Fix OpenRouter Kimi K2.6 developer role

closes #5159

packages/ai/CHANGELOG.md

@@ -5,9 +5,16 @@
 ### Breaking Changes
 
 - Changed direct provider stream functions to require explicit `options.apiKey`; top-level `stream*`/`complete*` helpers still resolve built-in environment auth.
+- Changed top-level stream helpers to delegate to `defaultModels()`, so setup failures are reported through stream error results consistently.
+- Moved the built-in API provider registration helpers from the `providers/register-builtins` subpath to `api-registry`/the root export.
+
+### Added
+
+- Added `Models`/`MutableModels`, `builtinModels()`, and `defaultModels()` for isolated or process-default model metadata, stream, and auth resolution.
 
 ### Fixed
 
+- Fixed OpenRouter Moonshot Kimi K2.6 requests to use `system` instead of unsupported `developer` messages ([#5159](https://github.com/earendil-works/pi/issues/5159)).
 - Fixed OpenCode Go Kimi K2.6 thinking requests to send `thinking` objects instead of invalid string values, and fixed OpenCode Zen Grok Build thinking requests to omit unsupported `reasoning_effort` ([#5169](https://github.com/earendil-works/pi/issues/5169)).
 
 ## [0.77.0] - 2026-05-28

packages/ai/scripts/generate-models.ts

@@ -1235,6 +1235,9 @@ async function generateModels() {
 			candidate.cost.cacheRead = 0.07;
 			candidate.maxTokens = 4096;
 		}
+		if (candidate.provider === "openrouter" && candidate.id.startsWith("moonshotai/kimi-k2.6")) {
+			candidate.compat = { ...candidate.compat, supportsDeveloperRole: false };
+		}
 		if (candidate.provider === "openrouter" && candidate.id === "z-ai/glm-5") {
 			candidate.cost.input = 0.6;
 			candidate.cost.output = 1.9;

packages/ai/src/models.generated.ts

@@ -9947,6 +9947,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "openrouter",
 			baseUrl: "https://openrouter.ai/api/v1",
+			compat: {"supportsDeveloperRole":false},
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {
@@ -9964,6 +9965,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "openrouter",
 			baseUrl: "https://openrouter.ai/api/v1",
+			compat: {"supportsDeveloperRole":false},
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {