Skip to content

Examples

Eight end-to-end example apps live in examples/ inside the repository. Every example builds, tests, and runs against a deterministic in-tree stub provider so CI never depends on a live LLM.

Run any example

From the repository root:

bash
pnpm install
pnpm --filter ./examples/<name> build
pnpm --filter ./examples/<name> test
GRAPHORIN_LLM_RECIPE=stub pnpm --filter ./examples/<name> dev

personal-assistant-cli

A 30-minute hands-on tour through Graphorin — wire createAgent({...}) to a six-tier Memory backed by SQLite + local embeddings, hook it up to one of three opt-in local-LLM stacks, and stream a real conversation through your terminal.

  • Recipes: stub, ollama, llamacpp-server, llamacpp-node.
  • Demonstrates: agent loop, memory, sensitivity gating, durable HITL, the GRAPHORIN_OFFLINE=1 contract.
  • Source: examples/personal-assistant-cli/.

multi-agent-crew

A supervisor + two-workers acceptance demo. Wires createAgent({ handoffs: [...] }) so the agent runtime auto-generates transfer_to_<worker> tools, emits a HandoffRecord per transfer, and lets one shared session reconstruct the full multi-agent conversation from JSONL export.

three-agent-harness

A three-agent crew driving the evaluatorOptimizer({...}) loop with a typed Zod rubric. Runs the generator → evaluator iteration with a strict iteration cap and a deterministic stub.

approval-workflow

A durable HITL workflow built on @graphorin/workflow. The flow validates an order, pauses for human approval (pause(...)), and ships once the approver resumes the thread — even on a different process.

document-pipeline

A document-ingestion pipeline that fans out work via Dispatch(...) across parallel workflow tasks and merges the results into the semantic memory tier.

background-consolidator

Drives @graphorin/memory's background consolidator across a long-running session. Demonstrates the per-tier ceiling and the tier: 'cheap' upgrade.

slack-bot-integration

An adapter sketch showing how to mount Graphorin as the brain behind a Slack bot — without making channel adapters part of the framework. The Slack bot is the consumer; Graphorin emits typed events the bot turns into messages.

local-stack-cli

A single-binary CLI that wires the standalone server, the triggers daemon, the encrypted-file secrets store, and a couple of skills into a turnkey local stack — useful as a starting point for production-shaped local deployments.

Tracing across every example

Set GRAPHORIN_TRACE=console to print finished spans from @graphorin/observability to your terminal:

bash
GRAPHORIN_TRACE=console GRAPHORIN_LLM_RECIPE=stub \
  pnpm --filter ./examples/personal-assistant-cli dev

The shared helper examples/example-trace-helper/ wires the console exporter consistently across the eight apps.

Production deployment templates

Three deployment templates live alongside the examples — they're not standalone apps, but reference manifests for shipping the standalone server:

Next steps


Graphorin · v0.1.0 · MIT License · © 2026 Oleksiy Stepurenko