AnyoneLearnAI
TutorialsPathsPracticeAI ToolsReference
LevelEasyStart

Topics

Level Easy Change

All
All
All
All
All
  • 01Prompt recipe card
  • 02Prompt vs RAG vs fine-tune
  • 03Safe sharing checklist
  • 04Verify AI answers checklist
  • 05Pick a model for your task
  • 06Token & cost quick reference
  • 07Prompt injection defense patterns
  • 08RAG quality checklist
  • 09Streaming vs batch inference
  • 10Local vs cloud models
  • 11Agent loop cheatsheet
  • 12Chat memory strategies
  • 13Literacy trust checklist
  • 14Embedding dimensions cheat sheet
  • 15Production RAG ship card
  • 16Image prompt card
  • 17Hybrid search patterns card
  • 18Agent ship checklist
  • 19Eval gate card
  • 20Guardrails layers card
  • 21LoRA vs RAG decision card
  • 22Research bot ship card
  • 23ML Python starter card
  • 24Production ops signals
  • 25Canary ship card
  • 26Re-rank patterns
  • 27AI incident response card
  • 28How to summarize a long document
  • 29How to write a system prompt
  • 30How to chunk documents for RAG
  • 31How to study a chapter with AI
  • 32How to build a simple chatbot
  • 33How to debug a bad AI answer
  • 34How to call OpenAI API with streaming
  • 35How to add function calling
  • 36How to evaluate RAG quality
  • 37How to use AI for code review
  • 38How to build a 5-document RAG app
  • 39How to write a few-shot prompt
  • 40How to compare two models
  • 41How to summarize meeting notes
  • 42How to create an image with AI
  • 43How to automate a workflow (no-code)
  • 44How to pick an embedding model
  • 45How to reduce hallucinations in answers
  • 46How to set up a Python virtual environment
  • 47How to deploy a small AI app
  • 48How to manage chat memory
  • 49How to secure API keys
  • 50How to ship with a canary deploy
  • 51How to test prompt changes
  • 52How to wire Anthropic Messages API
  • 53How to run a local LLM with Ollama
  • 54How to handle API rate limits
  • 55How to analyze a CSV with AI
  • 56How to build a minimal agent loop
  • 57How to run a canary deploy
  • 58How to spot synthetic media
  • 59How to read an AI regulation summary
  • 60How to embed documents in batch
  • 61How to log production LLM calls
  • 62How to add citations to RAG answers
  • 63How to write a professional email with AI
  • 64How to compare reasoning vs fast models
  • 65How to deploy a RAG app (checklist)
  • 66How to wire an MCP server
  • 67How to wire hybrid search for RAG
  • 68How to trace a failed RAG answer
  • 69How to ship an agent with an eval gate
  • 70How to run golden-task eval locally
  • 71How to add a tool allowlist
  • 72How to design an AI workflow checkpoint
  • 73How to red-team prompt injection
  • 74How to train a LoRA adapter
  • 75How to wire a multi-agent crew
  • 76How to call a vision API
  • 77How to set production AI alerts
  • 78How to profile token spend
  • 79How to verify research bot answers
  • 80How to run an A/B model test
  • 81How to triage an AI incident
  • 82Canary traffic split (sketch)
  • 83Injection test fixture (pytest sketch)
  • 84OpenAI chat — minimal
  • 85OpenAI chat — streaming
  • 86OpenAI embeddings
  • 87Anthropic messages — minimal
  • 88Minimal RAG loop
  • 89Function call handler
  • 90Chunk text by words
  • 91JSON schema prompt
  • 92OpenAI chat with tools
  • 93Next.js streaming API route
  • 94Summarize chat history
  • 95OpenAI batch embeddings
  • 96Validate JSON from LLM
  • 97Load API key from environment
  • 98Retry OpenAI on 429
  • 99Ollama chat API
  • 100Cosine similarity (NumPy)
  • 101OpenAI embeddings — batch
  • 102Anthropic Messages — streaming
  • 103RAG prompt with citations
  • 104Synthetic media disclosure label
  • 105RAG health check route
  • 106Minimal agent loop (Python)
  • 107Merge keyword + vector hits
  • 108Pass request ID through LLM calls
  • 109Golden eval exit code
  • 110Tool allowlist check
  • 111OpenAI vision message
  • 112LoRA training config sketch
  • 113A/B testing
  • 114Adapter (LoRA)
  • 115Adversarial prompt
  • 116Agent
  • 117Agent memory
  • 118AGI
  • 119AI
  • 120AI copyright
  • 121AI literacy
  • 122AI monitoring
  • 123AI regulation
  • 124Alignment
  • 125Anthropic API
  • 126API key
  • 127API rate limit
  • 128Audit trail
  • 129Baseline metric
  • 130Batch inference
  • 131Batch job
  • 132Batching (inference)
  • 133Benchmark
  • 134Bias
  • 135Canary deployment
  • 136Chain-of-thought prompting
  • 137Chatbot
  • 138Chunk overlap
  • 139Chunking
  • 140Citation (in AI answers)
  • 141Classification
  • 142Clustering
  • 143Completion
  • 144Computer vision
  • 145Confusion matrix
  • 146Consent
  • 147Context window
  • 148Conversation memory
  • 149Cosine similarity
  • 150Cost allocation
  • 151Cost per token
  • 152Cross-encoder
  • 153Cross-validation
  • 154Data drift
  • 155Data governance
  • 156Data pipeline
  • 157Dataset split
  • 158Dead letter queue
  • 159Decision tree
  • 160Deep learning
  • 161Deepfake
  • 162Diffusion model
  • 163Distillation
  • 164Embedding
  • 165Embedding batch
  • 166Embedding index
  • 167Embedding model
  • 168Eval gate
  • 169Eval set
  • 170Experiment tracking
  • 171Exponential backoff
  • 172Fairness
  • 173Faithfulness
  • 174Feature engineering
  • 175Few-shot prompting
  • 176Fine-tuning
  • 177Function calling
  • 178Generative AI
  • 179Git
  • 180GitHub
  • 181Golden task
  • 182GPU
  • 183Gradient descent
  • 184Groundedness
  • 185Grounding
  • 186Guardrails
  • 187Hallucination
  • 188Hugging Face
  • 189Human in the loop
  • 190Hybrid search
  • 191Idempotency
  • 192Incident response
  • 193Inference
  • 194Inference endpoint
  • 195Input filter
  • 196Jailbreak
  • 197JSON mode
  • 198JSON schema
  • 199KV cache
  • 200LangChain
  • 201Latency
  • 202LLM
  • 203LLM tracing
  • 204Local inference
  • 205LoRA
  • 206Loss function
  • 207Machine Learning
  • 208matplotlib
  • 209Max tokens
  • 210MCP
  • 211Metadata filter (vector search)
  • 212Model card
  • 213Model router
  • 214Model routing
  • 215Model serving
  • 216Model weights
  • 217Multi-agent system
  • 218Multimodal
  • 219Narrow AI
  • 220Neural network
  • 221NLP
  • 222Observability
  • 223Ollama
  • 224On-call
  • 225On-device AI
  • 226Open weights
  • 227Open-source model
  • 228OpenAI API
  • 229Orchestrator agent
  • 230Output validation
  • 231Overfitting
  • 232P95 latency
  • 233pandas
  • 234PEFT
  • 235Postmortem
  • 236Pre-training
  • 237Precision@k
  • 238Privacy
  • 239Production AI architecture
  • 240Prompt
  • 241Prompt caching
  • 242Prompt engineering
  • 243Prompt injection
  • 244Quality gate
  • 245Quantization
  • 246RAG
  • 247Random forest
  • 248Rate limit
  • 249Re-ranking
  • 250Reasoning model
  • 251Recall@k
  • 252Red teaming
  • 253Reinforcement learning
  • 254Responsible AI
  • 255Retrieval
  • 256Risk tier
  • 257RLHF
  • 258Rollback
  • 259Root cause analysis
  • 260Runbook
  • 261Schema validation
  • 262Semantic cache
  • 263Semantic search
  • 264Service recovery
  • 265Similarity search
  • 266SLO (service level objective)
  • 267Span (tracing)
  • 268Speech recognition
  • 269SSE (Server-Sent Events)
  • 270Streaming (LLM responses)
  • 271Structured output
  • 272Supervised learning
  • 273Sync API
  • 274Synthetic data
  • 275Synthetic media
  • 276System message
  • 277System prompt
  • 278Temperature
  • 279Text-to-image
  • 280Threshold
  • 281Token
  • 282Token budget
  • 283Tokenizer
  • 284Tool / function calling
  • 285Tool allowlist
  • 286Top-k
  • 287Top-p (nucleus sampling)
  • 288Traffic split
  • 289Transformer
  • 290Transparency
  • 291Unsupervised learning
  • 292User message
  • 293Vector
  • 294Vector database
  • 295Vector index
  • 296Virtual environment
  • 297Vision encoder
  • 298Webhook
  • 299Workflow automation
  • 300Workflow orchestration
  • 301Workflow trigger
  • 302Zero-shot prompting
All

Tutorials · Paths · Practice

Reference · Glossary

P95 latency

The response time below which **95% of requests** finish — catches slow tail better than average alone.

When to use

SLOs and alerts for chat and RAG APIs.

When not to

As the only metric — pair with error rate and quality proxies.

Example

p95 = 1.8s while mean = 0.9s → a few slow retrieves or long generations need traces.

Learn by doing →

Related terms

  • Latency
  • SLO (service level objective)
  • Baseline metric
← All terms
AnyoneLearnAI · Tutorials · Paths · Practice · AI Tools · Reference
HomeLearnPracticePaths

Topics

Level Easy Change

All
All
All
All
All
  • 01Prompt recipe card
  • 02Prompt vs RAG vs fine-tune
  • 03Safe sharing checklist
  • 04Verify AI answers checklist
  • 05Pick a model for your task
  • 06Token & cost quick reference
  • 07Prompt injection defense patterns
  • 08RAG quality checklist
  • 09Streaming vs batch inference
  • 10Local vs cloud models
  • 11Agent loop cheatsheet
  • 12Chat memory strategies
  • 13Literacy trust checklist
  • 14Embedding dimensions cheat sheet
  • 15Production RAG ship card
  • 16Image prompt card
  • 17Hybrid search patterns card
  • 18Agent ship checklist
  • 19Eval gate card
  • 20Guardrails layers card
  • 21LoRA vs RAG decision card
  • 22Research bot ship card
  • 23ML Python starter card
  • 24Production ops signals
  • 25Canary ship card
  • 26Re-rank patterns
  • 27AI incident response card
  • 28How to summarize a long document
  • 29How to write a system prompt
  • 30How to chunk documents for RAG
  • 31How to study a chapter with AI
  • 32How to build a simple chatbot
  • 33How to debug a bad AI answer
  • 34How to call OpenAI API with streaming
  • 35How to add function calling
  • 36How to evaluate RAG quality
  • 37How to use AI for code review
  • 38How to build a 5-document RAG app
  • 39How to write a few-shot prompt
  • 40How to compare two models
  • 41How to summarize meeting notes
  • 42How to create an image with AI
  • 43How to automate a workflow (no-code)
  • 44How to pick an embedding model
  • 45How to reduce hallucinations in answers
  • 46How to set up a Python virtual environment
  • 47How to deploy a small AI app
  • 48How to manage chat memory
  • 49How to secure API keys
  • 50How to ship with a canary deploy
  • 51How to test prompt changes
  • 52How to wire Anthropic Messages API
  • 53How to run a local LLM with Ollama
  • 54How to handle API rate limits
  • 55How to analyze a CSV with AI
  • 56How to build a minimal agent loop
  • 57How to run a canary deploy
  • 58How to spot synthetic media
  • 59How to read an AI regulation summary
  • 60How to embed documents in batch
  • 61How to log production LLM calls
  • 62How to add citations to RAG answers
  • 63How to write a professional email with AI
  • 64How to compare reasoning vs fast models
  • 65How to deploy a RAG app (checklist)
  • 66How to wire an MCP server
  • 67How to wire hybrid search for RAG
  • 68How to trace a failed RAG answer
  • 69How to ship an agent with an eval gate
  • 70How to run golden-task eval locally
  • 71How to add a tool allowlist
  • 72How to design an AI workflow checkpoint
  • 73How to red-team prompt injection
  • 74How to train a LoRA adapter
  • 75How to wire a multi-agent crew
  • 76How to call a vision API
  • 77How to set production AI alerts
  • 78How to profile token spend
  • 79How to verify research bot answers
  • 80How to run an A/B model test
  • 81How to triage an AI incident
  • 82Canary traffic split (sketch)
  • 83Injection test fixture (pytest sketch)
  • 84OpenAI chat — minimal
  • 85OpenAI chat — streaming
  • 86OpenAI embeddings
  • 87Anthropic messages — minimal
  • 88Minimal RAG loop
  • 89Function call handler
  • 90Chunk text by words
  • 91JSON schema prompt
  • 92OpenAI chat with tools
  • 93Next.js streaming API route
  • 94Summarize chat history
  • 95OpenAI batch embeddings
  • 96Validate JSON from LLM
  • 97Load API key from environment
  • 98Retry OpenAI on 429
  • 99Ollama chat API
  • 100Cosine similarity (NumPy)
  • 101OpenAI embeddings — batch
  • 102Anthropic Messages — streaming
  • 103RAG prompt with citations
  • 104Synthetic media disclosure label
  • 105RAG health check route
  • 106Minimal agent loop (Python)
  • 107Merge keyword + vector hits
  • 108Pass request ID through LLM calls
  • 109Golden eval exit code
  • 110Tool allowlist check
  • 111OpenAI vision message
  • 112LoRA training config sketch
  • 113A/B testing
  • 114Adapter (LoRA)
  • 115Adversarial prompt
  • 116Agent
  • 117Agent memory
  • 118AGI
  • 119AI
  • 120AI copyright
  • 121AI literacy
  • 122AI monitoring
  • 123AI regulation
  • 124Alignment
  • 125Anthropic API
  • 126API key
  • 127API rate limit
  • 128Audit trail
  • 129Baseline metric
  • 130Batch inference
  • 131Batch job
  • 132Batching (inference)
  • 133Benchmark
  • 134Bias
  • 135Canary deployment
  • 136Chain-of-thought prompting
  • 137Chatbot
  • 138Chunk overlap
  • 139Chunking
  • 140Citation (in AI answers)
  • 141Classification
  • 142Clustering
  • 143Completion
  • 144Computer vision
  • 145Confusion matrix
  • 146Consent
  • 147Context window
  • 148Conversation memory
  • 149Cosine similarity
  • 150Cost allocation
  • 151Cost per token
  • 152Cross-encoder
  • 153Cross-validation
  • 154Data drift
  • 155Data governance
  • 156Data pipeline
  • 157Dataset split
  • 158Dead letter queue
  • 159Decision tree
  • 160Deep learning
  • 161Deepfake
  • 162Diffusion model
  • 163Distillation
  • 164Embedding
  • 165Embedding batch
  • 166Embedding index
  • 167Embedding model
  • 168Eval gate
  • 169Eval set
  • 170Experiment tracking
  • 171Exponential backoff
  • 172Fairness
  • 173Faithfulness
  • 174Feature engineering
  • 175Few-shot prompting
  • 176Fine-tuning
  • 177Function calling
  • 178Generative AI
  • 179Git
  • 180GitHub
  • 181Golden task
  • 182GPU
  • 183Gradient descent
  • 184Groundedness
  • 185Grounding
  • 186Guardrails
  • 187Hallucination
  • 188Hugging Face
  • 189Human in the loop
  • 190Hybrid search
  • 191Idempotency
  • 192Incident response
  • 193Inference
  • 194Inference endpoint
  • 195Input filter
  • 196Jailbreak
  • 197JSON mode
  • 198JSON schema
  • 199KV cache
  • 200LangChain
  • 201Latency
  • 202LLM
  • 203LLM tracing
  • 204Local inference
  • 205LoRA
  • 206Loss function
  • 207Machine Learning
  • 208matplotlib
  • 209Max tokens
  • 210MCP
  • 211Metadata filter (vector search)
  • 212Model card
  • 213Model router
  • 214Model routing
  • 215Model serving
  • 216Model weights
  • 217Multi-agent system
  • 218Multimodal
  • 219Narrow AI
  • 220Neural network
  • 221NLP
  • 222Observability
  • 223Ollama
  • 224On-call
  • 225On-device AI
  • 226Open weights
  • 227Open-source model
  • 228OpenAI API
  • 229Orchestrator agent
  • 230Output validation
  • 231Overfitting
  • 232P95 latency
  • 233pandas
  • 234PEFT
  • 235Postmortem
  • 236Pre-training
  • 237Precision@k
  • 238Privacy
  • 239Production AI architecture
  • 240Prompt
  • 241Prompt caching
  • 242Prompt engineering
  • 243Prompt injection
  • 244Quality gate
  • 245Quantization
  • 246RAG
  • 247Random forest
  • 248Rate limit
  • 249Re-ranking
  • 250Reasoning model
  • 251Recall@k
  • 252Red teaming
  • 253Reinforcement learning
  • 254Responsible AI
  • 255Retrieval
  • 256Risk tier
  • 257RLHF
  • 258Rollback
  • 259Root cause analysis
  • 260Runbook
  • 261Schema validation
  • 262Semantic cache
  • 263Semantic search
  • 264Service recovery
  • 265Similarity search
  • 266SLO (service level objective)
  • 267Span (tracing)
  • 268Speech recognition
  • 269SSE (Server-Sent Events)
  • 270Streaming (LLM responses)
  • 271Structured output
  • 272Supervised learning
  • 273Sync API
  • 274Synthetic data
  • 275Synthetic media
  • 276System message
  • 277System prompt
  • 278Temperature
  • 279Text-to-image
  • 280Threshold
  • 281Token
  • 282Token budget
  • 283Tokenizer
  • 284Tool / function calling
  • 285Tool allowlist
  • 286Top-k
  • 287Top-p (nucleus sampling)
  • 288Traffic split
  • 289Transformer
  • 290Transparency
  • 291Unsupervised learning
  • 292User message
  • 293Vector
  • 294Vector database
  • 295Vector index
  • 296Virtual environment
  • 297Vision encoder
  • 298Webhook
  • 299Workflow automation
  • 300Workflow orchestration
  • 301Workflow trigger
  • 302Zero-shot prompting
All

Tutorials · Paths · Practice