[JUnit 5 + Mockito — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/junit5-mockito/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Standard unit testing framework combination for Java MCP servers and Spring Boot services. JUnit 5 provides the test lifecycle model and parameterised test support, while Mockito enables clean mocking of Spring beans, OpenFeign clients, and LLM provider interfaces without starting a full application]]></description>
      <guid isPermaLink="false">junit5-mockito-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[k6 — Assess]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/k6/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Standard backend performance tool for API latency and error-rate gates with CI-friendly scripts and threshold-based validation.
]]></description>
      <guid isPermaLink="false">k6-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Karate — Assess]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/karate/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Primary API automation framework for cross-team regression coverage, readable assertions, and stable parallel execution in CI.
]]></description>
      <guid isPermaLink="false">karate-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Kubernetes + Helm — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/platforms-and-operations/kubernetes-helm/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Container orchestration platform and package manager for deploying and managing all AI microservices. Kubernetes provides declarative workload management, auto-scaling, and self-healing, while Helm charts encode deployment configurations and enable environment-specific overrides through values files]]></description>
      <guid isPermaLink="false">kubernetes-helm-2026-06-01</guid>
      <category>platforms and operations</category>
    </item>
    <item>
      <title><![CDATA[LangChain/LangGraph — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/langchain-langgraph/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Orchestration frameworks for building LLM-powered agents and multi-step pipelines. LangChain provides composable abstractions for chains, tools, memory, and retrievers, while LangGraph extends this with stateful, graph-based agent workflows supporting cycles and branching. Adopted as the standard ag]]></description>
      <guid isPermaLink="false">langchain-langgraph-2026-06-01</guid>
      <category>languages and frameworks</category>
    </item>
    <item>
      <title><![CDATA[LangSmith — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/platforms-and-operations/langsmith/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[LLM observability, tracing, and evaluation platform from LangChain. LangSmith captures full execution traces of LangChain/LangGraph agent runs, including LLM inputs/outputs, tool invocations, latency, and token costs. It provides a UI for debugging agent behaviour, running evaluation datasets, and c]]></description>
      <guid isPermaLink="false">langsmith-2026-06-01</guid>
      <category>platforms and operations</category>
    </item>
    <item>
      <title><![CDATA[Layered Architecture — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/layered-architecture/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Standard three-tier layered decomposition adopted for all AI microservices: controller (API boundary, request validation) → service (business logic, orchestration) → backend (data access, external integrations). This separation of concerns enforces single-responsibility at each layer, simplifies uni]]></description>
      <guid isPermaLink="false">layered-architecture-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[Lighthouse CI — Trial]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/lighthouse-ci/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Standard frontend quality gate for Core Web Vitals and accessibility budgets on critical routes in CI pipelines.
]]></description>
      <guid isPermaLink="false">lighthouse-ci-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Logback (async + rolling) — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/logback-async-rolling/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Structured logging framework for Java Spring Boot services, configured with asynchronous appenders and rolling file policies. Async appenders decouple log I/O from request threads, preventing slow log writes from impacting service latency. Logback is configured to produce JSON-formatted output (via ]]></description>
      <guid isPermaLink="false">logback-async-rolling-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Microsoft Azure — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/platforms-and-operations/microsoft-azure/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Primary cloud platform for all AI COE workloads. Azure provides the compute, networking, identity, and managed AI services underpinning the GPP AI service stack, including Azure Kubernetes Service for container orchestration, Azure OpenAI for production LLM access, and Azure AI Foundry for MLOps. Az]]></description>
      <guid isPermaLink="false">microsoft-azure-2026-06-01</guid>
      <category>platforms and operations</category>
    </item>
    <item>
      <title><![CDATA[Model Context Protocol (MCP) 2024-11-05 — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/model-context-protocol-mcp/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Core architectural protocol for connecting AI models to external tools, data sources, and services. MCP (spec version 2024-11-05) defines a standardised JSON-RPC-based interface for exposing tools, resources, and prompt templates that LLMs can invoke at runtime. Adopted as the foundational integrati]]></description>
      <guid isPermaLink="false">model-context-protocol-mcp-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[Multi-Phase AI Pipeline (5-phase) — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/multi-phase-ai-pipeline/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[A structured five-phase query orchestration pattern used to process natural-language requests in AI services: (1) intent classification, (2) entity extraction, (3) context enrichment, (4) LLM inference, (5) response formatting and validation. This decomposition improves testability, observability, a]]></description>
      <guid isPermaLink="false">multi-phase-ai-pipeline-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[Multi-Provider LLM Architecture — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/multi-provider-llm-architecture/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Architecture pattern that abstracts multiple LLM providers (Azure OpenAI GPT-4.1, Anthropic Claude) behind a unified interface, enabling runtime provider selection, automatic failover, and A/B evaluation without changing consuming services. Implemented via the Provider/Strategy pattern and circuit b]]></description>
      <guid isPermaLink="false">multi-provider-llm-architecture-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[NumPy — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/numpy/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Core numerical computing library used for embedding generation, vector mathematics, and similarity calculations in the AI microservice stack. NumPy provides highly optimised array operations that underpin cosine similarity, dot-product ranking, and batch embedding transformations. It is a transitive]]></description>
      <guid isPermaLink="false">numpy-2026-06-01</guid>
      <category>languages and frameworks</category>
    </item>
    <item>
      <title><![CDATA[NUnit — Assess]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/nunit/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Supported language-native unit and integration test framework for C#/.NET teams where it fits existing team practices.
]]></description>
      <guid isPermaLink="false">nunit-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[OAuth2 / JWT + mTLS — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/oauth2-jwt-mtls/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Dual-layer authentication pattern combining OAuth2 bearer tokens (JWT) for application-level identity and mutual TLS (mTLS) for service-to-service transport security. JWT tokens are issued by Keycloak (OIDC) and carry entitlement claims used for downstream authorisation decisions. mTLS ensures that ]]></description>
      <guid isPermaLink="false">oauth2-jwt-mtls-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[OpenAI SDK — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/openai-sdk/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Official Python and Java SDKs for integrating with OpenAI-compatible LLM endpoints, including Azure OpenAI. Adopted as the primary integration layer for GPT-4.1 in the AI COE stack. The SDK handles authentication, retries, streaming responses, and token counting, abstracting provider-specific HTTP d]]></description>
      <guid isPermaLink="false">openai-sdk-2026-06-01</guid>
      <category>languages and frameworks</category>
    </item>
    <item>
      <title><![CDATA[OpenAPI-First / Code Generation — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/openapi-first-code-generation/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[API-first design approach where the OpenAPI 3.x specification is the source of truth, with server stubs and client SDKs generated from the spec rather than hand-written. This pattern enforces contract consistency between producers and consumers, reduces integration bugs, and enables automatic docume]]></description>
      <guid isPermaLink="false">openapi-first-code-generation-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[Pact — Assess]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/pact/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Use for consumer-driven contract workflows when cross-team API compatibility risk justifies broker-based governance and verification.
]]></description>
      <guid isPermaLink="false">pact-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Playwright — Assess]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/playwright/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Default browser automation stack for smoke and critical user journeys because auto-waiting and trace artifacts reduce flakiness and debugging time.
]]></description>
      <guid isPermaLink="false">playwright-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[PostgreSQL Database — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/platforms-and-operations/postgresql-database/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Open-source relational database used for agent memory persistence (LangGraph checkpointing), vector storage via the pgvector extension, and general-purpose relational data needs in the AI COE stack. PostgreSQL + pgvector provides a unified store for both structured relational data and high-dimension]]></description>
      <guid isPermaLink="false">postgresql-database-2026-06-01</guid>
      <category>platforms and operations</category>
    </item>
    <item>
      <title><![CDATA[Postman/Newman — Hold]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/postman-newman/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Useful for quick API prototyping, but avoid as a primary CI regression approach because collection-heavy suites become brittle and drift over time.
]]></description>
      <guid isPermaLink="false">postman-newman-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Prometheus / Micrometer — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/platforms-and-operations/prometheus-micrometer/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Metrics collection stack for all AI microservices. Micrometer provides a vendor-neutral metrics facade in Java/Spring Boot services that exports to Prometheus, while Python services expose a /metrics endpoint using the prometheus-client library. Key metrics captured include LLM latency, token usage,]]></description>
      <guid isPermaLink="false">prometheus-micrometer-2026-06-01</guid>
      <category>platforms and operations</category>
    </item>
    <item>
      <title><![CDATA[Provider / Strategy Pattern — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/provider-strategy-pattern/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Design pattern used to abstract LLM and cache provider implementations behind a common interface, enabling runtime provider selection and seamless switching between Azure OpenAI, Anthropic, Redis, and Hazelcast without changing orchestration code. The strategy pattern also allows easy injection of m]]></description>
      <guid isPermaLink="false">provider-strategy-pattern-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[pytest — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/pytest/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Standard unit and integration test framework for Python AI microservices. pytest provides a concise fixture model, parametrised tests, rich assertion introspection, and an extensive plugin ecosystem (pytest-asyncio, pytest-cov, pytest-mock). Adopted as the mandatory test runner for all Python servic]]></description>
      <guid isPermaLink="false">pytest-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Python 3.11+ — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/python/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Primary language for all AI microservices in the GPP AI COE stack. Python 3.11+ delivers measurable performance improvements through faster CPython interpreter, improved error messages, and tomllib stdlib addition. Aligns with the broader ecosystem of ML/AI libraries (NumPy, LangChain, FastAPI) that]]></description>
      <guid isPermaLink="false">python-2026-06-01</guid>
      <category>languages and frameworks</category>
    </item>
    <item>
      <title><![CDATA[python-json-logger — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/python-json-logger/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Python logging formatter that serialises log records to JSON, enabling log aggregation platforms (e.g., ELK, Azure Monitor) to parse structured fields without regex. Used together with structlog to produce consistent, queryable JSON log output from all Python AI microservices. Adopted as the standar]]></description>
      <guid isPermaLink="false">python-json-logger-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Redis 8+ — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/platforms-and-operations/redis-8/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[In-memory data platform used as the vector store and semantic cache backend in development and lower environments. Redis 8 natively bundles vector search, full-text search, and JSON capabilities directly into the core server — the separate Redis Stack modules and the RediSearch module (deprecated si]]></description>
      <guid isPermaLink="false">redis-8-2026-06-01</guid>
      <category>platforms and operations</category>
    </item>
    <item>
      <title><![CDATA[REST Assured — Assess]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/rest-assured/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Recommended as a Java service-local API testing option for fast developer feedback alongside the primary cross-team API suite.
]]></description>
      <guid isPermaLink="false">rest-assured-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Retry with Exponential Backoff — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/retry-with-exponential-backoff/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Resilience pattern applied to transient failures when calling LLM APIs and downstream services. Requests are retried with exponentially increasing delays plus jitter to prevent thundering-herd stampede on provider rate limits or transient network errors. Adopted as a mandatory resilience primitive i]]></description>
      <guid isPermaLink="false">retry-with-exponential-backoff-2026-06-01</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[Selenium — Hold]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/selenium/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Not recommended for new UI automation due to higher flakiness and maintenance overhead compared with modern browser automation tooling.
]]></description>
      <guid isPermaLink="false">selenium-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[SonarQube — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/sonarqube/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Continuous code quality and security analysis platform integrated into the CI pipeline for all AI COE services. SonarQube tracks code smells, bugs, coverage trends, and OWASP security vulnerabilities across Python and Java codebases. Quality gate thresholds block merges when coverage drops or new cr]]></description>
      <guid isPermaLink="false">sonarqube-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Spring — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/spring/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[The Spring ecosystem is the standard framework suite for building Java-based MCP servers and backend services in the GPP AI stack. Key packages in use:

Spring Boot — production-ready defaults for health checks, metrics, and configuration management. Brings native GraalVM support, improved observabi]]></description>
      <guid isPermaLink="false">spring-2026-06-01</guid>
      <category>languages and frameworks</category>
    </item>
    <item>
      <title><![CDATA[structlog — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/structlog/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Structured logging library for Python services that produces machine-readable log output. structlog wraps Python&#39;s standard logging module with a processor pipeline that adds contextual fields (request ID, user ID, LLM provider, latency) to every log entry. Combined with python-json-logger, it e]]></description>
      <guid isPermaLink="false">structlog-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Testcontainers — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/testcontainers/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Library that provides throwaway Docker-based instances of real services (Oracle, PostgreSQL, Elasticsearch, Redis, Hazelcast) for integration tests. Testcontainers eliminates the need for shared test databases or mocks of infrastructure dependencies, giving integration tests high fidelity against pr]]></description>
      <guid isPermaLink="false">testcontainers-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[Uvicorn — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/uvicorn/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[ASGI server used to run FastAPI applications in production and development. Uvicorn is a lightning-fast, production-grade ASGI server built on uvloop and httptools, providing the concurrency layer that enables FastAPI&#39;s async request handling at scale. Standard choice for serving Python AI micro]]></description>
      <guid isPermaLink="false">uvicorn-2026-06-01</guid>
      <category>languages and frameworks</category>
    </item>
    <item>
      <title><![CDATA[xUnit — Assess]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/xunit/</link>
      <pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Recommended .NET option for service-local tests; combine with HttpClient for fast API feedback in C# teams alongside shared cross-team regression suites.
]]></description>
      <guid isPermaLink="false">xunit-2026-06-01</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[GitHub Copilot — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/tools/github-copilot/</link>
      <pubDate>Wed, 14 May 2025 00:00:00 GMT</pubDate>
      <description><![CDATA[AI-powered coding assistant used across the AI COE for code generation, review, and agentic development workflows.
]]></description>
      <guid isPermaLink="false">github-copilot-2025-05-14</guid>
      <category>tools</category>
    </item>
    <item>
      <title><![CDATA[RAG Pipelines — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/methods-and-patterns/rag-pipelines/</link>
      <pubDate>Wed, 14 May 2025 00:00:00 GMT</pubDate>
      <description><![CDATA[Retrieval-Augmented Generation patterns for grounding LLM responses with enterprise data. Core architectural pattern for AI COE applications.
]]></description>
      <guid isPermaLink="false">rag-pipelines-2025-05-14</guid>
      <category>methods and patterns</category>
    </item>
    <item>
      <title><![CDATA[TypeScript — Adopt]]></title>
      <link>https://finastra-ai.github.io/architecture-tech-radar/languages-and-frameworks/typescript/</link>
      <pubDate>Wed, 14 May 2025 00:00:00 GMT</pubDate>
      <description><![CDATA[Used for Next.js applications, VS Code extensions, and MCP server implementations across the AI COE platform.
]]></description>
      <guid isPermaLink="false">typescript-2025-05-14</guid>
      <category>languages and frameworks</category>
    </item>
  </channel>
</rss>