chore: bump interp-engine to 1.6.0 in inference, graph and nla

AGENTS.md

@@ -107,7 +107,7 @@ live here at `interp-engine/`; it is now its own repository,
 with its own tests, docs, lint gates and release process. Its comparison harness against
 TransformerLens, nnsight, vLLM and SGLang went with it.
 
-Each of the three apps pins an **exact version** (`interp-engine[...]==1.3.3`) in its
+Each of the three apps pins an **exact version** (`interp-engine[...]==1.6.0`) in its
 `pyproject.toml`. Exact rather than floored because the engine is first-party and moves in lockstep
 with these services: a relock for some unrelated dependency must not silently swap the code that
 serves every activation.

apps/graph/pyproject.toml

@@ -12,7 +12,7 @@ dependencies = [
   "pydantic>=2.11.4",
   "torch>=2.11.0",
   "transformers>=4.57.1",
-  "interp-engine==1.3.3",
+  "interp-engine==1.6.0",
   "python-dotenv>=1.1.0,<2",
   "hf-transfer>=0.1.9,<0.2",
   "psutil>=7.0,<8",

apps/inference/pyproject.toml

@@ -46,7 +46,10 @@ dependencies = [
   # 1.3.0 renamed the vllm-freeze backend to vllm-static and split generation-only out as
   # vllm-generate; freeze_points/frozen_points became static_points. See AGENT_INTEGRATION.md
   # in the engine repo. This service's call sites moved with it in the same commit.
-  "interp-engine[vllm,quant]==1.3.3",
+  # 1.6.0 is additive on the surface this service calls: speed and memory work (half the
+  # static-buffer VRAM on vllm-static), per-model point fixes for Gemma 4, and a new
+  # `interp_engine.memory` sizer nothing here imports yet.
+  "interp-engine[vllm,quant]==1.6.0",
 ]
 
 [dependency-groups]