Blog

Understand the augmentation step of RAG: how retrieved documents are structured into prompts, how metadata and citations improve response quality, and how to avoid common pitfalls like prompt injection.

Understand the retrieval step of RAG: Learn how database queries, keyword search, vector search, and hybrid approaches find the right information to ground your LLM responses.

The Corvus.Json.CodeGeneration library analyses JSON Schema and builds a TypeDeclaration tree that maps schema patterns to code patterns. The code generator, CLI tool, source generator, and Blazor playground all consume the same abstraction - and you can too.

The final post in the Optimising DAX series: the CALCULATE trap, variables and IF.EAGER, slicer costs, and a practical approach to isolating slow queries.

Data materialisation is when the storage engine gives up on efficient processing and rebuilds the entire table. This post explains what triggers it, why it's expensive, and how to delay it.

A deep dive into the cross-assembly linking, PDB-based source links, TFM scanning, enrichment merging, and search indexing that power our API reference documentation pipeline.

We generate about 8,800 API reference pages from 16 libraries across two engine versions, with source links, TFM availability badges, and hand-authored enrichments - all from a single PowerShell pipeline. Here's how.

A flat copilot-instructions.md works for small projects. When your codebase hits 500k lines, 19 projects, and multiple code generators, you need something more structured. Here's how we built a modular skills library for Corvus.Text.Json - and the design principles that make it work.

When you execute a DAX query, two engines divide up the work. Understanding the split between the formula engine and the storage engine is key to writing performant DAX.

Traditional cloud SLAs measure API uptime, not behavioural stability. This post talks about how silent AI model drift can quietly break your agents and the questions you might want to ask vendors.

Microsoft's new WSL Containers (wslc) promises container management on Windows without Docker Desktop. I tried it with dev containers, Aspire, and Docker Compose - here's what works, what doesn't, and whether you should switch yet.

Implement a custom IMessageTransport for brokers not covered by the built-in packages - PublishAsync, SubscribeAsync, RequestAsync, dead-letter routing, and optional health check support.

Test Corvus-generated AsyncAPI producers and consumers without a running broker - unit test handler logic with typed payloads, verify producer output with message capture, and integration test the full publish-consume round-trip using InMemoryMessageTransport.

Generate producers and consumers for only the channels you need - use channel filtering, tag selection, and the asyncapi-show preview command to carve focused messaging code from large AsyncAPI specifications.

Configure durable message delivery for Corvus-generated AsyncAPI consumers - Kafka consumer groups, AMQP acknowledgement, NATS JetStream, MQTT persistent sessions, and Azure Service Bus PeekLock settlement.

Add observability to Corvus-generated AsyncAPI producers and consumers - OpenTelemetry distributed tracing, message throughput metrics, processing loop heartbeats, and ASP.NET Core health checks for transport connectivity.

Generate typed request/reply methods from AsyncAPI specs with Corvus - correlated messaging with zero-allocation correlation IDs, typed response payloads, timeout management, and support for both AsyncAPI 3.0 reply objects and 2.6 x-corvus-reply extensions.

Generate typed message consumers from AsyncAPI specs with Corvus - handler interfaces with pre-validated payloads, configurable error policies, dead-letter routing, and Polly resilience middleware.

Generate strongly-typed message producers from AsyncAPI specs using the Corvus CLI - validated payloads, channel address templates, and pluggable transports for NATS, Kafka, AMQP, MQTT, and Azure Service Bus.

Test Corvus-generated OpenAPI code at every level - unit test handler logic in isolation with typed params, integration test the full round-trip with WebApplicationFactory, and verify client behaviour with in-memory transports.

Generate code for only the endpoints you need - use path filtering and the openapi-show preview command to carve focused, domain-specific clients and server stubs from large OpenAPI specifications.

Generate typed webhook receivers and senders from OpenAPI specs with Corvus - callback server stubs for receiving notifications, callback clients for sending them, runtime expression resolution, and linked operations for hypermedia-style API navigation.

Rx.NET 7.0 is now available, with the potential to reduce deployable application size by up to 90MB.

Corvus generates per-operation OAuth2 scope constants from your OpenAPI spec - use them with standard .NET DelegatingHandler middleware and Azure.Identity for type-safe authentication without proprietary frameworks.

Generate typed streaming responses from OpenAPI specs with Corvus - SSE and NDJSON on the server with writer callbacks, IAsyncEnumerable on the client with pooled typed documents and automatic frame parsing.