feat: pass modelId to context engine assemble()

Context engine plugins can now receive the current model identifier
in assemble() params, enabling model-aware context formatting.

For example, a plugin could:
- Use XML formatting for Claude models
- Use terse/compressed context for small local models (7B)
- Adjust token budgets based on model context window size

The model param is optional and backward-compatible — existing
plugins that don't use it are unaffected.

src/agents/pi-embedded-runner/run/attempt.ts

@@ -2167,6 +2167,7 @@ export async function runEmbeddedAttempt(
               sessionKey: params.sessionKey,
               messages: activeSession.messages,
               tokenBudget: params.contextTokenBudget,
+              model: params.modelId,
             });
             if (assembled.messages !== activeSession.messages) {
               activeSession.agent.replaceMessages(assembled.messages);

src/context-engine/legacy.ts

@@ -40,6 +40,7 @@ export class LegacyContextEngine implements ContextEngine {
     sessionKey?: string;
     messages: AgentMessage[];
     tokenBudget?: number;
+    model?: string;
   }): Promise<AssembleResult> {
     // Pass-through: the existing sanitize -> validate -> limit -> repair pipeline
     // in attempt.ts handles context assembly for the legacy engine.

src/context-engine/types.ts

@@ -131,6 +131,9 @@ export interface ContextEngine {
     sessionKey?: string;
     messages: AgentMessage[];
     tokenBudget?: number;
+    /** Current model identifier (e.g. "claude-opus-4", "gpt-4o", "qwen2.5-7b").
+     *  Allows context engine plugins to adapt formatting per model. */
+    model?: string;
   }): Promise<AssembleResult>;
 
   /**