feat(inference): local InferenceGateway adapter — fail-closed, audits every call - #127
Merged
Conversation
… every call The local (agent-machine/Noetica) adapter of the InferenceGateway intersection seam (sourceos-spec inference-gateway-intersection.md). Serves only through a registered active InferenceProvider + an admitting consent decision, and emits a GatewayCallAudit (memory-mesh v0.1) on EVERY call including refusals — nothing served un-consented or un-audited. scripts/validate-inference-gateway.py (17 negative+positive controls) wired into make validate. Output returned only on outcome=ok.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The local fan-out of the InferenceGateway intersection seam (sourceos-spec
inference-gateway-intersection.md, #271). agent-machine/Noetica now implements the shared serving contract locally.src/agent_machine/inference_gateway.py—serve(request, provider, backend). Serves only through a registered activeInferenceProvider(contracts/inference-provider.schema.json) and an admitting consent decision, and emits aGatewayCallAudit(memory-mesh v0.1) on every call — including refusals (outcome="denied", no output). Nothing served un-consented or un-audited. Output only on outcome="ok".scripts/validate-inference-gateway.py— negative-control-first (a gateway never seen refusing == no gateway): 7 refusal modes each assert no output + a denied audit, plus the admitted path. Wired intomake validate.Verified:
make validate-inference-gateway→ 17/17 controls pass; refusals emit denied audits, admitted call returns output + matching receipt_hash.