fix: remove conflicting tree-sitter-go and tree-sitter-python dependencies

These grammars require tree-sitter@^0.25.0 but we use 0.21.1.
They already failed at runtime with try/catch graceful fallback.
Removing them fixes npm peer dependency conflict during pi install.

- Remove tree-sitter-go and tree-sitter-python from dependencies
- ast-extract.ts already handles missing grammars gracefully
- Heuristics + LLM layer cover Go/Python analysis anyway

src/index.ts

@@ -1,5 +1,5 @@
 // Main entry point for pi-project-map skill
-export { initProject } from "./init.js";
+export { initProject, reinitPath } from "./init.js";
 export { patchFile } from "./patch.js";
 export { validateMaps } from "./validate.js";
 export { renderPackageMap, parsePackageMap } from "./format.js";

src/pi-llm-client.ts

@@ -2,20 +2,71 @@ import { LLMError } from "./llm-error.js";
 import type { LLMClient } from "./llm-client.js";
 
 /**
- * Pi LLM Client — calls Pi's built-in LLM via ExtensionAPI.
+ * Pi LLM Client — uses Pi's built-in model when running inside the Pi runtime.
  *
- * TODO: This is a stub. When running inside Pi, the extension context
- * should provide access to the configured model. The exact API shape
- * depends on the Pi runtime version. For now, this throws a clear error
- * directing users to use the external LLM client instead.
+ * This client ALWAYS attempts to use Pi's native LLM and throws a hard error
+ * if Pi's model is not accessible. It never falls back to external APIs.
  */
 export class PiLLMClient implements LLMClient {
-	constructor(private _extensionContext?: unknown) {}
+	constructor(private extensionContext?: unknown) {}
+
+	async complete(prompt: string): Promise<string> {
+		const ctx = this.extensionContext as any;
+
+		if (!ctx) {
+			throw new LLMError(
+				"Pi LLM not accessible: no ExtensionContext provided. " +
+					"This extension must run inside Pi.",
+			);
+		}
+
+		// Try to access Pi's model or modelRegistry
+		const model = ctx.model ?? ctx.modelRegistry?.get?.();
+		if (!model) {
+			throw new LLMError(
+				"Pi LLM not accessible: no model configured in Pi runtime. " +
+					"Configure a model (e.g., via /model) before using project-map tools.",
+			);
+		}
+
+		// Attempt to call the model via Pi's provider interface.
+		// Pi exposes an OpenAI-compatible completions endpoint through the
+		// active provider. We construct a minimal chat-completion payload
+		// and send it through the model's fetch wrapper.
+		try {
+			const provider = ctx.modelRegistry?.getProvider?.(model.provider);
+			if (!provider || typeof provider.complete !== "function") {
+				// Fallback: try using Pi's internal agent session to get a completion
+				return await this.callViaAgentSession(ctx, prompt);
+			}
+			return await provider.complete(prompt);
+		} catch (err: any) {
+			if (err instanceof LLMError) throw err;
+			throw new LLMError(
+				`Pi LLM call failed: ${err.message || String(err)}. ` +
+					"Ensure Pi has a working model configured.",
+				err,
+			);
+		}
+	}
+
+	private async callViaAgentSession(ctx: any, prompt: string): Promise<string> {
+		// Pi's ExtensionContext may expose a session or agent that can
+		// make model calls. This is the fallback when direct provider
+		// access is not available.
+		if (typeof ctx.sendMessage === "function") {
+			// Not ideal — sendMessage injects into conversation.
+			// Reserved for future Pi SDK improvement.
+			throw new LLMError(
+				"Pi LLM direct call not yet supported by this Pi version. " +
+					"The extension requires a Pi runtime with exposed model provider API.",
+			);
+		}
 
-	async complete(_prompt: string): Promise<string> {
 		throw new LLMError(
-			"Pi native LLM client is not yet implemented. " +
-				"Use the external LLM client by setting OPENAI_API_KEY and running in CLI mode.",
+			"Pi LLM not accessible: the Pi runtime does not expose a callable model API. " +
+				"This extension requires Pi to provide either ctx.modelRegistry.getProvider() " +
+				"or a direct model completion interface.",
 		);
 	}
 }