On-VM embeddings
Record chunks are embedded with ONNX on your FastYoke VM. Vector search runs in Rust against tenant-scoped rag_chunks — no external embedding API.
Enterprise · Add-on
Ask your business data — retrieval stays on your VM
Yoker lives in the right-toolbar dock next to Messaging. It retrieves the most relevant chunks from your **entity records** and **text attachments**, then synthesizes an answer with citations. Embeddings run on your VM — tenant data is not sent to OpenAI or Anthropic for search.
Yoker answers. Yoke proposes.
| Yoker | Yoke | |
|---|---|---|
| Job | Answer questions about what you already stored | Propose workflow and schema improvements |
| Output | Chat reply + source chips | Brief in an inbox you approve |
| Retrieval | On-VM semantic search (RAG) | Template matchers over event log aggregates |
| Landing | Yoker docs | Yoke overview |
Privacy
- What makes Yoker different
Ephemeral sessions
The server does not persist chat history. The client keeps the thread during your session; each turn is forgotten server-side after the answer.
Cited sources
Answers include used_sources chips so operators can verify which records grounded the reply — critical for regulated workflows.
Corpus
- What Yoker knows today
Retrieval expands in waves; these ingestion sources ship now.
Entity records
Every entity record's data_payload is rendered to text, chunked (1500 chars / 200 overlap), embedded, and indexed per tenant.
Text attachments
PDF, Word, and plain-text attachments on entity records are parsed, chunked, and indexed alongside records — ideal for SOPs and contract PDFs.
Enable
- How to enable Yoker
- Enterprise / Fleet — Yoker is included.
- Pro / Team — requires the
yokeradd-on at $299/mo (check pricing for current rates). - Hobby / Solo — no add-on path.
Without entitlement, Yoker endpoints return 403 yoker_not_entitled with a CTA to activate the add-on.
Read the Yoker docs
API reference
Compliance workflows
Pair Yoker with Compliance Yoke for framework-readiness sweeps, then ask tenant-scoped questions about collected evidence — see Lab M13 in the training curriculum.
← All AI capabilities