Lucio Lelii
eaa5b04821
P2.2: Cache JSON schema generation (+102ms→12ms per request): - Add ConcurrentHashMap<Class<?>, JsonNode> to JsonSchemaProducer - Schema compilation moves from per-request to once-per-type P2.3: Filter executions at repository level (N users → 1 user query): - ExecutionsController.visibleExecutions() now uses getExecutionsByOwner() instead of getAllExecutions().filter() in-memory P2.4: Avoid Pattern.compile() per template resolution (3-5μs overhead): - Replace regex matching with simple string replace loop - Placeholder count typically < 50 keys P2.5: Cache ObjectMapper and WebClient in InternalOllamaLLMProvider: - Add static final ObjectMapper singleton (3-5ms creation cost) - Build WebClient once in constructor instead of per-request P2.5b: Make ObjectMapper static in SwitchExecutor and DelimitedParserExecutor: - Shared across all bean instances P2.6: Single-pass aggregation in UserStatsService (~4-5 passes→1 pass): - Add ExecutionStats record to collect all counters in one iteration - collectExecutionStats() consolidates running/succeeded/failed/simulations - Both getSystemStats() and buildUserStats() now use single pass Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com> |
||
|---|---|---|
| .mvn/wrapper | ||
| src | ||
| .gitattributes | ||
| .gitignore | ||
| Dockerfile | ||
| mvnw | ||
| mvnw.cmd | ||
| ollama-preloaded-docker | ||
| pom.xml | ||
| run_service.sh | ||