Table of Contents

🇫🇷 Version française

See also: Back to index

Examples catalog

Everything under examples/ runs on the same engine; this page is the editorial map. The authoritative inventory — every numbered example with its process type, agent/task counts and referenced tools — is the generated examples/INDEX.md: it is produced by scripts/generate_examples_index.py and CI fails when it drifts from the folders on disk, so no count is maintained by hand here.

The nine business categories

The numbered YAML crews live in nine thematic folders:

Category Folder Focus
01 — Enterprise 01-enterprise/ CRM, due diligence, compliance, HR, supply chain
02 — Science & Research 02-science-research/ Literature analysis, computational biology, open science
03 — Finance & Trading 03-finance-trading/ Algorithmic trading, fraud detection, portfolio consensus
04 — Health & Wellness 04-health-wellness/ Diagnostic support, clinical trials, personalized plans
05 — Education 05-education/ Course design, assessment, tutoring, adaptive learning
06 — Engineering & DevOps 06-engineering-devops/ CI/CD automation, infrastructure, code review, testing
07 — Creative & Media 07-creative-media/ Content generation, scripting, design workflows
08 — IoT & Smart Systems 08-iot-smart-systems/ Monitoring, predictive maintenance, anomaly detection
09 — Experimental 09-experimental/ Prototypes, advanced orchestration research

Beyond the numbered crews

Folder What it shows
rag/ The RAG subsystem: basic-ingestion/, hybrid-retrieval/, custom-reranker/, crew-yaml/, plus the offline evaluation dataset under eval/
raggable-tree/ Semantic codebase analysis: basic-indexing/, crew-yaml/, custom-adapter/
scripting/ The TypeScript DSL (.ork.ts): hello world → dynamic spawn, FSM/graph literals, custom tools, RAG (08-rag.ork.ts)
cli-ts-commands/ Interactive REPL commands in *.cmd.ts, loaded without .NET recompilation
local-embeddings/ On-device embeddings (no API key) wired into memory and agent selection
crew-multifile/ A crew described as a directory (orkeon run <dir>)
forge/promote-demo/ The Atelier's offline half: a ready orkeon forge session bundled as a workspace — forge list, then forge promote into an ordinary folder (no LLM needed)
service-host/ The orkeon-host daemon's example material: a README and appsettings.host.json
run-events/ The orkeon run --events jsonl protocol: README, sample-stream.jsonl, watch-run.py
appsettings/, others/ The shared settings profile matrix and miscellaneous one-offs

Notable examples

  • Due diligence with NIST audit — 01-enterprise/07-due-diligence-nist/: hierarchical process, compliance tooling, long-term memory.
  • Algorithmic trading, multi-strategy — 03-finance-trading/31-algo-trading/: parallel specialists with consensus aggregation.
  • RAG ingestion to cited answers — rag/basic-ingestion/: ingest, retrieve, generate with citations — then graduate to hybrid-retrieval/ and the corrective profile.
  • Index a codebase, then ask it questions — raggable-tree/basic-indexing/: Tree-sitter indexing plus the 15 analysis tools.

Running an example

orkeon run examples/crew-multifile/          # a crew directory
orkeon run examples/scripting/01-hello-world.ork.ts
./examples/run-example.sh 01-enterprise/01-research-assistant   # from source
docker run -it --rm -e ORKEON_RUNNER=shell ghcr.io/orkeon/orkeon-runners
# then inside the container: orkeon-example list && orkeon-example run 7

From C# (via DI):

// crewFactory : ICrewFactory, orchestrator : ICrewOrchestrationService
var crew = await crewFactory.CreateFromDirectoryAsync(
    "examples/01-enterprise/07-due-diligence-nist/", ct);
var result = await orchestrator.KickoffAsync(crew.Id, CrewInput.Empty(), ct);

Examples ship configuration, not datasets — see the example data policy for how to mount your own input.

Edit this page