feat(inference): multi-provider LLM upstreams (Anthropic, Ollama) + pluggable guardrail pipeline (OpenAI Moderation) - #488

Merged

Pal Lakatos-Toth (pallakatos) merged 19 commits into

sandole:feat/inference-multi-provider-guardrails

Aug 28, 2026

Commits on Aug 25, 2026

feat(inference): multi-provider LLM upstreams + pluggable guardrail pipeline
John Seong (sandole)
committed
b277cc7
fix(guardrails): accurate SSE block frames + bounded stream scan context
John Seong (sandole)
committed
e7fb257
fix(guardrails): tolerant SSE data: prefix + raw-text scan fallback
John Seong (sandole)
committed
17a2cf1
fix(anthropic): strip hop-by-hop headers when relaying pass-through responses
John Seong (sandole)
committed
5d256db
fix(guardrails,provider): close review blockers B1-B4 + windowed scanning
John Seong (sandole)
committed
a447a9a
fix(guardrails): scan non-JSON SSE data frames instead of fast-releasing
John Seong (sandole)
committed
0459ae0
docs(inference): trim verbose comment blocks to load-bearing lines
John Seong (sandole)
committed
ca19334
fix(guardrails): close bypass through Foundry proxy inference routes
John Seong (sandole)
committed
6500341
refactor(controller): extract reconciler pod-spec helpers into submodule
John Seong (sandole)
committed
7a1e226
fix(guardrails): fail-closed pipeline hardening + decompose module
John Seong (sandole)
committed
720456b
fix(inference): route fail-closed enforcement, Anthropic streaming usage, unified error codes
John Seong (sandole)
committed
be5cb8b
docs(inference): security audit + enforcement-scope updates
John Seong (sandole)
committed
708b240

Commits on Aug 27, 2026

Commits on Aug 28, 2026

Merge branch 'main' into feat/inference-multi-provider-guardrails
Pal Lakatos-Toth
committed
77cade7
docs(security): add independent guardrails audit sign-off
Pal Lakatos-Toth
committed
f320067