Deep Dive 74
- XOR in NeuroLink: a decision model that reads images and answers in numbers
- How we built a 14-gate AI content pipeline
- Testing proxyIR for exhaustiveness
- Two decision providers, one decide(): what Laya forced on NeuroLink
- Managing a 64-model catalog without chaos
- generate, stream, decide: a third inference type for NeuroLink
- Correlating traces across providers
- Cerebras integration deep dive
- git-tools: why we pass values, not flags
- Hardening the background-command tool
- What a green provider sweep proves, and what it skips
- The byte-cursor ledger: tracking proxy state precisely
- Designing the /accounts endpoint
- Why Azure is excluded from the model catalog
- Latency engineering for realtime voice agents
- Coordinating multiple agents without collisions
- Deterministic replay for proxy debugging
- Zero-downtime updates for the proxy
- Fixing SSRF and TOCTOU in fetch pinning
- Scraping the dependents graph to find real users
- What our first AI share-of-voice baseline actually measured
- Two backends, two stores, one TaskManager: how NeuroLink schedules AI work
- Embedding fast-path and session stickiness
- NeuroLink Brownfield Integration: Express & Fastify Adapters
- Voice as Three Stream Topologies: TTS, STT, and Full-Duplex Realtime
- Inside NeuroLink's RAG: The generate() Integration Bridge
- Grading the model: the scorer hierarchy and evaluation pipeline
- Inside ConversationMemoryFactory: How NeuroLink Picks and Wires a Memory Backend
- ModelPool's error-class fallback design
- Scaling LiveKit rooms for production voice agents
- Twenty-four providers, one BaseProvider: the adapter catalog
- Flagging call-graph claims automatically
- From User Input to Provider API: The Five-Stage Message Flow
- Seventeen file processors, six categories, one priority system
- Four-stage context compaction: what runs when the model window fills up
- What You Actually Inherit When You Extend BaseProvider
- Disaster recovery from a session log
- How we shipped 12 providers in one PR
- NeuroLink can now generate music: how it works
- Hitting the SmolLM2 ceiling
- Rolling out 4 new providers (2 local, 2 cloud) — and the AI SDK bug we hit
- stepIndex bookkeeping: keeping multi-step calls in order
- How We Test NeuroLink: 20 Continuous Test Suites and Counting
- Live Documentation: Building an MCP Docs Server Inside Docusaurus
- MCP Circuit Breaker: Preventing Cascading Failures in AI Tool Calls
- Yama: AI-Native Code Review Powered by NeuroLink
- Tool Annotations: How NeuroLink Auto-Infers Metadata for Smart Routing
- MCP Auto-Discovery: Finding and Registering Tools at Runtime
- Claude Proxy: Multi-Account OAuth Pooling for Heavy Claude Code Use
- Hippocampus: Persistent Memory That Learns Across Conversations
- The hold-and-rewrite pipeline for stale drafts
- Security considerations for voice agents
- The silent Vertex default in your RAG pipeline
- NeuroLink + Docusaurus: How We Document an AI SDK
- Generating 10,000 Descriptions a Day: Cost-Optimized Content Pipelines
- How We Scaled to 13 Providers: The Provider Registry Story
- When One Model Isn't Enough: Multi-Model Consensus for High-Stakes Decisions
- Multi-Agent Networks: Orchestrating AI Teams with NeuroLink
- The Workflow Engine: Multi-Model Orchestration with Judge Scoring
- How We Built Multi-Provider Failover: Never Losing an API Call
- How We Built MCP Integration: Supporting 4 Transport Protocols
- Model Evaluation and Scoring: RAGAS-Style Quality Assessment
- Context Compaction: Managing Long Conversations Without Losing Information
- Microservices with AI: Integrating NeuroLink into Distributed Systems
- Building Multi-Tenant AI SaaS with NeuroLink
- Event-Driven AI: Building Reactive Systems with NeuroLink
- The Event System: Real-Time Hooks for AI Observability
- How We Built the RAG Pipeline: 10 Chunking Strategies and Why
- Advanced RAG: 10 Chunking Strategies, Hybrid Search, and Reranking
- How We Built Streaming Tool Calls: Real-Time AI at Scale
- Building Auditable AI Pipelines: HITL, Guardrails, and Observability for Regulated Industries
- The Factory + Registry Pattern: How NeuroLink Breaks Circular Dependencies
- The Middleware System: Analytics, Guardrails, and Custom Pipelines
- How We Built NeuroLink's Provider Abstraction: 13 APIs, One Interface