A positional wire format
for AI pipelines.
ASHRU is an eleven-field, pipe-delimited transport protocol for semantic data. The schema lives in the spec, not the payload — built for repetitive LLM extraction at infrastructure scale.
V|deploy|engineer|api||||staging|p|0|
The 11-field row
| # | Field | Role |
|---|---|---|
| 0 | V | version (always present) |
| 1 | action | the verb |
| 2 | subject | the agent |
| 3 | object | the patient |
| 4 | instrument | with what |
| 5 | recipient | to whom |
| 6 | source | from where |
| 7 | location | where |
| 8 | tense | p / n / f / h / c |
| 9 | negated | 0 or 1 |
| 10 | attributes | free-form annotation |
Empty fields are sequential pipes (||).
Literal pipes inside values are escaped (\|).
The version column is reserved for future revisions of the protocol.
Design principles
Schema-implicit
The schema lives in the protocol spec, not the payload. No repeated field names. No nested brackets. Field positions carry the meaning.
Positional
Eleven fields, always in the same order. Missing data is an empty cell, not an omitted key. Parsers know the contract.
Probabilistic in, deterministic out
Extraction from raw text via an LLM is probabilistic. Once a row is emitted, parsing, validation, and graph ingestion are strictly deterministic.
Streaming-native
Line-oriented by construction. Pipes into Kafka, Unix tools, and graph database loaders without buffering.