@oh-my-pi/pi-coding-agent 12.6.0 → 12.7.1

package/src/modes/components/settings-defs.ts

@@ -146,10 +146,16 @@ const OPTION_PROVIDERS: Partial<Record<SettingPath, OptionProvider>> = {
 	],
 	// Provider options
 	"providers.webSearch": [
-		{ value: "auto", label: "Auto", description: "Priority: Exa > Perplexity > Anthropic" },
+		{
+			value: "auto",
+			label: "Auto",
+			description: "Priority: Exa > Jina > Perplexity > Anthropic > Gemini > Codex > Z.AI",
+		},
 		{ value: "exa", label: "Exa", description: "Requires EXA_API_KEY" },
+		{ value: "jina", label: "Jina", description: "Requires JINA_API_KEY" },
 		{ value: "perplexity", label: "Perplexity", description: "Requires PERPLEXITY_API_KEY" },
 		{ value: "anthropic", label: "Anthropic", description: "Uses Anthropic web search" },
+		{ value: "zai", label: "Z.AI", description: "Calls Z.AI webSearchPrime MCP" },
 	],
 	"providers.image": [
 		{ value: "auto", label: "Auto", description: "Priority: OpenRouter > Gemini" },

package/src/modes/controllers/extension-ui-controller.ts

@@ -9,6 +9,7 @@ import type {
 	ExtensionError,
 	ExtensionUIContext,
 	ExtensionUIDialogOptions,
+	TerminalInputHandler,
 } from "../../extensibility/extensions";
 import { HookEditorComponent } from "../../modes/components/hook-editor";
 import { HookInputComponent } from "../../modes/components/hook-input";
@@ -20,6 +21,7 @@ import { setTerminalTitle } from "../../utils/title-generator";
 export class ExtensionUiController {
 	#hookSelectorOverlay: OverlayHandle | undefined;
 	#hookInputOverlay: OverlayHandle | undefined;
+	#extensionTerminalInputUnsubscribers = new Set<() => void>();
 
 	readonly #dialogOverlayOptions = {
 		anchor: "bottom-center",
@@ -41,6 +43,7 @@ export class ExtensionUiController {
 			confirm: (title, message, _dialogOptions) => this.showHookConfirm(title, message),
 			input: (title, placeholder, _dialogOptions) => this.showHookInput(title, placeholder),
 			notify: (message, type) => this.showHookNotify(message, type),
+			onTerminalInput: handler => this.addExtensionTerminalInputListener(handler),
 			setStatus: (key, text) => this.setHookStatus(key, text),
 			setWorkingMessage: message => this.ctx.setWorkingMessage(message),
 			setWidget: (key, content) => this.setHookWidget(key, content),
@@ -157,6 +160,7 @@ export class ExtensionUiController {
 				this.ctx.statusContainer.clear();
 
 				// Create new session
+				this.clearExtensionTerminalInputListeners();
 				const success = await this.ctx.session.newSession({ parentSession: options?.parentSession });
 				if (!success) {
 					return { cancelled: true };
@@ -351,6 +355,7 @@ export class ExtensionUiController {
 				this.ctx.statusContainer.clear();
 
 				// Create new session
+				this.clearExtensionTerminalInputListeners();
 				const success = await this.ctx.session.newSession({ parentSession: options?.parentSession });
 				if (!success) {
 					return { cancelled: true };
@@ -450,6 +455,7 @@ export class ExtensionUiController {
 			confirm: async (_title: string, _message: string, _dialogOptions) => false,
 			input: async (_title: string, _placeholder?: string, _dialogOptions?: unknown) => undefined,
 			notify: () => {},
+			onTerminalInput: () => () => {},
 			setStatus: () => {},
 			setWorkingMessage: () => {},
 			setWidget: () => {},
@@ -726,6 +732,22 @@ export class ExtensionUiController {
 	/**
 	 * Show an extension error in the UI.
 	 */
+	addExtensionTerminalInputListener(handler: TerminalInputHandler): () => void {
+		const unsubscribe = this.ctx.ui.addInputListener(handler);
+		this.#extensionTerminalInputUnsubscribers.add(unsubscribe);
+		return () => {
+			unsubscribe();
+			this.#extensionTerminalInputUnsubscribers.delete(unsubscribe);
+		};
+	}
+
+	clearExtensionTerminalInputListeners(): void {
+		for (const unsubscribe of this.#extensionTerminalInputUnsubscribers) {
+			unsubscribe();
+		}
+		this.#extensionTerminalInputUnsubscribers.clear();
+	}
+
 	showExtensionError(extensionPath: string, error: string): void {
 		const errorText = new Text(theme.fg("error", `Extension "${extensionPath}" error: ${error}`), 1, 0);
 		this.ctx.chatContainer.addChild(errorText);

package/src/modes/controllers/selector-controller.ts

@@ -279,7 +279,9 @@ export class SelectorController {
 
 			// Provider settings - update runtime preferences
 			case "webSearchProvider":
-				setPreferredSearchProvider(value as "auto" | "exa" | "perplexity" | "anthropic");
+				setPreferredSearchProvider(
+					value as "auto" | "exa" | "jina" | "zai" | "perplexity" | "anthropic" | "gemini" | "codex",
+				);
 				break;
 			case "imageProvider":
 				setPreferredImageProvider(value as "auto" | "gemini" | "openrouter");

package/src/modes/interactive-mode.ts

@@ -720,6 +720,7 @@ export class InteractiveMode implements InteractiveModeContext {
 			this.#sttController.dispose();
 			this.#sttController = undefined;
 		}
+		this.#extensionUiController.clearExtensionTerminalInputListeners();
 		this.statusLine.dispose();
 		if (this.unsubscribe) {
 			this.unsubscribe();
@@ -919,6 +920,7 @@ export class InteractiveMode implements InteractiveModeContext {
 	}
 
 	handleClearCommand(): Promise<void> {
+		this.#extensionUiController.clearExtensionTerminalInputListeners();
 		return this.#commandController.handleClearCommand();
 	}
 

package/src/modes/rpc/rpc-mode.ts

@@ -164,6 +164,11 @@ export async function runRpcMode(session: AgentSession): Promise<never> {
 			);
 		}
 
+		onTerminalInput(): () => void {
+			// Raw terminal input not supported in RPC mode
+			return () => {};
+		}
+
 		notify(message: string, type?: "info" | "warning" | "error"): void {
 			// Fire and forget - no response needed
 			this.output({

package/src/prompts/system/system-prompt.md

@@ -110,31 +110,43 @@ Don't open a file hoping. Hope is not a strategy.
 
 <procedure>
 ## Execution
-**Step 0 — CHECKPOINT** (mandatory for multi-step, multi-file, or ambiguous tasks):
-- Distinct work streams? Dependencies between them?
-{{#has tools "task"}}
-- Parallelizable via Task tool, or necessarily sequential?
-{{/has}}
-{{#if skills.length}}
-- Skill matches domain? Read it first.
-{{/if}}
-{{#if rules.length}}
-- Applicable rule? Read it first.
-{{/if}}
-- Skip only when: single-file, ≤3 edits, requirements explicit.
-**Plan** if task has weight: 3–7 bullets, no more.
-**After each tool call**: interpret result → decide next action → execute. No echoing output.
-**If blocked**: exhaust tools/context/files first. Only then ask — minimum viable question.
-**If requested change includes refactor**: remove now-unused elements. Note removals.
+
+**Assess scope first.** 
+{{#if skills.length}}- If a skill matches the domain, read it before starting.{{/if}}
+{{#if rules.length}}- If an applicable rule exists, read it before starting.{{/if}}
+{{#has tools "task"}}- Consider if the task is parallelizable via Task tool? Make a conflict-free plan to delegate to subagents if possible.{{/has}}
+- If the task is multi-file or ambiguous, write a 3–7 bullet plan.
+
+**Do the work.**
+Every turn must include at least one tool call that advances the deliverable (edit, write, run, delegate, etc.). Planning and tracking do not count.
+
+**After each tool call**: 
+- Interpret result
+- Decide next action
+- Execute. No echoing output.
+
+**If blocked**: 
+- Exhaust tools/context/files first.
+- Only then ask — minimum viable question.
+
+**If requested change includes refactor**: 
+Cleanup dead code and unused elements, do not yield before the codebase is pristine.
 
 {{#has tools "todo_write"}}
 ### Task Tracking
-- Use `todo_write` proactively for non-trivial, multi-step work so progress stays visible.
-- Initialize todos before implementation for complex tasks, then keep them current while working.
-- Mark todo items complete immediately after finishing them; do not batch completion updates.
-- Keep todo items as focused logical units (one coherent outcome per item); split broad work into smaller items.
-- Keep exactly one item `in_progress` at a time and complete in order unless requirements change.
-- Skip `todo_write` for single trivial or purely informational requests.
+- Never create a todo list and then stop. A turn that contains only todo updates is a failed turn.
+- Use todos as you make progress to make multi-step progress visible, don't batch.
+- Skip entirely for single-step or trivial requests.
+{{/has}}
+
+{{#has tools "task"}}
+### Parallel Execution
+Use the Task tool when work genuinely forks into independent streams:
+- Editing 4+ files with no dependencies between edits
+- Investigating 2+ independent subsystems
+- Work that decomposes into pieces not needing each other's results
+
+Task tool is for **parallel execution**, not deferred execution. If you can do it now, do it now. Sequential is fine when steps depend on each other — don't parallelize for its own sake.
 {{/has}}
 
 ### Verification
@@ -244,10 +256,10 @@ Sequential work requires justification. If you cannot articulate why B depends o
 {{/has}}
 
 <output_style>
-- No explanatory scaffolding. No summary closings ("In summary…").
-- No filler. No emojis. No ceremony.
-- User execution-mode instructions (do-it-yourself vs delegate) override tool-use defaults.
+- State intent before tool calls in one sentence.
+- No summary closings ("In summary…"). No filler. No emojis. No ceremony.
 - Suppress: "genuinely", "honestly", "straightforward".
+- User execution-mode instructions (do-it-yourself vs delegate) override tool-use defaults.
 - Requirements conflict or are unclear → ask only after exhausting exploration.
 </output_style>
 
@@ -256,6 +268,8 @@ Complete the full request before yielding. Use tools for verifiable facts. Resul
 
 You have unlimited stamina; the user does not. Persist on hard problems. Don't burn their energy on problems you failed to think through.
 
+This matters. Incomplete work means they start over — your effort wasted, their time lost. The person waiting deserves your best work.
+
 Tests you didn't write: bugs shipped. Assumptions you didn't state: incidents to debug. Edge cases you didn't name: pages at 3am.
 
 Write what you can defend.
@@ -270,9 +284,15 @@ The person waiting deserves to receive it.
 User works in a high-reliability industry—defense, finance, healthcare, infrastructure—where bugs have material impact on people's lives, even death.
 </stakes>
 
+<prime_directive>
+**GET THE WORK DONE.** 
+Everything else is subordinate to producing the requested output. If you find yourself emitting metadata (todos, plans, status updates) without having made a single edit or produced a single artifact, you have failed.
+</prime_directive>
+
 <critical>
 Keep going until finished.
-- Quote only needed; rest noise.
+- Every turn must advance the deliverable. Metadata-only turns are failures.
+- Quote only what's needed; rest is noise.
 - Don't claim unverified correctness.
 - Do not ask when it may be obtained from available tools or repo context/files.
 - Touch only requested; no incidental refactors/cleanup.

package/src/session/agent-session.ts

@@ -51,10 +51,16 @@ import type {
 	ExtensionCommandContext,
 	ExtensionRunner,
 	ExtensionUIContext,
+	MessageEndEvent,
+	MessageStartEvent,
+	MessageUpdateEvent,
 	SessionBeforeBranchResult,
 	SessionBeforeCompactResult,
 	SessionBeforeSwitchResult,
 	SessionBeforeTreeResult,
+	ToolExecutionEndEvent,
+	ToolExecutionStartEvent,
+	ToolExecutionUpdateEvent,
 	TreePreparation,
 	TurnEndEvent,
 	TurnStartEvent,
@@ -102,6 +108,7 @@ import {
 	pythonExecutionToText,
 } from "./messages";
 import type { BranchSummaryEntry, CompactionEntry, NewSessionOptions, SessionManager } from "./session-manager";
+import { getLatestCompactionEntry } from "./session-manager";
 
 /** Session-specific events that extend the core AgentEvent */
 export type AgentSessionEvent =
@@ -226,6 +233,7 @@ const noOpUIContext: ExtensionUIContext = {
 	confirm: async (_title, _message, _dialogOptions) => false,
 	input: async (_title, _placeholder, _dialogOptions) => undefined,
 	notify: () => {},
+	onTerminalInput: () => () => {},
 	setStatus: () => {},
 	setWorkingMessage: () => {},
 	setWidget: () => {},
@@ -874,6 +882,51 @@ export class AgentSession {
 			};
 			await this.#extensionRunner.emit(hookEvent);
 			this.#turnIndex++;
+		} else if (event.type === "message_start") {
+			const extensionEvent: MessageStartEvent = {
+				type: "message_start",
+				message: event.message,
+			};
+			await this.#extensionRunner.emit(extensionEvent);
+		} else if (event.type === "message_update") {
+			const extensionEvent: MessageUpdateEvent = {
+				type: "message_update",
+				message: event.message,
+				assistantMessageEvent: event.assistantMessageEvent,
+			};
+			await this.#extensionRunner.emit(extensionEvent);
+		} else if (event.type === "message_end") {
+			const extensionEvent: MessageEndEvent = {
+				type: "message_end",
+				message: event.message,
+			};
+			await this.#extensionRunner.emit(extensionEvent);
+		} else if (event.type === "tool_execution_start") {
+			const extensionEvent: ToolExecutionStartEvent = {
+				type: "tool_execution_start",
+				toolCallId: event.toolCallId,
+				toolName: event.toolName,
+				args: event.args,
+			};
+			await this.#extensionRunner.emit(extensionEvent);
+		} else if (event.type === "tool_execution_update") {
+			const extensionEvent: ToolExecutionUpdateEvent = {
+				type: "tool_execution_update",
+				toolCallId: event.toolCallId,
+				toolName: event.toolName,
+				args: event.args,
+				partialResult: event.partialResult,
+			};
+			await this.#extensionRunner.emit(extensionEvent);
+		} else if (event.type === "tool_execution_end") {
+			const extensionEvent: ToolExecutionEndEvent = {
+				type: "tool_execution_end",
+				toolCallId: event.toolCallId,
+				toolName: event.toolName,
+				result: event.result,
+				isError: event.isError ?? false,
+			};
+			await this.#extensionRunner.emit(extensionEvent);
 		} else if (event.type === "auto_compaction_start") {
 			await this.#extensionRunner.emit({ type: "auto_compaction_start", reason: event.reason });
 		} else if (event.type === "auto_compaction_end") {
@@ -2664,9 +2717,9 @@ Be thorough - include exact file paths, function names, error messages, and tech
 		// The error shouldn't trigger another compaction since we already compacted.
 		// Example: opus fails → switch to codex → compact → switch back to opus → opus error
 		// is still in context but shouldn't trigger compaction again.
-		const compactionEntry = this.sessionManager.getBranch().find(e => e.type === "compaction");
+		const compactionEntry = getLatestCompactionEntry(this.sessionManager.getBranch());
 		const errorIsFromBeforeCompaction =
-			compactionEntry && assistantMessage.timestamp < new Date(compactionEntry.timestamp).getTime();
+			compactionEntry !== null && assistantMessage.timestamp < new Date(compactionEntry.timestamp).getTime();
 
 		// Case 1: Overflow - LLM returned context overflow error
 		if (sameModel && !errorIsFromBeforeCompaction && isContextOverflow(assistantMessage, contextWindow)) {
@@ -3959,6 +4012,35 @@ Be thorough - include exact file paths, function names, error messages, and tech
 		const contextWindow = model.contextWindow ?? 0;
 		if (contextWindow <= 0) return undefined;
 
+		// After compaction, the last assistant usage reflects pre-compaction context size.
+		// We can only trust usage from an assistant that responded after the latest compaction.
+		// If no such assistant exists, context token count is unknown until the next LLM response.
+		const branchEntries = this.sessionManager.getBranch();
+		const latestCompaction = getLatestCompactionEntry(branchEntries);
+
+		if (latestCompaction) {
+			// Check if there's a valid assistant usage after the compaction boundary
+			const compactionIndex = branchEntries.lastIndexOf(latestCompaction);
+			let hasPostCompactionUsage = false;
+			for (let i = branchEntries.length - 1; i > compactionIndex; i--) {
+				const entry = branchEntries[i];
+				if (entry.type === "message" && entry.message.role === "assistant") {
+					const assistant = entry.message;
+					if (assistant.stopReason !== "aborted" && assistant.stopReason !== "error") {
+						const contextTokens = calculateContextTokens(assistant.usage);
+						if (contextTokens > 0) {
+							hasPostCompactionUsage = true;
+						}
+						break;
+					}
+				}
+			}
+
+			if (!hasPostCompactionUsage) {
+				return { tokens: null, contextWindow, percent: null };
+			}
+		}
+
 		const estimate = this.#estimateContextTokens();
 		const percent = (estimate.tokens / contextWindow) * 100;
 
@@ -3966,9 +4048,6 @@ Be thorough - include exact file paths, function names, error messages, and tech
 			tokens: estimate.tokens,
 			contextWindow,
 			percent,
-			usageTokens: estimate.usageTokens,
-			trailingTokens: estimate.trailingTokens,
-			lastUsageIndex: estimate.lastUsageIndex,
 		};
 	}
 
@@ -3985,9 +4064,6 @@ Be thorough - include exact file paths, function names, error messages, and tech
 	 */
 	#estimateContextTokens(): {
 		tokens: number;
-		usageTokens: number;
-		trailingTokens: number;
-		lastUsageIndex: number | null;
 	} {
 		const messages = this.messages;
 
@@ -4014,9 +4090,6 @@ Be thorough - include exact file paths, function names, error messages, and tech
 			}
 			return {
 				tokens: estimated,
-				usageTokens: 0,
-				trailingTokens: estimated,
-				lastUsageIndex: null,
 			};
 		}
 
@@ -4028,9 +4101,6 @@ Be thorough - include exact file paths, function names, error messages, and tech
 
 		return {
 			tokens: usageTokens + trailingTokens,
-			usageTokens,
-			trailingTokens,
-			lastUsageIndex,
 		};
 	}
 

package/src/task/render.ts

@@ -684,7 +684,7 @@ function renderFindings(
 		const findingContinue = isLastFinding ? "   " : `${theme.tree.vertical}  `;
 
 		const { color } = getPriorityInfo(finding.priority);
-		const titleText = finding.title.replace(/^\[P\d\]\s*/, "");
+		const titleText = finding.title?.replace(/^\[P\d\]\s*/, "") ?? "Untitled";
 		const loc = `${path.basename(finding.file_path)}:${finding.line_start}`;
 
 		lines.push(

package/src/web/search/index.ts

@@ -1,7 +1,7 @@
 /**
  * Unified Web Search Tool
  *
- * Single tool supporting Anthropic, Perplexity, Exa, Jina, Gemini, and Codex
+ * Single tool supporting Anthropic, Perplexity, Exa, Jina, Gemini, Codex, and Z.AI
  * providers with provider-specific parameters exposed conditionally.
  *
  * When EXA_API_KEY is available, additional specialized tools are exposed:
@@ -33,7 +33,7 @@ import { SearchProviderError } from "./types";
 export const webSearchSchema = Type.Object({
 	query: Type.String({ description: "Search query" }),
 	provider: Type.Optional(
-		StringEnum(["auto", "exa", "jina", "anthropic", "perplexity", "gemini", "codex"], {
+		StringEnum(["auto", "exa", "jina", "zai", "anthropic", "perplexity", "gemini", "codex"], {
 			description: "Search provider (default: auto)",
 		}),
 	),
@@ -47,7 +47,7 @@ export const webSearchSchema = Type.Object({
 
 export type SearchParams = {
 	query: string;
-	provider?: "auto" | "exa" | "jina" | "anthropic" | "perplexity" | "gemini" | "codex";
+	provider?: "auto" | "exa" | "jina" | "zai" | "anthropic" | "perplexity" | "gemini" | "codex";
 	recency?: "day" | "week" | "month" | "year";
 	limit?: number;
 	/** Maximum output tokens. Defaults to 4096. */
@@ -56,6 +56,8 @@ export type SearchParams = {
 	temperature?: number;
 	/** Number of search results to retrieve. Defaults to 10. */
 	num_search_results?: number;
+	/** Disable provider fallback when explicit provider is selected (CLI/debug use). */
+	no_fallback?: boolean;
 };
 
 function formatProviderList(providers: SearchProvider[]): string {
@@ -68,6 +70,9 @@ function formatProviderError(error: unknown, provider: SearchProvider): string {
 			return "Anthropic web search returned 404 (model or endpoint not found).";
 		}
 		if (error.status === 401 || error.status === 403) {
+			if (error.provider === "zai") {
+				return error.message;
+			}
 			return `${getSearchProvider(error.provider).label} authorization failed (${error.status}). Check API key or base URL.`;
 		}
 		return error.message;
@@ -165,7 +170,12 @@ async function executeSearch(
 	_toolCallId: string,
 	params: SearchParams,
 ): Promise<{ content: Array<{ type: "text"; text: string }>; details: SearchRenderDetails }> {
-	const providers = await resolveProviderChain(params.provider);
+	const providers =
+		params.provider && params.provider !== "auto" && params.no_fallback
+			? (await getSearchProvider(params.provider).isAvailable())
+				? [getSearchProvider(params.provider)]
+				: []
+			: await resolveProviderChain(params.provider);
 
 	if (providers.length === 0) {
 		const message = "No web search provider configured.";
@@ -226,7 +236,7 @@ export async function runSearchQuery(
 /**
  * Web search tool implementation.
  *
- * Supports Anthropic, Perplexity, Exa, Jina, Gemini, and Codex providers with automatic fallback.
+ * Supports Anthropic, Perplexity, Exa, Jina, Gemini, Codex, and Z.AI providers with automatic fallback.
  * Session is accepted for interface consistency but not used.
  */
 export class SearchTool implements AgentTool<typeof webSearchSchema, SearchRenderDetails> {

package/src/web/search/provider.ts

@@ -5,6 +5,7 @@ import { ExaProvider } from "./providers/exa";
 import { GeminiProvider } from "./providers/gemini";
 import { JinaProvider } from "./providers/jina";
 import { PerplexityProvider } from "./providers/perplexity";
+import { ZaiProvider } from "./providers/zai";
 import type { SearchProviderId } from "./types";
 
 export type { SearchParams } from "./providers/base";
@@ -14,12 +15,13 @@ const SEARCH_PROVIDERS: Record<SearchProviderId, SearchProvider> = {
 	exa: new ExaProvider(),
 	jina: new JinaProvider(),
 	perplexity: new PerplexityProvider(),
+	zai: new ZaiProvider(),
 	anthropic: new AnthropicProvider(),
 	gemini: new GeminiProvider(),
 	codex: new CodexProvider(),
 } as const;
 
-const SEARCH_PROVIDER_ORDER: SearchProviderId[] = ["exa", "jina", "perplexity", "anthropic", "gemini", "codex"];
+const SEARCH_PROVIDER_ORDER: SearchProviderId[] = ["exa", "jina", "perplexity", "anthropic", "gemini", "codex", "zai"];
 
 export function getSearchProvider(provider: SearchProviderId): SearchProvider {
 	return SEARCH_PROVIDERS[provider];