Runtime Contract
| Endpoint | Method | Purpose |
|---|---|---|
/api/manifest | GET | Returns SEPL intent, extension points, policies, targets, and the canonical seed source. |
/api/schema | GET | Returns the machine-readable IR contract used by the workbench. |
/api/compile | POST | Transforms source prose into deterministic relation nodes and package affordances. |
/api/validate | POST | Checks IR structure, extension bindings, concepts, and truth notes. |
/api/run | POST | Compiles, validates, applies policy, computes source grounding, and returns telemetry. |
/api/execute | POST | Compiles controlled Castillan, executes it through the Python adapter, and returns node results plus learning evidence. |
/api/simulate | POST | Runs the planning pipeline against an existing IR object. |
/api/packages/install | POST | Installs and indexes a Python or Node package, returning its capability count. |
/api/llm/status | GET | Reports the configured local neural host, backend, model, and explicit errors. |
/api/gaia/status | GET | Alias for the GAIA local-host status contract. |
/api/training/status | GET | Reports whether PEFT can train from the configured Hugging Face checkpoint. |
/api/learning/trigger | POST | Attempts a manual training cycle and reports insufficient data or missing prerequisites as failure. |
/api/telemetry | GET | Reports runtime status plus real Python and Node availability inventory. |
Intermediate Representation
{
"language": "SEPL",
"version": "0.3.1",
"source_language": "Castilian Spanish",
"canonical_intent": "non-logical programming language workbench",
"source_hash": "sha256",
"node_count": 9,
"nodes": [
{
"id": "node-stablehash",
"order": 1,
"kind": "package_affordance",
"text": "Source sentence",
"extension_point": "package-affordance-index",
"capabilities": ["package-affordance-index"],
"concepts": ["python_ecosystem", "node_ecosystem"],
"grounding": 0.75
}
],
"package_affordances": {
"ecosystems": ["node-javascript-typescript", "python"],
"affordance": "package import mapping",
"execution_claim": "not executed by this prototype backend"
},
"truth_notes": [
"The backend performs deterministic semantic analysis and planning.",
"The planning backend does not execute package calls; /api/execute reports backed controlled Python execution separately."
]
}
Extension Points
| ID | Role |
|---|---|
public-language-parser | Parses Castilian Spanish and English prose into stable relation nodes. |
neural-substrate-map | Marks claims about the neural layer between public language and machine execution. |
package-affordance-index | Maps Node.js, JavaScript, TypeScript, and Python ecosystems as computational material. |
nonlogical-relation-linker | Preserves relations that intentionally exceed deductive or algebraic form. |
execution-plan-ledger | Produces reviewable plans while labeling unimplemented execution adapters truthfully. |
telemetry-and-validation | Reports grounding, runtime inventory, and mismatches without dummy success claims. |