Agent Engineering

NosisTech AI Agent Engineering Series: practical, small Python examples that show common AI agent patterns without hiding the important parts behind a large framework.

The repo is organized as one folder per post. Each example is intentionally compact: read the local README.md, copy the environment template when one is provided, install that folder's requirements, and run agent.py.

Patterns Included

Folder	Pattern
post-01-tool-invocation-agent	Tool invocation and chart generation
post-02-autonomous-decision	Autonomous triage with escalation rules
post-03-market-data	Market data lookup and summarization
post-04-financial-news	Financial news briefing
post-05-async-standup-summarizer	Async team update summarization
post-06-flowise	No-code Flowise workflow
post-07-praisonai	Multi-agent research workflow
post-08-memory-augmented-agent	Persistent memory pattern
post-09-knowledge-retrieval-agent	Retrieval-augmented generation
post-10-document-intelligence-agent	OCR, extraction, and review routing
post-11-letta-memgpt	Letta-style memory manager
post-12-anythingllm	AnythingLLM API integration
post-13-phidata	Memory, knowledge, and tool use
post-14-planning-agent	Plan generation and execution
post-15-data-analysis-agent	Statistical analysis plus narrative
post-16-marketing-content-assistant	Multi-role content workflow
post-17-physical-world-sensing-agent	Sensor simulation and action policy
post-18-langflow	Langflow exported workflow and API client
post-19-dspy	DSPy prediction and optimization
post-20-chain-of-agents-orchestrator	Sequential specialist agents
post-21-conflict-resolution-agent	Dual-review conflict escalation
post-22-financial-advisory	Market/news advisory briefing
post-23-crewai	CrewAI research crew
post-24-autogen	AutoGen examples
post-25-agency-swarm	Agency Swarm orchestration
post-28-explainable-agent	Explainability, confidence, and counterfactuals
post-29-compliance-driven-agent	Compliance-driven software engineering
post-30-ai-system-governance-intake-agent	AI system governance intake
post-31-llm-guard	LLM safety gateway
post-32-griptape	Griptape-style governance pipeline
post-33-pydantic-ai	Pydantic-style governance decisions
post-34-fabric	Fabric-style prompt pattern runner
post-35-prompt-injection-defense-agent	Prompt injection screening
post-36-hallucination-detection-agent	Claim-level hallucination review
post-37-verification-validation-agent	Verification and validation claim gate
post-38-promptfoo-redteam-eval-agent	Promptfoo-style redteam evaluation harness
post-39-pentestgpt-authorized-red-team-agent	Authorized red-team scope review
post-40-gdpr-ccpa-compliance-agent	GDPR and CCPA privacy screening
post-41-browser-use-web-task-planner	Browser-use style web task planning
post-42-gpt-researcher	Research planning and source trust review
post-43-goose	Goose-style safety-gated action planning
post-44-openhands	OpenHands-style sandbox planning
post-45-next-agent	SWE-agent-style next action planning
post-46-taskweaver	TaskWeaver-style stateful data analysis
post-47-self-improving-agent	Self-improving prompt adaptation loop
post-48-general-problem-solver	Trust-then-escalate general problem solving
post-49-evoagentx	EvoAgentX-style single-agent tool execution loop
post-50-metagpt	MetaGPT-style role handoff through shared messages
post-51-smolagents-code-agent	Smolagents-style code-shaped action loop
post-52-bayesian-operations-gateway	Bayesian-style operations risk gateway
post-53-knowledge-gap-discovery-agent	Knowledge gap discovery and hypothesis gating
post-54-collective-intelligence-consensus-engine	Multi-agent consensus with adversarial review
post-55-aegis-constraint-agent	Deterministic constraint gate with model-assisted explanation

Example Quick Start

cd post-05-async-standup-summarizer
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -r requirements.txt
cp .env.template .env
python agent.py

Most examples expect a LiteLLM-compatible endpoint and environment variables such as LITELLM_BASE_URL, MODEL_NAME, and LITELLM_API_KEY. Check each post's README because some examples also need data-provider keys or local input files.

Repository Notes

• Generated files such as reports, chart images, memory logs, local databases, and .env files should stay out of version control.
• The examples favor clarity over production architecture. They are meant to make agent patterns inspectable, not to be drop-in enterprise systems.
• Some integrations require external tools such as LiteLLM, Langflow, AnythingLLM, CrewAI, AutoGen, or Agency Swarm.