@oh-my-pi/pi-coding-agent 15.5.7 → 15.5.8

package/src/main.ts

@@ -9,7 +9,6 @@ import * as fs from "node:fs/promises";
 import * as os from "node:os";
 import * as path from "node:path";
 import { createInterface } from "node:readline/promises";
-import { keepaliveWhile } from "@oh-my-pi/pi-agent-core";
 import type { ImageContent } from "@oh-my-pi/pi-ai";
 import {
 	$env,
@@ -316,7 +315,7 @@ async function runInteractiveMode(
 	}
 
 	while (true) {
-		const input = await keepaliveWhile(mode.getUserInput());
+		const input = await mode.getUserInput();
 		await submitInteractiveInput(mode, session, input);
 	}
 }

package/src/mcp/transports/http.ts

@@ -16,8 +16,23 @@ import type {
 	MCPTransport,
 } from "../../mcp/types";
 import { toJsonRpcError } from "../../mcp/types";
-import { createMCPTimeout, getNeverAbortSignal, resolveMCPTimeoutMs } from "../timeout";
+import { createMCPTimeout, getNeverAbortSignal, isMCPTimeoutEnabled, resolveMCPTimeoutMs } from "../timeout";
 
+const HTTP_SSE_CONNECT_TIMEOUT_MS = 1_000;
+/**
+ * Best-effort startup deadline for the optional Streamable HTTP GET SSE listener.
+ *
+ * Returns `0` (disabled) when the operator has explicitly disabled MCP client-side
+ * timeouts via `timeout: 0` or `OMP_MCP_TIMEOUT_MS=0`, mirroring the rest of the
+ * MCP timeout surface. Otherwise caps the wait at one second and scales below
+ * short request timeouts so connect-time never exceeds the request budget.
+ */
+export function resolveSSEConnectTimeoutMs(configTimeout?: number): number {
+	const requestTimeout = resolveMCPTimeoutMs(configTimeout);
+	if (!isMCPTimeoutEnabled(requestTimeout)) return 0;
+	const boundedTimeout = Math.min(HTTP_SSE_CONNECT_TIMEOUT_MS, Math.floor(requestTimeout / 4));
+	return Math.max(1, boundedTimeout);
+}
 /**
  * HTTP transport for MCP servers.
  * Uses POST for requests, supports SSE responses.
@@ -73,6 +88,15 @@ export class HttpTransport implements MCPTransport {
 		}
 
 		let response: Response;
+		let timedOut = false;
+		const startupTimeoutMs = resolveSSEConnectTimeoutMs(this.config.timeout);
+		const timeoutId =
+			startupTimeoutMs > 0
+				? setTimeout(() => {
+						timedOut = true;
+						this.#sseConnection?.abort();
+					}, startupTimeoutMs)
+				: null;
 		try {
 			response = await fetch(this.config.url, {
 				method: "GET",
@@ -81,13 +105,16 @@ export class HttpTransport implements MCPTransport {
 			});
 		} catch (error) {
 			this.#sseConnection = null;
-			if (error instanceof Error && error.name !== "AbortError") {
+			if (error instanceof Error && error.name !== "AbortError" && !timedOut) {
 				this.onError?.(error);
 			}
 			return;
+		} finally {
+			if (timeoutId !== null) clearTimeout(timeoutId);
 		}
 
 		if (response.status === 405 || !response.ok || !response.body) {
+			await response.body?.cancel();
 			this.#sseConnection = null;
 			return;
 		}

package/src/modes/components/tool-execution.ts

@@ -1,3 +1,4 @@
+import type { SnapshotStore } from "@oh-my-pi/hashline";
 import type { AgentTool } from "@oh-my-pi/pi-agent-core";
 import {
 	Box,
@@ -105,10 +106,10 @@ function resolveEditModeForTool(toolName: string, tool: AgentTool | undefined):
 }
 
 export interface ToolExecutionOptions {
+	snapshots?: SnapshotStore;
 	showImages?: boolean; // default: true (only used if terminal supports images)
 	editFuzzyThreshold?: number;
 	editAllowFuzzy?: boolean;
-	hashlineAutoDropPureInsertDuplicates?: boolean;
 }
 
 export interface ToolExecutionHandle {
@@ -142,7 +143,7 @@ export class ToolExecutionComponent extends Container {
 	#showImages: boolean;
 	#editFuzzyThreshold: number | undefined;
 	#editAllowFuzzy: boolean | undefined;
-	#hashlineAutoDropPureInsertDuplicates: boolean | undefined;
+	#snapshots?: SnapshotStore;
 	#isPartial = true;
 	#tool?: AgentTool;
 	#ui: TUI;
@@ -189,7 +190,7 @@ export class ToolExecutionComponent extends Container {
 		this.#showImages = options.showImages ?? true;
 		this.#editFuzzyThreshold = options.editFuzzyThreshold;
 		this.#editAllowFuzzy = options.editAllowFuzzy;
-		this.#hashlineAutoDropPureInsertDuplicates = options.hashlineAutoDropPureInsertDuplicates;
+		this.#snapshots = options.snapshots;
 		this.#tool = tool;
 		this.#ui = ui;
 		this.#cwd = cwd;
@@ -266,12 +267,13 @@ export class ToolExecutionComponent extends Container {
 
 		try {
 			const isStreaming = !this.#argsComplete;
+			if (editMode === "hashline" && !this.#snapshots) return;
 			const previews = await strategy.computeDiffPreview(effectiveArgs, {
 				cwd: this.#cwd,
 				signal: controller.signal,
+				snapshots: this.#snapshots!,
 				fuzzyThreshold: this.#editFuzzyThreshold,
 				allowFuzzy: this.#editAllowFuzzy,
-				hashlineAutoDropPureInsertDuplicates: this.#hashlineAutoDropPureInsertDuplicates,
 				isStreaming,
 			});
 			if (controller.signal.aborted) return;

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

@@ -3,6 +3,7 @@ import { calculatePromptTokens } from "@oh-my-pi/pi-agent-core/compaction/compac
 import type { AssistantMessage, ImageContent } from "@oh-my-pi/pi-ai";
 import { type Component, Loader, TERMINAL, Text } from "@oh-my-pi/pi-tui";
 import { settings } from "../../config/settings";
+import { getFileSnapshotStore } from "../../edit/file-snapshot-store";
 import { AssistantMessageComponent } from "../../modes/components/assistant-message";
 import {
 	ReadToolGroupComponent,
@@ -329,10 +330,10 @@ export class EventController {
 						content.name,
 						renderArgs,
 						{
+							snapshots: getFileSnapshotStore(this.ctx.session),
 							showImages: settings.get("terminal.showImages"),
 							editFuzzyThreshold: settings.get("edit.fuzzyThreshold"),
 							editAllowFuzzy: settings.get("edit.fuzzyMatch"),
-							hashlineAutoDropPureInsertDuplicates: settings.get("edit.hashlineAutoDropPureInsertDuplicates"),
 						},
 						tool,
 						this.ctx.ui,
@@ -444,10 +445,10 @@ export class EventController {
 				event.toolName,
 				event.args,
 				{
+					snapshots: getFileSnapshotStore(this.ctx.session),
 					showImages: settings.get("terminal.showImages"),
 					editFuzzyThreshold: settings.get("edit.fuzzyThreshold"),
 					editAllowFuzzy: settings.get("edit.fuzzyMatch"),
-					hashlineAutoDropPureInsertDuplicates: settings.get("edit.hashlineAutoDropPureInsertDuplicates"),
 				},
 				tool,
 				this.ctx.ui,
@@ -598,7 +599,13 @@ export class EventController {
 		};
 		this.ctx.statusContainer.clear();
 		const reasonText =
-			event.reason === "overflow" ? "Context overflow detected, " : event.reason === "idle" ? "Idle " : "";
+			event.reason === "overflow"
+				? "Context overflow detected, "
+				: event.reason === "incomplete"
+					? "Response incomplete, "
+					: event.reason === "idle"
+						? "Idle "
+						: "";
 		const actionLabel = event.action === "handoff" ? "Auto-handoff" : "Auto context-full maintenance";
 		this.ctx.autoCompactionLoader = new Loader(
 			this.ctx.ui,

package/src/modes/interactive-mode.ts

@@ -4,7 +4,13 @@
  */
 import * as fs from "node:fs/promises";
 import * as path from "node:path";
-import { type Agent, type AgentMessage, type AgentToolResult, ThinkingLevel } from "@oh-my-pi/pi-agent-core";
+import {
+	type Agent,
+	type AgentMessage,
+	type AgentToolResult,
+	EventLoopKeepalive,
+	ThinkingLevel,
+} from "@oh-my-pi/pi-agent-core";
 import type { CompactionOutcome } from "@oh-my-pi/pi-agent-core/compaction";
 import {
 	type AssistantMessage,
@@ -619,7 +625,9 @@ export class InteractiveMode implements InteractiveModeContext {
 		};
 		this.#scheduleLoopAutoSubmit();
 		this.#scheduleGoalContinuation();
-		return promise;
+
+		using _ = new EventLoopKeepalive();
+		return await promise;
 	}
 
 	#scheduleLoopAutoSubmit(): void {

package/src/modes/utils/ui-helpers.ts

@@ -2,6 +2,7 @@ import type { AgentMessage } from "@oh-my-pi/pi-agent-core";
 import type { AssistantMessage, ImageContent, Message } from "@oh-my-pi/pi-ai";
 import { type Component, Spacer, Text, TruncatedText } from "@oh-my-pi/pi-tui";
 import { settings } from "../../config/settings";
+import { getFileSnapshotStore } from "../../edit/file-snapshot-store";
 import { AssistantMessageComponent } from "../../modes/components/assistant-message";
 import { BashExecutionComponent } from "../../modes/components/bash-execution";
 import { BranchSummaryMessageComponent } from "../../modes/components/branch-summary-message";
@@ -377,10 +378,10 @@ export class UiHelpers {
 						content.name,
 						renderArgs,
 						{
+							snapshots: getFileSnapshotStore(this.ctx.session),
 							showImages: settings.get("terminal.showImages"),
 							editFuzzyThreshold: settings.get("edit.fuzzyThreshold"),
 							editAllowFuzzy: settings.get("edit.fuzzyMatch"),
-							hashlineAutoDropPureInsertDuplicates: settings.get("edit.hashlineAutoDropPureInsertDuplicates"),
 						},
 						tool,
 						this.ctx.ui,

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

@@ -59,6 +59,9 @@ With most FS/bash-like tools, static references to them will automatically resol
    - `/<path>`: JSON field extraction
 - `artifact://<id>`: Artifact content
 - `local://<name>.md`: Plan artifacts and shared content with subagents
+{{#if hasObsidian}}
+- `vault://<vault>/<path>`: Obsidian vault content (read/edit). `vault://` lists vaults; `vault://_/…` targets the active vault. File-scoped `?op=outline|backlinks|links|tags|properties|tasks|base|…`; vault-scoped `?op=search&q=…|daily|tasks|orphans|unresolved|bases|…`.
+{{/if}}
 - `mcp://<uri>`: MCP resource
 - `issue://<N>` (or `issue://<owner>/<repo>/<N>`): GitHub issue view; cached on disk so re-reads are free. Bare `issue://` (or `issue://<owner>/<repo>`) lists recent issues; supports `?state=open|closed|all&limit=&author=&label=`.
 - `pr://<N>` (or `pr://<owner>/<repo>/<N>`): GitHub PR view; same cache. Append `?comments=0` to drop the comments section. Bare `pr://` (or `pr://<owner>/<repo>`) lists recent PRs; supports `?state=open|closed|merged|all&limit=&author=&label=`.

package/src/prompts/tools/ast-edit.md

@@ -14,7 +14,7 @@ Performs structural AST-aware rewrites via native ast-grep.
 </instruction>
 
 <output>
-- Replacement summary, per-file replacement counts, and change diffs as `¶src/foo.ts#1a2b`, `-12:before`, `+12:after` lines in hashline mode
+- Replacement summary, per-file replacement counts, and change diffs as `¶src/foo.ts#0a`, `-12:before`, `+12:after` lines in hashline mode
 - Parse issues when files cannot be processed
 </output>
 

package/src/prompts/tools/ast-grep.md

@@ -18,7 +18,7 @@ Performs structural code search using AST matching via native ast-grep.
 
 <output>
 - Grouped matches with file path, byte range, line/column ranges, metavariable captures
-- Match lines are numbered under a file-hash header in hashline mode: `¶src/foo.ts#1a2b`, `*42:content` for the matched line, ` 43:content` for context
+- Match lines are numbered under a file snapshot tag header in hashline mode: `¶src/foo.ts#0a`, `*42:content` for the matched line, ` 43:content` for context
 - Summary counts (`totalMatches`, `filesWithMatches`, `filesSearched`) and parse issues when present
 </output>
 

package/src/prompts/tools/read.md

@@ -8,7 +8,7 @@ Read files, directories, archives, SQLite databases, images, documents, internal
 
 ## Parameters
 
-- `path` — required. Local path, internal URI (`skill://`, `agent://`, `artifact://`, `memory://`, `rule://`, `local://`, `mcp://`), or URL. Append `:<sel>` for line ranges, raw mode, or special modes (e.g. `src/foo.ts:50-200`, `src/foo.ts:raw`, `db.sqlite:users:42`).
+- `path` — required. Local path, internal URI (`skill://`, `agent://`, `artifact://`, `memory://`, `rule://`, `local://`, `vault://`, `mcp://`), or URL. Append `:<sel>` for line ranges, raw mode, or special modes (e.g. `src/foo.ts:50-200`, `src/foo.ts:raw`, `db.sqlite:users:42`).
 
 ## Selectors
 
@@ -28,7 +28,7 @@ Append `:<sel>` to `path`. The bare path falls back to the default mode.
 
 - Reading a directory path returns a depth-limited dirent listing.
 {{#if IS_HL_MODE}}
-- Reading a file with an explicit selector emits a file-hash header and numbered lines: `¶src/foo.ts#1a2b` then `41:def alpha():`. Copy the `¶PATH#HASH` header for anchored edits; ops use bare line numbers. NEVER fabricate the hash.
+- Reading a file with an explicit selector emits a file snapshot tag header and numbered lines: `¶src/foo.ts#0a` then `41:def alpha():`. Copy the `¶PATH#TAG` header for anchored edits; ops use bare line numbers. NEVER fabricate the tag.
 {{else}}
 {{#if IS_LINE_NUMBER_MODE}}
 - Reading a file with an explicit selector returns lines prefixed with line numbers: `41|def alpha():`.
@@ -70,7 +70,7 @@ For `.sqlite`, `.sqlite3`, `.db`, `.db3`:
 
 # Internal URIs
 
-`skill://<name>`, `agent://<id>`, `artifact://<id>`, `memory://root`, `rule://<name>`, `local://<name>.md`, `mcp://<uri>` resolve transparently and accept the same line selectors as filesystem paths. Use `artifact://<id>` to recover full output that a previous bash/eval/tool result spilled or truncated.
+`skill://<name>`, `agent://<id>`, `artifact://<id>`, `memory://root`, `rule://<name>`, `local://<name>.md`, `vault://<vault>/<path>`, `mcp://<uri>` resolve transparently and accept the same line selectors as filesystem paths. Use `artifact://<id>` to recover full output that a previous bash/eval/tool result spilled or truncated.
 
 <critical>
 - You MUST use `read` for every file, directory, archive, and URL inspection. `cat`, `head`, `tail`, `less`, `more`, `ls`, `tar`, `unzip`, `curl`, `wget` are FORBIDDEN — any such bash call is a bug, regardless of how short or convenient it looks.

package/src/prompts/tools/search.md

@@ -9,7 +9,7 @@ Searches files using powerful regex matching.
 
 <output>
 {{#if IS_HL_MODE}}
-- Text output emits a file-hash header per matched file plus numbered lines: `¶src/login.ts#3c4d`, `*42:if (user.id) {` (match), ` 43:return user;` (context). Copy the header for anchored edits; ops use bare line numbers.
+- Text output emits a file snapshot tag header per matched file plus numbered lines: `¶src/login.ts#1f`, `*42:if (user.id) {` (match), ` 43:return user;` (context). Copy the header for anchored edits; ops use bare line numbers.
 {{else}}
 {{#if IS_LINE_NUMBER_MODE}}
 - Text output is line-number-prefixed

package/src/sdk.ts

@@ -10,6 +10,7 @@ import {
 } from "@oh-my-pi/pi-agent-core";
 import {
 	type CredentialDisabledEvent,
+	isUsageLimitError,
 	type Message,
 	type Model,
 	type SimpleStreamOptions,
@@ -23,6 +24,7 @@ import type { Component } from "@oh-my-pi/pi-tui";
 import {
 	$env,
 	$flag,
+	extractRetryHint,
 	getAgentDbPath,
 	getAgentDir,
 	getProjectDir,
@@ -1885,13 +1887,36 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 					...streamOptions,
 					openrouterVariant: streamOptions?.openrouterVariant ?? openrouterVariant,
 					onAuthError: async (provider, oldKey, error) => {
+						const message = error instanceof Error ? error.message : String(error);
+						// streamSimple invokes this for both 401 auth failures AND
+						// rotatable usage-limit errors (Codex usage_limit_reached,
+						// Anthropic usage_limit_reached, etc.). The two need
+						// different storage actions: a real 401 means the credential
+						// is bad and should be marked suspect; a usage limit just
+						// means this account is parked until reset and should be
+						// temporarily blocked so a sibling can pick the request up.
+						if (isUsageLimitError(message)) {
+							const retryAfterMs = extractRetryHint(undefined, message);
+							const switched = await modelRegistry.authStorage.markUsageLimitReached(provider, agent.sessionId, {
+								retryAfterMs,
+								signal: streamOptions?.signal,
+							});
+							logger.debug("Retrying provider request after usage-limit block", {
+								provider,
+								switched,
+								retryAfterMs,
+								error: message,
+							});
+							if (!switched) return undefined;
+							return modelRegistry.getApiKeyForProvider(provider, agent.sessionId);
+						}
 						await modelRegistry.authStorage.invalidateCredentialMatching(provider, oldKey, {
 							signal: streamOptions?.signal,
 							sessionId: agent.sessionId,
 						});
 						logger.debug("Retrying provider request after credential invalidation", {
 							provider,
-							error: error instanceof Error ? error.message : String(error),
+							error: message,
 						});
 						return modelRegistry.getApiKeyForProvider(provider, agent.sessionId);
 					},

package/src/session/agent-session.ts

@@ -18,6 +18,7 @@ import * as fs from "node:fs";
 import * as path from "node:path";
 import { scheduler } from "node:timers/promises";
 import { isPromise } from "node:util/types";
+import type { InMemorySnapshotStore } from "@oh-my-pi/hashline";
 import {
 	type AfterToolCallContext,
 	type AfterToolCallResult,
@@ -104,6 +105,8 @@ import { onAppendOnlyModeChanged } from "../config/settings";
 import { RawSseDebugBuffer } from "../debug/raw-sse-buffer";
 import { loadCapability } from "../discovery";
 import { expandApplyPatchToEntries, normalizeDiff, normalizeToLF, ParseError, previewPatch, stripBom } from "../edit";
+import { getFileSnapshotStore } from "../edit/file-snapshot-store";
+import { namespaceSessionId as namespacePythonSessionId } from "../eval/py";
 import {
 	disposeKernelSessionsByOwner,
 	executePython as executePythonCommand,
@@ -209,7 +212,11 @@ import { YieldQueue } from "./yield-queue";
 /** Session-specific events that extend the core AgentEvent */
 export type AgentSessionEvent =
 	| AgentEvent
-	| { type: "auto_compaction_start"; reason: "threshold" | "overflow" | "idle"; action: "context-full" | "handoff" }
+	| {
+			type: "auto_compaction_start";
+			reason: "threshold" | "overflow" | "idle" | "incomplete";
+			action: "context-full" | "handoff";
+	  }
 	| {
 			type: "auto_compaction_end";
 			action: "context-full" | "handoff";
@@ -738,6 +745,7 @@ export class AgentSession {
 	readonly sessionManager: SessionManager;
 	readonly settings: Settings;
 	readonly yieldQueue: YieldQueue;
+	fileSnapshotStore?: InMemorySnapshotStore;
 
 	#powerAssertion: MacOSPowerAssertion | undefined;
 
@@ -4156,6 +4164,7 @@ export class AgentSession {
 				const fileMentionMessages = await generateFileMentionMessages(fileMentions, this.sessionManager.getCwd(), {
 					autoResizeImages: this.settings.get("images.autoResize"),
 					useHashLines: resolveFileDisplayMode(this).hashLines,
+					snapshotStore: getFileSnapshotStore(this),
 				});
 				messages.push(...fileMentionMessages);
 			}
@@ -5662,10 +5671,14 @@ export class AgentSession {
 	 * Check if context maintenance or promotion is needed and run it.
 	 * Called after agent_end and before prompt submission.
 	 *
-	 * Three cases (in order):
-	 * 1. Overflow + promotion: promote to larger model, retry without maintenance
-	 * 2. Overflow + no promotion target: run context maintenance, auto-retry on same model
-	 * 3. Threshold: Context over threshold, run context maintenance (no auto-retry)
+	 * Four cases (in order):
+	 * 1. Input overflow + promotion: promote to larger model, retry without maintenance.
+	 * 2. Input overflow + no promotion target: run context maintenance, auto-retry on same model.
+	 * 3. Output incomplete (stopReason === "length", e.g. `response.incomplete`): the
+	 *    model burned its output budget without producing an actionable deliverable
+	 *    (reasoning-only or truncated). Drop the dead turn, try promotion, otherwise
+	 *    run compaction/handoff and retry.
+	 * 4. Threshold: context over threshold, run context maintenance (no auto-retry).
 	 *
 	 * @param assistantMessage The assistant message to check
 	 * @param skipAbortedCheck If false, include aborted messages (for pre-prompt check). Default: true
@@ -5724,10 +5737,49 @@ export class AgentSession {
 			}
 			return false;
 		}
+
+		// Case 3: Output-side incomplete — `response.incomplete` from OpenAI Responses
+		// (and Codex) maps to stopReason === "length". The model burned its
+		// `max_output_tokens` budget on reasoning/text and emitted no actionable
+		// deliverable. Same recovery class as overflow: promotion if available,
+		// otherwise compaction/handoff. Unlike overflow, the *input* is fine, so we
+		// allow the handoff strategy to actually run.
+		if (sameModel && !errorIsFromBeforeCompaction && assistantMessage.stopReason === "length") {
+			const messages = this.agent.state.messages;
+			if (messages.length > 0 && messages[messages.length - 1].role === "assistant") {
+				this.agent.replaceMessages(messages.slice(0, -1));
+			}
+
+			const promoted = await this.#tryContextPromotion(assistantMessage);
+			if (promoted) {
+				logger.debug("Context promotion triggered by response.incomplete (length stop)", {
+					from: `${assistantMessage.provider}/${assistantMessage.model}`,
+				});
+				this.#scheduleAgentContinue({ delayMs: 100, generation });
+				return false;
+			}
+
+			const incompleteCompactionSettings = this.settings.getGroup("compaction");
+			if (incompleteCompactionSettings.enabled && incompleteCompactionSettings.strategy !== "off") {
+				logger.debug("Compaction triggered by response.incomplete (length stop, no promotion target)", {
+					model: `${assistantMessage.provider}/${assistantMessage.model}`,
+					strategy: incompleteCompactionSettings.strategy,
+				});
+				await this.#runAutoCompaction("incomplete", true, false, allowDefer);
+			} else {
+				// Neither promotion nor compaction is available — surface the dead-end so
+				// the user understands why the turn yielded with nothing.
+				logger.warn("response.incomplete with no recovery path (promotion + compaction both unavailable)", {
+					model: `${assistantMessage.provider}/${assistantMessage.model}`,
+				});
+			}
+			return false;
+		}
+
 		const compactionSettings = this.settings.getGroup("compaction");
 		if (!compactionSettings.enabled || compactionSettings.strategy === "off") return false;
 
-		// Case 2: Threshold - turn succeeded but context is getting large
+		// Case 4: Threshold - turn succeeded but context is getting large
 		// Skip if this was an error (non-overflow errors don't have usage data)
 		if (assistantMessage.stopReason === "error") return false;
 		const pruneResult = await this.#pruneToolOutputs();
@@ -6450,7 +6502,7 @@ export class AgentSession {
 	 * @returns true when a deferred handoff was scheduled. Inline runs always return false.
 	 */
 	async #runAutoCompaction(
-		reason: "overflow" | "threshold" | "idle",
+		reason: "overflow" | "threshold" | "idle" | "incomplete",
 		willRetry: boolean,
 		deferred = false,
 		allowDefer = true,
@@ -6459,10 +6511,14 @@ export class AgentSession {
 		if (compactionSettings.strategy === "off") return false;
 		if (reason !== "idle" && !compactionSettings.enabled) return false;
 		const generation = this.#promptGeneration;
+		// "overflow" and "incomplete" force inline execution because they are recovery
+		// paths the caller wants resolved before scheduling the next turn. "idle" is
+		// triggered by the idle loop and does its own scheduling.
 		if (
 			!deferred &&
 			allowDefer &&
 			reason !== "overflow" &&
+			reason !== "incomplete" &&
 			reason !== "idle" &&
 			compactionSettings.strategy === "handoff"
 		) {
@@ -6477,6 +6533,9 @@ export class AgentSession {
 			return true;
 		}
 
+		// "overflow" forces context-full because the input itself is broken — a handoff
+		// LLM call would hit the same overflow. "incomplete" is an output-side problem,
+		// so a handoff request on the existing context is still viable.
 		let action: "context-full" | "handoff" =
 			compactionSettings.strategy === "handoff" && reason !== "overflow" ? "handoff" : "context-full";
 		await this.#emitSessionEvent({ type: "auto_compaction_start", reason, action });
@@ -6777,8 +6836,18 @@ export class AgentSession {
 			if (willRetry) {
 				const messages = this.agent.state.messages;
 				const lastMsg = messages[messages.length - 1];
-				if (lastMsg?.role === "assistant" && (lastMsg as AssistantMessage).stopReason === "error") {
-					this.agent.replaceMessages(messages.slice(0, -1));
+				if (lastMsg?.role === "assistant") {
+					const lastAssistant = lastMsg as AssistantMessage;
+					// Drop the prior turn before retry when it carries no actionable deliverable:
+					// - "error": failure was kept in history but must not re-enter the next turn's prompt.
+					// - reason === "incomplete" && stopReason === "length": truncated output (typically
+					//   reasoning-only) — re-running it produces the same dead-end.
+					const shouldDrop =
+						lastAssistant.stopReason === "error" ||
+						(reason === "incomplete" && lastAssistant.stopReason === "length");
+					if (shouldDrop) {
+						this.agent.replaceMessages(messages.slice(0, -1));
+					}
 				}
 
 				this.#scheduleAgentContinue({ delayMs: 100, generation });
@@ -6812,7 +6881,9 @@ export class AgentSession {
 				errorMessage:
 					reason === "overflow"
 						? `Context overflow recovery failed: ${errorMessage}`
-						: `Auto-compaction failed: ${errorMessage}`,
+						: reason === "incomplete"
+							? `Incomplete response recovery failed: ${errorMessage}`
+							: `Auto-compaction failed: ${errorMessage}`,
 			});
 		} finally {
 			if (this.#autoCompactionAbortController === autoCompactionAbortController) {
@@ -7521,7 +7592,7 @@ export class AgentSession {
 				});
 			const result = await executePythonCommand(code, {
 				cwd,
-				sessionId,
+				sessionId: namespacePythonSessionId(sessionId),
 				kernelOwnerId: this.#evalKernelOwnerId,
 				kernelMode: this.settings.get("python.kernelMode"),
 				onChunk,

package/src/system-prompt.ts

@@ -11,6 +11,7 @@ import { systemPromptCapability } from "./capability/system-prompt";
 import type { SkillsSettings } from "./config/settings";
 import { type ContextFile, loadCapability, type SystemPrompt as SystemPromptFile } from "./discovery";
 import { loadSkills, type Skill } from "./extensibility/skills";
+import { hasObsidian } from "./internal-urls/vault-protocol";
 import customSystemPromptTemplate from "./prompts/system/custom-system-prompt.md" with { type: "text" };
 import projectPromptTemplate from "./prompts/system/project-prompt.md" with { type: "text" };
 import systemPromptTemplate from "./prompts/system/system-prompt.md" with { type: "text" };
@@ -569,6 +570,7 @@ export async function buildSystemPrompt(options: BuildSystemPromptOptions = {}):
 		mcpDiscoveryServerSummaries,
 		eagerTasks,
 		secretsEnabled,
+		hasObsidian: hasObsidian(),
 	};
 	const rendered = prompt.render(resolvedCustomPrompt ? customSystemPromptTemplate : systemPromptTemplate, data);
 	const systemPrompt = [rendered];

package/src/tools/ast-edit.ts

@@ -1,11 +1,13 @@
 import * as path from "node:path";
-import { computeFileHash, formatHashlineHeader } from "@oh-my-pi/hashline";
+import { formatHashlineHeader } from "@oh-my-pi/hashline";
 import type { AgentTool, AgentToolContext, AgentToolResult, AgentToolUpdateCallback } from "@oh-my-pi/pi-agent-core";
 import { type AstReplaceChange, type AstReplaceFileChange, astEdit } from "@oh-my-pi/pi-natives";
 import type { Component } from "@oh-my-pi/pi-tui";
 import { Text } from "@oh-my-pi/pi-tui";
 import { $envpos, prompt, untilAborted } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
+import { getFileSnapshotStore } from "../edit/file-snapshot-store";
+import { normalizeToLF } from "../edit/normalize";
 import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import type { Theme } from "../modes/theme/theme";
 import astEditDescription from "../prompts/tools/ast-edit.md" with { type: "text" };
@@ -281,14 +283,15 @@ export class AstEditTool implements AgentTool<typeof astEditSchema, AstEditToolD
 			}
 
 			const useHashLines = resolveFileDisplayMode(this.session).hashLines;
-			const hashContexts = new Map<string, { fileHash: string }>();
+			const hashContexts = new Map<string, { tag: string }>();
 			if (useHashLines) {
+				const snapshotStore = getFileSnapshotStore(this.session);
 				for (const relativePath of fileList) {
 					const absolutePath = path.resolve(this.session.cwd, relativePath);
 					try {
-						const fullText = await Bun.file(absolutePath).text();
-						const fileHash = computeFileHash(fullText);
-						hashContexts.set(relativePath, { fileHash });
+						const fullText = normalizeToLF(await Bun.file(absolutePath).text());
+						const tag = snapshotStore.recordContiguous(absolutePath, 1, fullText.split("\n"), { fullText });
+						hashContexts.set(relativePath, { tag });
 					} catch {
 						// Best-effort: if a file disappears between ast-edit and rendering, emit plain line output.
 					}
@@ -326,7 +329,7 @@ export class AstEditTool implements AgentTool<typeof astEditSchema, AstEditToolD
 					const rendered = renderChangesForFile(relativePath);
 					const count = fileReplacementCounts.get(relativePath) ?? 0;
 					const hashContext = hashContexts.get(relativePath);
-					const hashSuffix = hashContext ? `#${hashContext.fileHash}` : "";
+					const hashSuffix = hashContext ? `#${hashContext.tag}` : "";
 					return {
 						headerSuffix: `${hashSuffix} (${formatCount("replacement", count)})`,
 						modelLines: rendered.model,
@@ -346,7 +349,7 @@ export class AstEditTool implements AgentTool<typeof astEditSchema, AstEditToolD
 					}
 					const hashContext = hashContexts.get(relativePath);
 					if (hashContext) {
-						outputLines.push(formatHashlineHeader(relativePath, hashContext.fileHash));
+						outputLines.push(formatHashlineHeader(relativePath, hashContext.tag));
 					}
 					outputLines.push(...rendered.model);
 					displayLines.push(...rendered.display);