Software development guides

Topics

• AI agents in production

  AI agents and tool calls in production

  • Designing agent tools
  • Idempotency and retries
  • Approval gates and handoff
  • Budgets and spend control
  • Queues, deadlines and fallbacks
  • Prompt injection and fencing
• Retrieval and embeddings

  Storing and retrieving knowledge for AI

  • Chunking and freshness
  • Embeddings in Postgres
  • Vector index tuning
  • Choosing embedding models
  • Hybrid search and reranking
  • Retrieval quality
• AI testing and evals

  Testing and evaluating AI behavior

  • Golden sets and labels
  • Scoring model output
  • Judges and calibration
  • Stubbing and replay
  • Model swaps and gates
  • Drift detection
• Operating AI in production

  Running AI features reliably

  • AI SLOs and alarms
  • Tracing agent runs
  • Logs and audit trails
  • Incidents and runbooks
  • On-call for AI systems
  • Cost and capacity
• API design and integration

  API design and integration

  • REST and OpenAPI design
  • Errors and status codes
  • Pagination and cursors
  • Webhooks
  • Contracts and versioning
  • SDKs and developer docs
• Database reliability

  Database reliability and migrations

  • Query plans and slow queries
  • Postgres locks
  • Migrations without downtime
  • Database performance tuning
  • Schema change review
  • Rollouts and rollbacks
• Deployment and releases

  Deployment, environments and releases

  • CI/CD and smoke tests
  • Preview and staging
  • Test data and fixtures
  • Edge functions and regions
  • Inference routing
  • Prompt and model releases
• Security and privacy

  Security, privacy and permissions

  • Secrets and rotation
  • Permissions and least privilege
  • Tenant isolation
  • PII and data retention
  • Residency and egress
  • Injection defenses