Skip to content

graff — a fast agentic coding harness in Zig: multi-provider, MCP, workflows, DGM evolution loop, TS/Python SDKs

Python

Document optional formal learning checks

Rach Pradhan66d64f09a084
diff --git a/CHANGELOG.md b/CHANGELOG.md
index cd01b014..92478a47 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -18,6 +18,12 @@ current is part of cutting a release.
- WebSocket prewarm sends the selected model in its first frame and records a bounded rejection reason when the server refuses it. A rejected prewarm can still fall back through the existing transport recovery path.
- WebSocket turns deliver response deltas as they arrive, including before a slow completion frame, and flush held text once at completion. Mock transport checks cover live output, retry, fallback, and duplicate-output prevention.
+### Optional formal evidence for local learning
+
+- Bounded lifecycle models cover effort changes, asynchronous tools, permission ownership, and connection leases. Deliberately weakened configurations provide counterexamples for the invariants those models are meant to guard.
+- Prompt evolution and local learning can require pinned formal-check evidence before scoring, checkpoint restore, or promotion. Drifted inputs or missing evidence stop that optional path; ordinary learning keeps its prior defaults.
+- Formal evidence remains separate from held-out task evaluation: a model check cannot make an incorrect candidate pass, and these finite models do not claim proof that the implementation matches them.
+
## v0.0.302.5
- Dry-run route checks preserve an explicitly qualified selection while checking whether it is available. This keeps the reported route aligned with the route the next request would use.