Part of the governance and compliance evidence epic. Related: stamping the policy version on redaction evidence.
Work lands in: philterd/phinder.
Why
"One policy, enforced identically everywhere" only holds if every enforcement point runs the same engine. Today Phinder lags Philter's Phileas version, so the same policy could behave differently between them. This is hygiene, but it is a credibility risk for the consistency claim.
Current state
Embedded-engine versions diverge:
| Enforcement point |
Phileas version |
| Philter |
4.0.0-SNAPSHOT |
| Phinder |
3.3.0 |
Philter AI Proxy and Philter MCP delegate to Philter over REST, so they stay consistent automatically.
What to build
Acceptance criteria
The same policy JSON produces the same redaction across Philter and Phinder, verified by a shared test fixture.
Size
About 1 week. Mechanical bump plus testing.
Part of the governance and compliance evidence epic. Related: stamping the policy version on redaction evidence.
Work lands in:
philterd/phinder.Why
"One policy, enforced identically everywhere" only holds if every enforcement point runs the same engine. Today Phinder lags Philter's Phileas version, so the same policy could behave differently between them. This is hygiene, but it is a credibility risk for the consistency claim.
Current state
Embedded-engine versions diverge:
Philter AI Proxy and Philter MCP delegate to Philter over REST, so they stay consistent automatically.
What to build
phinderto match Philter's release, and test it against the current policy schema.Acceptance criteria
The same policy JSON produces the same redaction across Philter and Phinder, verified by a shared test fixture.
Size
About 1 week. Mechanical bump plus testing.