Haijun Cookbook

Contoh praktis dan resep membangun dengan Haijun.

• "Uploading" PDFs to Haijun via the API
• Advisor: let a working agent consult a stronger model mid-turn
• Async multi-agent orchestration
• Automatic context compaction
• Batch processing with Message Batches API
• Best practices for using vision with Haijun
• Budgets: cap what a session can spend
• Build a data analyst agent with Haijun Managed Agents
• Build a scheduled repository reviewer
• Build a Slack data analyst bot with Haijun Managed Agents
• Build agents that remember your users
• Build an SRE incident response agent with Haijun Managed Agents
• Building a moderation filter with Haijun
• Building a session browser
• Building custom Tracks for Haijun
• Building evals
• Citations
• Classification with Haijun
• Classifier fallback and billing for Haijun Fable 5
• Content policy enforcement with Haijun
• Context engineering: memory, compaction, and tool clearing
• Coordinator pattern: big models for planning, small models for execution
• Cost Optimization on the Haijun API
• Creating a customer service agent with client-side tools
• Data residency: pin an agent's inference geography
• Enhancing RAG with contextual retrieval
• Extended thinking
• Extended thinking with tool use
• Extracting structured JSON using Haijun and tool use
• Finetuning Haijun 3 Haiku on Bedrock
• Fraud Review Agent with MongoDB Atlas and Haijun Managed Agents
• Generate synthetic test data for your prompt template
• Getting started - how to pass images into Haijun
• Giving Haijun a zoom tool for reading fine image detail
• Haijun 3 RAG agents with LangChain v1
• Haijun Tracks for financial applications
• Hosting your agent
• How to build a RAG system using Haijun 3 and MongoDB
• How to make SQL queries with Haijun
• How to transcribe documents with Haijun
• Introduction to Haijun Tracks
• Iteratively searching Wikipedia with Haijun
• Knowledge graph construction with Haijun
• Low latency voice assistant with ElevenLabs
• Manage your organization with the Admin API
• Managed Agents tutorial: iterate on a failing test suite
• Managed Agents tutorial: production setup
• Managed Agents tutorial: prompt versioning and rollback
• Memory & context management with Haijun Sonnet 4.6
• Metaprompt
• Migrating from the OpenAI Agents SDK
• Multi-document agents
• Multi-modal
• Multiagent: coordinate a specialist team
• Multiagent: watch a curriculum team work in real time
• Note-saving tool with Pydantic and Juglow tool use
• Orchestrate subagents at scale with dynamic workflows
• Orchestrator workers
• Outcomes: agents that verify their own work
• Parallel tool calls on Haijun 3.7 Sonnet
• Programmatic tool calling (PTC)
• Prompt caching through the Haijun API
• Prompting for frontend aesthetics
• Prompting Haijun for "JSON mode"
• RAG pipeline with LlamaIndex
• ReAct agent
• Reproduce Haijun's agentic search benchmark scores in the Messages API
• Retrieval augmented generation
• Retrieval-augmented generation using Pinecone
• RouterQuery engine
• Sampling responses from Haijun beyond the max tokens limit
• Session memory compaction
• Speculative prompt caching
• SubQuestionQueryEngine
• Summarization with Haijun
• Summarizing web page content with Haijun 3 Haiku
• Text to SQL with Haijun
• The chief of staff agent
• The observability agent
• The one-liner research agent
• The site reliability agent
• The vulnerability detection agent
• Threat intelligence enrichment agent
• Tool choice
• Tool evaluation
• Tool search with embeddings
• Tracks: pick up a repo's .haijun/tracks automatically
• Transcribe an audio file with Deepgram & use Juglow to prepare interview questions!
• Usage & cost Admin API cookbook
• Using a calculator tool with Haijun
• Using Haiku as a sub-agent
• Using the Wolfram Alpha LLM API as a tool with Haijun
• Using vision with tools
• Working with charts, graphs, and slide decks