Current knowledge extraction contract
This page is generated from the current source build’s runtime constants and Schemas. It describes current main, without claiming npm availability. After changing the implementation, run yarn build:runtime && yarn build:docs; yarn build:docs:check verifies the generated content.
| Current contract | Value |
|---|---|
| Authored prompt-json | knowledge-extraction-v29 |
| Configured strict prompt | knowledge-extraction-v30 |
| Structural wire | omk-knowledge-extraction-wire/v1 |
| Domain response Schema | 4 |
| Extraction run Schema | 5 |
| Entity history Schema | 2 |
| System + input UTF-8 capacity | 1,048,576 bytes |
| Producer outcome JSON declaration ceiling | 536,805,376 bytes |
| Serialized study plan capacity | 16,777,216 bytes |
| Complete study inspection input capacity | 134,217,728 bytes |
| Default generation policy | knowledge-generation-single/v1 |
| Default OMK invocation count | 1 |
| Opt-in bounded policy | knowledge-generation-bounded/v1 |
| Bounded OMK invocation ceiling | 3 |
| Bounded reask ceiling (included in invocations) | 2 |
| Bounded output-token ceiling per invocation | 128,000 |
Byte capacities are not token, model-context or spending budgets. The outcome ceiling bounds the producer’s declared returned JSON; persisted archive envelopes have a separate file capacity. Bounded policy requires explicit opt-in, and reasks count toward total invocations; see the knowledge extraction guide for model and token-reservation conditions.
These current values do not rewrite historical prompts, runs, Schemas or measurement conclusions; different identities cannot inherit quality acceptance. See the entity extraction contract for historical identities and provenance limits.