ARCANADA
← Autonomy / Output Guard
BUILDING L1 · target L3

Output Guard

Ecosystem-wide repair pipeline for structured LLM output.

A TypeScript and Python monorepo that repairs structured output from language models — JSON, YAML, TOML — so that a malformed response becomes a usable one or an explicit failure, never a silent misparse. Shared across the ecosystem rather than reimplemented per service.

Capabilities

  • Repairs malformed JSON, YAML and TOML from model output
  • TypeScript and Python packages from one source of truth
  • Turns an unparseable response into an explicit failure rather than a guess

Current autonomy level

L1
What levels mean →

Weakest link

Repair is heuristic: a document that parses after repair is not thereby correct, and the pipeline cannot tell a fixed structure from a plausible-looking wrong one.

Roadmap to L3

  1. Schema-aware repair, so a repaired document is checked against what was expected
  2. Record repair rates per producer to find the models that need it most

Links

Next step

Review the current build and its development plan

BUILDING means active development, not a promise of public availability. Use the current capabilities and roadmap above as the reference.

Review the path to the target level