examples

Langchain-hs Examples

A comprehensive suite of 43 verified example executables demonstrating each core component of the langchain-hs ecosystem using Ollama (local/offline), OpenAI (cloud API), and Google Gemini (Interactions v1 API).

How to Build & Run

From the repository root:

# Build all example executables
cabal build examples

# Run a specific example (e.g., Gemini simple example)
cabal run simplegemini

Or with Stack:

# Build with Stack
stack --stack-yaml examples/stack.yaml build

# Run a specific example
stack --stack-yaml examples/stack.yaml run <executable-name>

---

Environment Setup

• Ollama: Ensure Ollama is running locally (ollama serve). Pull required models (e.g. ollama pull llama3 or ollama pull gemma3).
• OpenAI: Set your API key:

  export OPENAI_API_KEY="sk-..."

• Google Gemini: Set your API key in environment or store in ~/.gemini_api_key:

  export GEMINI_API_KEY="AIzaSy..."

---

Available Executables

Component	Ollama Executable	OpenAI Executable	Gemini Executable	Description
Chat Models	simpleollama	simpleopenai	simplegemini	Basic invocation, templating, prompt generation, and batch
Streaming	streamollama	streamopenai	streamgemini	Conduit-based reactive token streaming (SSE)
Tools & Calling	toolollama	toolopenai	toolgemini	Function definition, invocation, and tool binders
RAG & Embeddings	ragollama	ragopenai	raggemini	Document loading, chunking, and embedding vectors (gemini-embedding-001)
Langchain Monad	monadollama	monadopenai	—	Decoupled LangchainT reader/error transformer
Structured Output	jsonollama	jsonopenai	—	Typed JSON Schema generation and parsing
Prompt Template	prompttemplate	prompttemplate	prompttemplate	Pure string template interpolation with pre-bound variables
Hybrid Retrievers	retrieverollama	retrieveropenai	—	BM25 + vector similarity hybrid retrieval
Memory Systems	memoryollama	memoryopenai	—	Conversation buffer and sliding window memory
Retrieval QA	retrievalqaollama	retrievalqaopenai	—	Question-answering pipeline over loaded documents
ReAct Agent	reactollama	reactopenai	—	Multi-turn reasoning and tool execution loop
Plan-and-Execute	planandexecuteollama	planandexecuteopenai	—	Two-phase planning and step execution agent
Guardrails	guardrailollama	guardrailopenai	—	Content safety and output length validation
Resilience	resilienceollama	resilienceopenai	—	Circuit breakers, retry policies, and fallbacks
Observability	observabilityollama	observabilityopenai	—	OpenTelemetry distributed tracing and structured logging
MCP Integration	mcpollama	mcpopenai	—	Model Context Protocol client over stdio
StateGraph	stategraphollama	stategraphopenai	—	Cyclic state machines with STM checkpoints
Multi-Agent	multiagentollama	multiagentopenai	—	Supervisor agent routing and multi-agent coordination
HITL	hitlollama	hitlopenai	—	Human-in-the-loop interrupts and state resumption
Runnables	runnableollama	runnableopenai	—	Pure AST composition (`