01 / Core Loops & Tools
Basics of initialization, tool registering, and model arguments validation.
- 01_basic_runtime.py
Runs a simple conversation loop between the model adapter and memory context.
- 02_tool_execution.py
Demonstrates automatic Pydantic-based JSON schema generation and successful function calls.
- 12_structured_data_extraction.py
Uses Pydantic models for verifying complex arguments fed into registered tool loops.
- 14_manual_state_injection.py
Manipulates agent conversational history and injects states programmatically mid-run.
02 / Governance & Safety
Deterministic security controls, approval interrupts, and recovery scenarios.
- 03_policy_interrupt.py
Configures a policy evaluator to automatically reject calls exceeding safety boundaries.
- 04_resume_after_approval.py
Saves states to checkpoints, pauses execution, and resumes once approval triggers.
- 05_credit_review_demo.py
Simulates a real banking workflow with multiple policy validations and audit trails.
- 13_error_recovery_retry.py
Implements self-healing agent pipelines that handle and retry on tool failure.
03 / Orchestration & Graphs
StateGraph orchestrations, symbolic piping, and conditional routing.
- 07_multi_agent_graph.py
Creates manual transitions, routing states across specialized support agents.
- 08_symbolic_chaining.py
Chains agents into sequential graphs using the symbolic chaining syntax sugar (
>>). - 09_conditional_router.py
Routes turns based on conversational intent and prints Mermaid diagrams of transitions.
- 11_researcher_writer_graph.py
Deploys a complex 3-agent pipeline involving researchers, writers, and editorial nodes.
04 / Production & Cloud
Context window optimization, cloud databases, and observability logs.
- 06_ollama_local_run.py
Bypasses network latency by executing agent loops entirely locally using Ollama models.
- 10_performance_bench.py
Benchmarks state caching hits and context window trimming rules during massive logs.
- 15_aws_lambda_handler.py
Serverless deployment handler template combining AWS Lambda with Amazon DynamoDB tables.
- 16_distributed_redis_state.py
Maintains distributed state caches across clustered deployments using Redis adapters.
- 17_cloudwatch_tracing.py
Writes structured audit event metrics directly to Amazon CloudWatch JSON logs.