API Versioning as a Translation Layer
API versioning as an architecture: requests enter through a version header, pass through a chain of reversible version changes, and reach controllers that only ever see the latest model.
Loading…
Topic
Architecture
830 posts about Architecture. Every summary links to the original.
- 830 posts
- 50 companies
- 31 explainers
See the idea
Architecture explainers
BIAN as a Contract Layer: Modernizing Core Banking One Domain at a Time
BIAN gives banks a shared vocabulary and service-domain contracts that sit between business concepts and the systems implementing them. See how domains, contracts, and governed access stack into one architecture.
Argo CD: The Reconciliation Loop Between Git and Your Cluster
Argo CD is a GitOps controller that treats Git as the desired state and continuously reconciles it against what actually runs in each cluster. It maps the parts of that loop and why pull-based agents change its scale.
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Newest first
Posts
Build Better AI Agents: 5 Developer Tips from the Agent Bake-Off
The Google Cloud AI Agent Bake-Off is presented as a test of building autonomous agents under severe time constraints for e-commerce, banking modernization, and startup go-to-market problems. Its five patterns recommend decomposing monolithic prompts into specialized sub-agents under a supervisor, designing modularly because model capabilities can quickly replace custom harnesses, and treating multimodality as native. It also advocates open protocols—including MCP, A2A, UCP, AP2, A2UI, and AG-UI—for agent and tool connections instead of bespoke wrappers. Finally, it separates probabilistic reasoning from execution: strict JSON schemas such as Pydantic validate LLM outputs before deterministic Python functions or SQL perform calculations and writes. The stated conclusion is that reliable production agents depend on rigorous software architecture rather than elaborate prompts or flashy single-shot demos.
Managing context in long-run agentic applications
Long-running multi-agent systems struggle with context management because accumulating raw message histories degrades inference quality, increases latency, and exceeds context window limits. In a collaborative security investigation platform, passing unrestricted history can also introduce confirmation bias across specialized agents. To maintain coherence across unbounded rounds of investigation, the system eliminates raw message history carryover between invocations. Instead, it coordinates agents through three structured context channels: a Director's Journal for orchestration memory, a Critic's Review that scores findings to filter hallucinations, and a Critic's Timeline of validated chronological events. This architecture provides agents with tailored context without overwhelming their inference capacity.
Using AI agents to build Stripe-Salesforce integrations faster
An experiment using Claude Code to build a Stripe App for Salesforce Flow exposed context gaps: the generated integration misnamed fields, targeted a nonexistent custom field, and omitted AsyncAfterCommit, causing a “Callout from triggers not supported” deployment error. The proposed remedy is a repository of grounding context files that present invocable signatures, request models, platform constraints, LWC patterns, and validated examples in concise, machine-readable form. Core files are about 35K tokens, while the full reference catalog is about 230K tokens and can be loaded on demand. With the context loaded, the agent selected the v02_CreateCustomers invocable directly and generated a helper Apex class for nested Checkout Session inputs, along with callout-safe Flow structure and other safeguards. The author reports minutes instead of hours and fewer iterations in personal usage, while noting that org-specific details still require human verification.
Enabling Multi-Region Data Systems at Notion
Notion’s data residency work keeps customer data processed and stored in the region where it was generated, supporting EU requirements across Search, analytics, scheduled events, and AI. To make this possible, Notion modularized infrastructure around isolated private networks and workspace IDs, then deployed regional data lakes, Kafka clusters, Spark jobs, Elasticsearch, and vector databases. Debezium captures Postgres changes into regional Kafka, while Airflow remains a centralized orchestration layer that routes jobs to regional EMR clusters; event logging similarly routes traffic by workspace and sanitizes data before Snowflake. Centralized mapping tables, shared region helpers, Terraform, and localized sanitization let teams retain a unified Snowflake environment without exporting customer data. The work took months, and Notion says it is applying these patterns to Japan and South Korea while aiming to provision future regional data stacks within days rather than months.
The three villains to agentic observability: retention, sampling and rollups
Observability practices increasingly depend on data sampling, short retention windows, and rollups to control storage costs. While human operators historically compensated for these compromises with intuition, AI agents lack implicit operational context and require full-fidelity telemetry to diagnose issues and justify conclusions. Head sampling drops unobserved spans early, tail sampling risks missing implicit non-error anomalies, and aggressive log expiration prevents long-term cyclical trend analysis. Addressing these constraints requires decoupling storage and compute via high-compression object storage and dedicated compute pools. Deploying real-time analytical databases such as ClickHouse enables high-concurrency, low-latency SQL querying, log clustering, and structured in-database summarization necessary for automated reasoning loops.
MongoDB Predictive Auto-Scaling: An Experiment
MongoDB Atlas initially relied on a reactive auto-scaler that took minutes to adjust instance sizes and only transitioned between adjacent tiers. Because these delays caused transient server over-utilization or unnecessary customer costs, researchers prototyped predictive auto-scaling to forecast demand. The experimental architecture used short-term and long-term forecasters for customer-driven metrics, an estimator using boosted decision trees to project CPU utilization, and a planner to pick the most cost-effective tier. In testing, the prototype kept replica sets closer to target CPU ranges while saving an average of nine cents per replica set hourly. MongoDB subsequently implemented a production version that scales replica sets up proactively before expected spikes while retaining reactive downscaling.
Frontend Engineering at Palantir: Plotlines in Three.js
Palantir’s frontend engineers are building a 3D implementation of Gaia to visualize Foundry Ontology objects with geospatial properties, including ground locations, satellite positions, and paths through space. The system uses Zodiac, a Three.js library backed by Foundry-based APIs: the backend stores, filters, and computes data, while the frontend receives simplified geometry and style information for rendering. Performance techniques include converting latitude and longitude to Earth-fixed coordinates once, batching similarly styled lines into single draw calls, and using instanced meshes plus batched updates for dense satellite scenes; orbital propagation uses TLEs client-side, while heavier analysis remains server-side. The article also describes frame-synchronized animation callbacks that update material properties without rebuilding geometry, while noting that 3D APIs continue to iterate across usability, platform integration, and performance as new data sources and workflows are added.
Frontend Engineering at Palantir: Building a Backend-less Cross-Application API
Palantir’s Application Frameworks team builds the “backend of the frontend”: frameworks that let applications communicate during workflows spanning multiple interfaces. The work addresses environments with denied, disrupted, intermittent, or limited (DDIL) connectivity, where relying on backend services can introduce unreliable latency. The team generalized the Gaia Controller prototype into Shared State, which lets applications define named data models and exchange updates organized by Media Type, and Commands, which exposes arbitrary logic through command IDs and request objects that return responses. Both frameworks use frontend-only browser APIs for client-side interaction, support workflows such as pairing Gaia with Workshop, and are also being used as tools by AIP agents; the post says adoption is growing across Palantir.
Improving storage efficiency in Magic Pocket, our immutable blob store
After introducing a direct erasure-coding service, Dropbox encountered severe storage fragmentation in its immutable blob store, Magic Pocket, due to a large number of under-filled volumes. Because their legacy compaction strategy was only designed to top off near-full volumes, the team created new multi-strategy compaction methods to aggressively consolidate sparse volumes. These approaches leverage dynamic programming and streaming pipelines to rapidly reclaim unused disk capacity at exabyte scale.
Welcome Gemma 4: Frontier multimodal intelligence on device
Google DeepMind’s Gemma 4 family is available on Hugging Face as Apache 2-licensed, multimodal models intended for use across agents, inference engines, fine-tuning libraries, and on-device deployments. The family accepts image and text inputs across five sizes, while E2B, E4B, and 12B Unified also support audio; context windows range from 128k to 256k tokens. Its design combines alternating local and global attention, dual RoPE configurations, Per-Layer Embeddings, shared KV caches, and model-specific vision and audio handling, including an encoder-free unified 12B variant. The post documents integrations for Transformers, llama.cpp, MLX, WebGPU, Rust, and other tools, alongside fine-tuning examples and demos. Reported instruction-tuned results include a 1452 estimated text-only LMArena score for 31B and benchmark results spanning reasoning, coding, vision, audio, and long-context tasks.
Rill and ClickHouse: real-time operational BI for a metered world
Granular operations across cloud computing, payment platforms, and artificial intelligence workloads produce high-volume event streams that traditional business intelligence tools struggle to unify and aggregate. Operational BI platform Rill addresses this challenge by combining declarative data loading through the dlt framework with ClickHouse's high-performance analytical aggregation engine. Data transformations, SQL-based measure expressions, and dashboard configurations are defined entirely as code using SQL and YAML. This declarative, metrics-first architecture allows developers to validate analytics locally, track changes in Git, and eliminate intermediate caching layers when querying live data. By querying ClickHouse directly and linking AI-generated insights back to underlying queries, the system provides traceable, real-time operational analytics for enterprise environments.
Falcon Perception
Falcon Perception is a 0.6B-parameter early-fusion Transformer for open-vocabulary grounding and segmentation from natural-language prompts. It processes image patches, text, and task tokens in one sequence, using hybrid attention so image tokens receive bidirectional context while text and task tokens decode causally. Its Chain-of-Perception interface predicts each instance as coordinate, size, and segmentation tokens, with specialized Fourier-feature and dot-product heads for dense outputs. The post also introduces PBench, which profiles attributes, OCR-guided identification, spatial understanding, relations, and crowded scenes. Falcon Perception reaches 68.0 Macro-F1 on SA-Co versus 62.3 for SAM 3, while presence calibration remains weaker at MCC 0.64 versus 0.82; the 0.3B Falcon OCR model scores 80.3 on olmOCR and 88.6 on OmniDocBench.
Any Custom Frontend with Gradio's Backend
gradio.Server enables applications built with React, Svelte, or plain HTML/JavaScript to use Gradio’s backend facilities instead of Gradio components. The post demonstrates this with Text Behind Image, whose browser UI layers a background, styled text, and a transparent foreground produced by BiRefNet, then exports a PNG client-side. A roughly 50-line Python backend extends FastAPI, exposes a standard HTML route, and defines an @app.api("remove_background") endpoint; @spaces.GPU handles ZeroGPU allocation while Gradio supplies queuing, concurrency control, SSE streaming, and gradio_client compatibility. The frontend calls the endpoint through the Gradio JavaScript client, so image-processing requests pass through the queue while rendering and compositing remain in the browser. The result is a custom frontend that retains Spaces hosting and Gradio infrastructure.
Granite 4.0 3B Vision: Compact Multimodal Intelligence for Enterprise Documents
Granite 4.0 3B Vision is a compact vision-language model designed to extract information from enterprise documents, including tables, charts, forms, and semantically meaningful key-value pairs. It combines a million-scale ChartNet dataset with a code-guided synthesis pipeline, DeepStack Injection for routing semantic and high-resolution visual features to different layers, and a LoRA adapter architecture built on Granite 4.0 Micro. The modular design supports standalone image processing, text-only fallback, and integration with Docling for multi-page document pipelines. Reported evaluations show 86.4% on Chart2Summary, leading the evaluated models, second place on Chart2CSV at 62.1%, leading scores on the listed table benchmarks, and 85.5% zero-shot exact-match accuracy on VAREX.
Training mRNA Language Models Across 25 Species for $165
OpenMed describes an end-to-end protein engineering pipeline that connects ESMFold structure prediction, ProteinMPNN sequence design, and custom mRNA codon optimization. The codon-language-model study compares CodonBERT, ModernBERT, and several CodonRoBERTa variants using 250,000 E. coli coding sequences, a 69-token codon vocabulary, masked-language modeling, and training on four 80GB A100 GPUs. CodonRoBERTa-large-v2 achieved perplexity 4.10, CAI Spearman correlation 0.404, and 7.7% synonymous recovery, while ModernBERT-base reached perplexity 26.24 and correlation 0.070. The work then trained a multi-species suite on 381,283 sequences spanning 25 species, producing four production models in 55 GPU-hours and documenting planned extensions such as larger training data and additional prediction heads.
ClickHouse BYOC on Google Cloud now Generally Available
ClickHouse Bring Your Own Cloud is now generally available on Google Cloud across all public ClickHouse Cloud Google regions without requiring a waitlist. The architecture separates the ClickHouse-managed control plane from the customer data plane, which runs entirely within the user Google Cloud project on Google Kubernetes Engine and Google Cloud Storage. An outbound-only, encrypted Tailscale zero-trust tunnel restricts control plane communication strictly to orchestration traffic, ensuring query data never leaves the customer environment. Deployment involves a three-step Terraform setup that establishes least-privilege IAM roles, configures VPC networking, and automates cluster provisioning. ClickHouse manages database and infrastructure maintenance through ArgoCD, employing a make-before-break strategy to add updated replicas before decommissioning older nodes across customizable release channels.
Predicting Rider Conversion in Sparse Data Environments with Bayesian Trees
Lyft requires real-time predictions of whether a rider will request a ride after viewing price and ETA details to balance supply and demand and personalize user experiences. Traditional machine learning models overfit on high-cardinality, sparse context intersections, whereas deep neural networks introduce unacceptable inference latency during live user sessions. To address this, engineers created a hierarchical Bayesian tree framework that organizes session data into increasingly granular context partitions. Each node houses a parametric model trained top-down, applying Gaussian Bayesian priors and L2 penalties relative to parent parameters to smooth predictions in data-rare child segments. Using simple parametric models at each node also permits monotonic constraints, guaranteeing logically consistent predictions aligned with domain expectations.
How Padlet uses ClickHouse Cloud to power real-time classroom analytics
Padlet needed a real-time analytics system to provide teachers with immediate insights into student engagement across 40 million monthly unique users. To avoid the operational overhead of managing database clusters with a small engineering team, the company adopted ClickHouse Cloud and ClickPipes for managed data ingestion. Active user sessions emit 30-second heartbeat events that stream alongside page views into ClickHouse. Materialized views pre-aggregate engagement duration and view counts by date, while native HyperLogLog sketches process unique visitor calculations efficiently across arbitrary time ranges. In production, this architecture ingested roughly eight billion events in a single month and delivered analytical query results with a 45-millisecond median latency.
How we’re building a data platform for a new user: agents
AI agents are emerging as autonomous database users that generate SQL, execute queries, and iterate in seconds. Because traditional analytics workflows built for human analysts cannot keep up with these rapid loops, ClickHouse is adapting its platform for agentic workloads. The proposed Agentic Data Stack combines LibreChat for conversational interfaces, ClickHouse with Model Context Protocol servers for the data layer, and Langfuse for observability. Supporting developer tooling includes chDB 4 with a Pandas-like API, IDE integrations, and ClickHouse Agent Skills to enforce database design best practices. Platforms like Anthropic and Character.AI already leverage ClickHouse's high-concurrency aggregation and real-time ingestion capabilities to handle massive data scales required by modern AI applications.
Building Biz Ask Anything: From Prototype to Product
Yelp expanded Yelp Assistant to business pages by developing a question-answering architecture that extracts concise, evidence-backed answers from reviews, photos, and structured facts. The production system replaces static snapshots with near-real-time streaming pipelines for reviews and structured properties alongside weekly batch ingestion for menus and website data. For each incoming query, the system evaluates intent, retrieves relevant context through keyword and embedding searches, and constructs targeted prompts with dynamically selected few-shot examples. Synthesized answers are augmented with supporting visuals and citations before being streamed token-by-token to users. Generating suggested questions directly from specific business content rather than generic categories improved user engagement by roughly fifty percent and reduced unanswerable queries by twenty-six percent.
Related topics
All topics →Weekly digest
The best engineering reads, once a week.