Agentic AI

67 problems

• Assemble the Critic's audit inputEasy
• Build the chat messages listEasy
• Cache-before-search, write-after-fetchMedium
• Which ReAct step are we on?Easy
• PASS or FAIL?Easy
• Count the dedup savingsEasy
• Find the bottleneck executorMedium
• Format retrieved docs as contextEasy
• Map your components to the frameworksEasy
• Rebuild findings from gather resultsEasy
• One model, three roles — route by promptEasy
• Accumulate findings across attemptsEasy
• MockLLM keyword routingMedium
• MockSearchEngine keyword retrievalEasy
• The Orchestrator feedback loopMedium
• Parallel execution profileMedium
• Parse the planner's sub-question listEasy
• Parse the SEARCH / ANSWER protocolEasy
• Parse the Critic's structured verdictMedium
• Agents are nouns; these are the verbsEasy
• Plan → execute → synthesizeMedium
• The RAG run: search → ground → answerMedium
• The ReAct loop with a max_steps guardMedium
• The whole system, expressedMedium
• The SharedMemory storeEasy
• When does the retry loop stop?Easy
• Aggregate trace eventsEasy
• Run totals from the traceEasy
• Validate a chat messageEasy
• Sequential vs parallel wall-clockEasy
• Bellman equation for value iterationMedium
• Citation Attribution AssemblyMedium
• Critic Verdict ParserEasy
• Exponential Backoff + JitterEasy
• Generalised Advantage Estimation (GAE)Medium
• GRPO objectiveHard
• JSON Tool-Arg RepairMedium
• KV-Cache Eviction (Sliding Window)Medium
• LinUCB Contextual BanditHard
• Multi-Tool Router (by Embedding)Medium
• Noisy top-k gatingMedium
• Paged-Attention KV Block AllocationMedium
• PII Detection & RedactionMedium
• Plan-DAG Cycle DetectionMedium
• Planner Output ParserEasy
• Policy gradient with REINFORCEMedium
• PPO clipped objectiveMedium
• Prefix-Cache Hit AccountingEasy
• Prioritized experience replayMedium
• Prompt-Injection HeuristicMedium
• Q-learning for MDPsMedium
• ReAct Loop DriverMedium
• Reflexion Retry With FeedbackMedium
• Refusal / Confidence GateEasy
• Run-Level Cost / Budget CeilingMedium
• Semaphore-Bounded Concurrent TasksMedium
• Shadow / Canary Rollback DecisionMedium
• Shared-Memory Tool-Call CacheMedium
• Sparse mixture-of-experts layerHard
• Streaming SSE Delta AssemblerEasy
• Task-DAG Parallel SchedulerHard
• TD(0) value updateEasy
• Thompson Sampling (Bernoulli-Beta)Medium
• Token-Budget Context ManagerMedium
• Tool-Call Parse + Schema ValidationMedium
• Trace Aggregation SummaryEasy
• UCB1 multi-armed banditEasy