f3413e6 · Oversight

Remove internal handoff cruft from the public tree

SESSION_RESUME.md and docs/RUNBOOK.md were internal operational
notes for the original author's homelab: CT numbers, Proxmox node
hostnames, private network IPs, and a step-by-step runbook for a
deployment that has nothing to do with the open protocol. They
have no place in a public repository and should not have shipped.

oversight_core/decoy.py: drop the homelab reference in the module
docstring and point the default OLLAMA_URL at loopback. Callers can
still override via the env var as documented.

docs/V05_REKOR_PLAN.md: replace a lingering "CT 222" reference with
neutral phrasing.

f3413e6 Zion Boggan committed on Apr 22, 2026 (2 months ago)

docs/V05_REKOR_PLAN.md +1 -1

		@@ -43,7 +43,7 @@ v0.5 lands as a stretch goal if evaluation work comes together first.

		## 2. Non-goals for v0.5

	-	- No self-hosted Rekor on [container]. Recorded as out-of-scope (revisit point 3).
	+	- No self-hosted Rekor for the reference deployment. Recorded as out-of-scope (revisit point 3).
		- No removal of legacy `oversight_core/tlog.py`. It stays as fallback verifier.
		- No Hardware KeyProvider work - that's v0.6 alongside format adapters.
		- No new entry-type negotiation with Sigstore. We use vanilla DSSE.

oversight_core/decoy.py +6 -4

		@@ -15,11 +15,13 @@ realism. Recent research (SPADE 2025, HoneyGPT) shows this is an open area
		with no strong commercial shipment.

		Backend options (pick via `backend` arg or OVERSIGHT_DECOY_BACKEND env):
	-	- "ollama" - POST to a local Ollama server (recommended; uses GPU node)
	+	- "ollama" - POST to a local Ollama server (recommended)
		- "openai" - OpenAI-compatible API (for testing)
		- "static" - hardcoded templates (works offline; lowest quality)

	-	Point OLLAMA_URL at any Ollama instance; default is loopback.
	+	Override the Ollama endpoint and model with the ``OLLAMA_URL`` and
	+	``OVERSIGHT_DECOY_MODEL`` environment variables. Defaults target a
	+	loopback Ollama install.
		"""

		from __future__ import annotations
		@@ -33,8 +35,8 @@ from typing import Optional
		import httpx


	-	DEFAULT_OLLAMA = os.environ.get("OLLAMA_URL", "http://[redacted-rfc1918]")
	-	DEFAULT_MODEL = os.environ.get("OVERSIGHT_DECOY_MODEL", "dolphin-mistral:7b-v2.8")
	+	DEFAULT_OLLAMA = os.environ.get("OLLAMA_URL", "http://127.0.0.1:11434")
	+	DEFAULT_MODEL = os.environ.get("OVERSIGHT_DECOY_MODEL", "llama3.1:8b")


		# Realistic decoy filenames. These are deliberately interesting to an attacker