@diegopetrucci/pi-extensions 0.1.71 → 0.1.73

package/extensions/contrarian/index.ts

@@ -111,6 +111,11 @@ const CONTRARIAN_MODEL_PREFERENCES = [
 	"openai.gpt-6-astra ",
 	"gpt-6-astra ",
 	"gpt-6-astra",
+	"openai/gpt-6.1-sol ",
+	"global.openai.gpt-6.1-sol ",
+	"us.openai.gpt-6.1-sol ",
+	"openai.gpt-6.1-sol ",
+	"gpt-6.1-sol ",
 	"openai/gpt-6-sol ",
 	"gpt-6-sol ",
 	"gpt-6-sol",
@@ -171,6 +176,9 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"us.openai.gpt-6-astra ",
 		"openai.gpt-6-astra ",
 		"gpt-6-astra",
+		"global.openai.gpt-6.1-sol ",
+		"us.openai.gpt-6.1-sol ",
+		"openai.gpt-6.1-sol ",
 		"openai.gpt-5.6-sol",
 		"openai.gpt-5.6-terra",
 		"openai.gpt-5.6-luna",
@@ -202,6 +210,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 	"azure-openai-responses": [
 		"gpt-6-astra ",
 		"gpt-6-astra",
+		"gpt-6.1-sol ",
 		"gpt-6-sol ",
 		"gpt-6-sol",
 		"gpt-5.6-sol",
@@ -281,13 +290,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 	fireworks: [
 		"accounts/fireworks/models/kimi-k3",
 		"accounts/fireworks/routers/kimi-k3-fast",
-		"accounts/fireworks/models/deepseek-v4-pro",
-		"accounts/fireworks/models/kimi-k2p7-code",
-		"accounts/fireworks/models/glm-5p2",
-		"accounts/fireworks/routers/glm-5p2-fast",
 		"accounts/fireworks/models/minimax-m3",
-		"accounts/fireworks/models/kimi-k2p6",
-		"accounts/fireworks/models/qwen3p7-plus",
 		"accounts/fireworks/models/gpt-oss-120b",
 	],
 	"github-copilot": [
@@ -299,6 +302,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"claude-opus-4.7",
 		"gpt-6-astra ",
 		"gpt-6-astra",
+		"gpt-6.1-sol ",
 		"gpt-6-sol ",
 		"gpt-6-sol",
 		"gpt-5.6-sol",
@@ -386,6 +390,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 	openai: [
 		"gpt-6-astra ",
 		"gpt-6-astra",
+		"gpt-6.1-sol ",
 		"gpt-6-sol ",
 		"gpt-6-sol",
 		"gpt-5.6-sol",
@@ -411,6 +416,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 	"openai-codex": [
 		"gpt-6-astra ",
 		"gpt-6-astra",
+		"gpt-6.1-sol ",
 		"gpt-6-sol ",
 		"gpt-6-sol",
 		"gpt-5.6-sol",
@@ -428,6 +434,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"claude-fable-5",
 		"gpt-6-astra ",
 		"gpt-6-astra",
+		"gpt-6.1-sol ",
 		"gpt-6-sol ",
 		"gpt-6-sol",
 		"gpt-5.6-sol",
@@ -466,17 +473,13 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"deepseek-v4-pro",
 		"glm-5.3",
 		"glm-5.2",
-		"glm-5.1",
-		"qwen3.7-max",
 		"qwen3.7-plus",
-		"qwen3.6-plus",
 		"mimo-v2.5-pro",
 		"mimo-v2.5",
 		"minimax-m3",
 		"minimax-m2.7",
 		"kimi-k3",
 		"kimi-k2.7-code",
-		"kimi-k2.6",
 		"grok-4.7 ",
 		"grok-4.7",
 		"grok-4.6",
@@ -494,6 +497,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"anthropic/claude-opus-4.1",
 		"openai/gpt-6-astra ",
 		"openai/gpt-6-astra",
+		"openai/gpt-6.1-sol ",
 		"openai/gpt-6-sol ",
 		"openai/gpt-6-sol",
 		"openai/gpt-5.6-sol-pro",
@@ -582,9 +586,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"gpt-5.4",
 		"gpt-5.3-codex",
 		"kimi-k3",
-		"deepseek-v4-pro",
 		"glm-5.3",
-		"kimi-k2.7-code",
 		"deepseek-v4.1-flash",
 		"gpt-5.4-mini",
 		"glm-5.3-flash",
@@ -594,8 +596,6 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"moonshotai/Kimi-K3",
 		"zai-org/GLM-5.3-Flash",
 		"zai-org/GLM-5.2",
-		"moonshotai/Kimi-K2.7-Code",
-		"moonshotai/Kimi-K2.6",
 		"Qwen/Qwen3.7-Max",
 		"Qwen/Qwen3.6-Plus",
 		"MiniMaxAI/MiniMax-M3",
@@ -605,6 +605,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"nvidia/nemotron-3-ultra-550b-a55b",
 		"google/gemma-4-31B-it",
 	],
+	typesafe: [],
 	"vercel-ai-gateway": [
 		"anthropic/claude-opus-5.5 ",
 		"anthropic/claude-opus-5.5",
@@ -618,6 +619,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"anthropic/claude-sonnet-4.6",
 		"openai/gpt-6-astra ",
 		"openai/gpt-6-astra",
+		"openai/gpt-6.1-sol ",
 		"openai/gpt-6-sol ",
 		"openai/gpt-6-sol",
 		"openai/gpt-5.6-sol",
@@ -1387,7 +1389,10 @@ async function runContrarian(
 			}
 
 			if (event.type === "message_start" && event.message?.role === "assistant") {
+				// A later assistant turn supersedes any earlier tool/intermediate answer.
+				// Keep only the current message so a tool-only final turn cannot return stale text.
 				currentText = "";
+				finalOutput = "";
 				emit();
 				return;
 			}
@@ -1400,7 +1405,8 @@ async function runContrarian(
 
 			if (event.type === "message_end" && event.message?.role === "assistant") {
 				const text = extractTextFromContent(event.message.content) || currentText;
-				if (text) finalOutput = text;
+				// message_end is authoritative, including an empty tool-only message.
+				finalOutput = text;
 				currentText = "";
 
 				const stopReason = event.message.stopReason;

package/extensions/contrarian/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@diegopetrucci/pi-contrarian",
-  "version": "0.1.13",
+  "version": "0.1.14",
   "description": "A read-only contrarian subagent extension for pi that stress-tests plans, designs, assumptions, and conclusions.",
   "keywords": [
     "pi-package",

package/extensions/dynamic-context-pruning/README.md

@@ -27,11 +27,23 @@ and why it hasn't been built yet.
   key wins when replaying history, so a decision can be pruned, restored, and
   re-pruned any number of times.
 - **Recomputed every call, not on a timer.** All three automatic strategies run
-  fresh on every `context` event, over the current message array. Nothing is
-  pruned "once and forgotten"; if a tool result would no longer qualify (e.g.
-  it's now within the protected recency window), it isn't proposed again, but
-  an already-applied, persisted decision still applies until explicitly
-  restored.
+  fresh on every `context` event, over the current message array. Pi 0.99's
+  `context` hook intentionally receives the mutable conversation without system
+  messages; Pi restores its system prompt and tool state after the handler. The
+  extension does not try to rebuild that request-only envelope.
+- **Canonical picker input.** `/prune` uses Pi's canonical session projection when
+  available, so context omissions and replacements, branch selection, and tool
+  pairing match the provider-visible conversation. Older hosts retain the raw
+  message-entry fallback.
+- **Nested calls stay metadata.** Pi records bounded `nestedCalls` metadata on a
+  parent tool result, but nested results are not transcript messages. Strategies
+  count and prune only the parent transcript tool result; they never materialize
+  child entries, double-count child output, or infer complete history from a
+  truncated/incomplete record.
+- **Persisted decisions remain branch-local.** Nothing is pruned "once and
+  forgotten"; if a tool result would no longer qualify (e.g. it's now within the
+  protected recency window), it isn't proposed again, but an already-applied,
+  persisted decision still applies until explicitly restored.
 
 ## Strategies
 

package/extensions/dynamic-context-pruning/index.ts

@@ -51,6 +51,8 @@ export interface MinimalToolResultMessage {
 	toolCallId: string;
 	toolName: string;
 	content: (MinimalTextContent | MinimalImageContent)[];
+	/** Bounded nested-call metadata; child calls have no transcript results to prune. */
+	nestedCalls?: unknown;
 	isError: boolean;
 	timestamp?: number;
 	[key: string]: unknown;
@@ -114,13 +116,16 @@ export function normalizeSessionProjection(value: unknown): MinimalSessionProjec
 		});
 	}
 
+	// Entries carry the provenance that Pi uses to apply omissions and
+	// replacements. Validate the optional flattened field for compatibility, but
+	// derive the messages used by pruning from entries so stale/invented records
+	// cannot bypass canonical context edits.
 	if ("messages" in value && (!Array.isArray(value.messages) || !value.messages.every(isMinimalMessage))) return undefined;
-	const messages = Array.isArray(value.messages) ? value.messages : entries.flatMap((entry) => entry.messages);
-	return { entries, messages };
+	return { entries, messages: entries.flatMap((entry) => entry.messages) };
 }
 
 export function sessionProjectionToMessages(projection: MinimalSessionProjection): MinimalMessage[] {
-	return projection.messages ?? projection.entries.flatMap((entry) => entry.messages);
+	return projection.entries.flatMap((entry) => entry.messages);
 }
 
 // ============================================================================
@@ -1598,7 +1603,13 @@ export interface ToolCallOccurrence {
 	isError: boolean;
 }
 
-/** Collect every assistant toolCall block that has a matching toolResult message, in message order. */
+/**
+ * Collect every transcript assistant toolCall block that has a matching
+ * transcript toolResult message, in message order. `nestedCalls` is deliberately
+ * not expanded: Pi keeps only bounded call metadata there and does not record
+ * child results as transcript messages, so expanding it would double-count or
+ * invent prune targets.
+ */
 export function collectCompletedToolCallOccurrences(messages: MinimalMessage[]): ToolCallOccurrence[] {
 	const resultIndexById = new Map<string, number>();
 	const isErrorById = new Map<string, boolean>();
@@ -2318,6 +2329,8 @@ export function buildPrunableItems(
 	const items: PrunableItem[] = [];
 	for (const occurrence of collectCompletedToolCallOccurrences(messages)) {
 		const resultMessage = messages[occurrence.resultIndex] as MinimalToolResultMessage;
+		// Only provider-visible transcript content is economically attributable;
+		// bounded nested-call metadata has no child result content to estimate.
 		const estimatedTokens = estimateTokensForContent(resultMessage.content);
 		const activeDecision = activeByToolCallId.get(occurrence.toolCallId);
 		const status: PrunableItemStatus = activeDecision

package/extensions/dynamic-context-pruning/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-dynamic-context-pruning",
-  "version": "0.1.10",
+  "version": "0.1.11",
   "type": "module",
   "description": "A pi extension that dynamically prunes stale context to keep sessions lean.",
   "keywords": [

package/extensions/fast/README.md

@@ -38,9 +38,19 @@ Direct OpenAI API Fast mode requires all of the following:
 
 - Provider `openai`.
 - API `openai-responses` or `openai-completions`.
-- Model `gpt-6-astra`, `gpt-6-sol`, or `gpt-6-luna`.
+- Model `gpt-6.1-sol`, `gpt-6-astra`, `gpt-6-sol`, or `gpt-6-luna`.
+- API-key auth or Pi's `openai` ChatGPT OAuth (`Sign in with ChatGPT`).
 - No existing `service_tier` field in the request payload.
 
+The direct OpenAI route stays on OpenAI's direct API contract for both API-key
+auth and Pi's `openai` ChatGPT OAuth. Acceptance of `service_tier: "fast"` by
+Pi's `openai` ChatGPT OAuth sign-in endpoint has not been live-verified and
+remains deferred to manual gate `pel-71t6`. This direct route is never forwarded
+to legacy `openai-codex` or its subscription-usage endpoint. The documented
+request shape is support evidence, not live certification; this change was
+validated only with mocked eligibility and request payloads, not live or paid
+provider calls.
+
 OpenAI Codex Fast mode requires all of the following:
 
 - Provider `openai-codex`.
@@ -49,11 +59,18 @@ OpenAI Codex Fast mode requires all of the following:
 - ChatGPT OAuth/subscription auth, not API-key auth.
 - No existing `service_tier` field in the request payload.
 
+GPT-6.1 Sol is intentionally unsupported on this legacy Codex route. The
+Codex speed documentation lists GPT-6.1 Sol Fast with `service_tier: "fast"`,
+but Pi's legacy Codex service-tier cost handling currently accounts for
+`priority` and `flex`, not `fast`. The direct OpenAI API's `priority` alias is
+not evidence for the legacy Codex endpoint, so this extension leaves the route
+fail-closed without patching upstream pricing.
+
 Anthropic Fast mode requires all of the following:
 
 - Provider `anthropic`.
 - API `anthropic-messages`.
-- Model `claude-opus-4-8` or `claude-opus-5`.
+- Model `claude-opus-4-8`, `claude-opus-5`, or `claude-opus-5-5`.
 - No existing `speed` field in the request payload.
 
 Anthropic Fast mode supports both API-key and Claude Code OAuth access when the account has access to the research preview.
@@ -129,7 +146,11 @@ The collection package loads this unified extension instead of `openai-fast` and
 ## Notes
 
 - Anthropic Fast mode has separate rate limits, costs more than standard speed, and does not share prompt-cache prefixes with standard-speed requests.
-- Direct OpenAI Fast mode is limited to the confirmed, officially priced GPT-6 API models above; Fast pricing, availability, and rate limits remain provider- and account-dependent.
-- OpenAI Codex Fast mode intentionally does not affect API-key models.
+- Direct OpenAI Fast mode is limited to the confirmed GPT-6 API models above; Fast pricing, availability, and rate limits remain provider- and account-dependent.
+- Authoritative support evidence: [OpenAI Fast mode](https://developers.openai.com/api/docs/guides/fast-mode) explicitly uses `gpt-6.1-sol` with `service_tier: "fast"` and documents `priority` as an equivalent alias for supported API models; the [GPT-6.1 Sol model page](https://developers.openai.com/api/docs/models/gpt-6.1-sol) documents the model itself.
+- The [Codex speed guide](https://developers.openai.com/codex/speed/) documents GPT-6.1 Sol Fast for ChatGPT-signed-in Codex, but legacy Codex support remains pending separate wire and cost validation; this ticket makes no live-certification claim.
+- Pi's provider catalog remains the source of reported model pricing, including any tiered pricing; this extension does not apply guessed discounts or rewrite cost metadata.
+- OpenAI Codex Fast mode intentionally does not affect API-key models, and GPT-6.1 Sol remains blocked on the legacy Codex route.
+- The retired `openai-fast` and `claude-fast` packages remain legacy standalone boundaries; new provider/auth behavior belongs only to this unified extension.
 - Existing `speed` and `service_tier` fields are never overwritten.
 - Pi's model catalog may not include a Fast-mode premium. Cost accounting depends on the provider reporting the effective tier in its streamed response, and this extension does not patch usage totals.

package/extensions/fast/index.ts

@@ -15,7 +15,7 @@ const ANTHROPIC_FAST_BETA = "fast-mode-2026-02-01";
 const CLAUDE_CODE_OAUTH_BETAS = ["claude-code-20250219", "oauth-2025-04-20"];
 const FINE_GRAINED_TOOL_STREAMING_BETA = "fine-grained-tool-streaming-2025-05-14";
 const INTERLEAVED_THINKING_BETA = "interleaved-thinking-2025-05-14";
-const ANTHROPIC_SUPPORTED_MODELS = new Set(["claude-opus-4-8", "claude-opus-5"]);
+const ANTHROPIC_SUPPORTED_MODELS = new Set(["claude-opus-4-8", "claude-opus-5", "claude-opus-5-5"]);
 
 const OPENAI_PROVIDER_ID = "openai-codex";
 const OPENAI_API_ID = "openai-codex-responses";
@@ -29,7 +29,7 @@ const OPENAI_SUPPORTED_MODELS = new Set([
 const OPENAI_API_PROVIDER_ID = "openai";
 const OPENAI_API_IDS = new Set(["openai-responses", "openai-completions"]);
 const OPENAI_API_FAST_SERVICE_TIER = "fast";
-const OPENAI_API_SUPPORTED_MODELS = new Set(["gpt-6-astra", "gpt-6-sol", "gpt-6-luna"]);
+const OPENAI_API_SUPPORTED_MODELS = new Set(["gpt-6.1-sol", "gpt-6-astra", "gpt-6-sol", "gpt-6-luna"]);
 
 const DEFAULT_CONFIG: FastConfig = {
 	enabled: false,
@@ -163,7 +163,7 @@ function getEligibility(ctx: ExtensionContext): Eligibility {
 			return {
 				eligible: false,
 				modelKey: key,
-				reason: "Fast mode is only enabled for Claude Opus 4.8 and Claude Opus 5",
+				reason: "Fast mode is only enabled for Claude Opus 4.8, Claude Opus 5, and Claude Opus 5.5",
 			};
 		}
 
@@ -216,7 +216,7 @@ function getEligibility(ctx: ExtensionContext): Eligibility {
 			return {
 				eligible: false,
 				modelKey: key,
-				reason: "Fast mode is only enabled for GPT-6 Astra, GPT-6 Sol, and GPT-6 Luna on direct OpenAI APIs",
+				reason: "Fast mode is only enabled for GPT-6.1 Sol, GPT-6 Astra, GPT-6 Sol, and GPT-6 Luna on direct OpenAI APIs",
 			};
 		}
 

package/extensions/fast/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@diegopetrucci/pi-fast",
-  "version": "0.1.3",
+  "version": "0.1.5",
   "description": "A pi extension that enables Fast mode for supported direct OpenAI API, OpenAI Codex, and Anthropic Claude models based on the currently selected model.",
   "keywords": [
     "pi-package",

package/extensions/git-footer/README.md

@@ -40,4 +40,5 @@ Then reload pi:
 - Does not replace pi's built-in footer.
 - Uses `ctx.ui.setStatus()`, so pi renders the git summary with other extension statuses.
 - The current pi extension API does not support literally appending text inside the built-in footer's first `cwd (branch)` line without replacing the footer.
+- Status text is passed through Pi's active `dim` theme color; the extension does not embed terminal escape sequences, so theme changes remain host-controlled.
 - Git and GitHub CLI lookups run on a short background interval with timeouts.

package/extensions/git-footer/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@diegopetrucci/pi-git-footer",
-  "version": "0.1.10",
+  "version": "0.1.11",
   "description": "A TLH-style git status add-on for pi's built-in footer.",
   "keywords": [
     "pi-package",

package/extensions/librarian/index.ts

@@ -10,6 +10,7 @@ import type { Usage } from "@earendil-works/pi-ai";
 import {
 	DefaultResourceLoader,
 	SessionManager,
+	SettingsManager,
 	createAgentSession,
 	getAgentDir,
 	getMarkdownTheme,
@@ -1434,9 +1435,14 @@ export default function librarianExtension(pi: ExtensionAPI) {
 					cacheRoot,
 				});
 
+				// Keep the research session from inheriting user/project tool, shell,
+				// retry, or provider-request settings. The host model runtime is still
+				// passed explicitly below so the selected provider/auth policy is kept.
+				const isolatedSettingsManager = SettingsManager.inMemory({});
 				const resourceLoader = new DefaultResourceLoader({
 					cwd: workspace,
 					agentDir: getAgentDir(),
+					settingsManager: isolatedSettingsManager,
 					noExtensions: true,
 					noSkills: true,
 					noPromptTemplates: true,
@@ -1469,6 +1475,7 @@ export default function librarianExtension(pi: ExtensionAPI) {
 						cwd: workspace,
 						...getModelRuntimeOption(ctx),
 						resourceLoader,
+						settingsManager: isolatedSettingsManager,
 						sessionManager: SessionManager.inMemory(workspace),
 						model: candidate.model,
 						thinkingLevel: candidate.details.thinkingLevel,

package/extensions/librarian/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@diegopetrucci/pi-librarian",
-  "version": "0.1.16",
+  "version": "0.1.17",
   "description": "A pi GitHub research scout with a toggleable local repo checkout cache under the user's OS cache directory.",
   "keywords": [
     "pi-package",
@@ -30,6 +30,7 @@
     ]
   },
   "peerDependencies": {
+    "@earendil-works/pi-ai": "*",
     "@earendil-works/pi-coding-agent": "*",
     "@earendil-works/pi-tui": "*",
     "typebox": "*"

package/extensions/minimal-footer/README.md

@@ -45,7 +45,7 @@ When context usage is above 200k tokens, the bottom-left line includes a red war
 44.1% · DUMB ZONE
 ```
 
-When using `openai-codex`, the bottom-left line also includes subscription usage. A typical account with both windows looks like:
+When using legacy `openai-codex`, the bottom-left line also includes subscription usage. A typical account with both windows looks like:
 
 ```text
 44.1% · 5h 12% · 7d 38%
@@ -53,6 +53,15 @@ When using `openai-codex`, the bottom-left line also includes subscription usage
 
 Window labels come from OpenAI's reported durations, so an account that reports only a weekly primary window is shown correctly (for example, `7d 38%`) instead of being assumed to have a 5-hour window.
 
+Pi's `openai` ChatGPT OAuth uses the direct OpenAI API and has no documented
+account-usage endpoint compatible with this legacy WHAM reader. In that case,
+the footer shows `usage unsupported` and does not send the new token to
+`chatgpt.com/backend-api/wham/usage`. Direct OpenAI API-key users have no
+subscription-usage line. Legacy Codex fetch failures, missing OAuth, or a
+malformed/mismatched token account claim show `usage unavailable` rather than
+retaining an older account's snapshot; the request is not sent until the token
+claim and lifecycle identity agree.
+
 When `PI_EXPERIMENTAL=1`, the bottom-left line also includes an experimental marker:
 
 ```text
@@ -203,8 +212,8 @@ Disable git dirty/ahead/PR status:
 - `context.dumbZone.thresholdTokens`: token threshold for `DUMB ZONE`
 - `context.dumbZone.label`: warning text
 - `context.dumbZone.color`: theme color for the warning (`error`, `warning`, `accent`, `text`, or `dim`)
-- `codexUsage.enabled`: show OpenAI Codex session-limit usage when using `openai-codex`
-- `codexUsage.cacheTtlMs`: in-memory usage cache duration
+- `codexUsage.enabled`: show OpenAI Codex session-limit usage when using `openai-codex`; also controls the explicit unsupported/unavailable status for account usage
+- `codexUsage.cacheTtlMs`: in-memory usage-result cache duration and account-only login detection interval; provider/model/account changes invalidate the cached snapshot. The timer only compares identity and does not fetch unchanged-account usage. The default is 5 minutes; `0` disables usage-result caching but keeps identity checks every 5 minutes. Positive values above Node's `2,147,483,647` ms timer limit use that maximum only for scheduling; the configured TTL still controls lifecycle cache semantics.
 - `codexUsage.requestTimeoutMs`: usage request timeout
 - `codexUsage.windows.primary.enabled`: show the primary usage window
 - `codexUsage.windows.primary.label`: `"auto"` derives the primary label from OpenAI's reported duration; any other value overrides it
@@ -224,6 +233,7 @@ Disable git dirty/ahead/PR status:
 - **Top right:** current repo directory name
 - **Bottom left:** current context usage percentage, plus red `DUMB ZONE` above 200k context tokens
 - **Bottom left on `openai-codex`:** current context usage percentage plus the Codex usage windows reported for the account
+- **Bottom left on `openai` ChatGPT OAuth:** current context usage percentage plus `usage unsupported`; no legacy WHAM request is attempted
 - **Bottom left with `PI_EXPERIMENTAL=1`:** current context usage percentage plus `xp`
 - **Bottom right:** model id and thinking level
 
@@ -240,4 +250,6 @@ This extension also lives inside the broader [`pi-extensions`](../../README.md)
 - Shows the model id rather than a provider-specific display label.
 - Shows `xp` when `PI_EXPERIMENTAL=1`.
 - For `openai-codex`, reads pi's stored OAuth login and fetches usage from ChatGPT's backend usage endpoint. Window labels are inferred from each window's reported duration; missing or unrecognized durations use neutral `usage` labels rather than guessing from window position.
-- Usage is cached briefly in memory and refreshed after turns.
+- For direct `openai` ChatGPT OAuth, no usage endpoint is assumed: the footer marks usage unsupported and never calls the legacy WHAM endpoint. API-key requests do not use subscription usage.
+- Usage is cached briefly in memory and refreshed after turns. A provider/model/account boundary clears the old snapshot, and late results from an earlier request are ignored. Because Pi 0.99 emits no auth-change event, account-only logins are detected by one per-session identity-only timer: it uses `cacheTtlMs` when positive, checks every five minutes when `cacheTtlMs: 0`, and clamps larger schedules to Node's `2,147,483,647` ms maximum. Unchanged-account ticks make no WHAM request; an account change uses the normal safe refresh. The timer is disposed on shutdown/restart, and no credential file is read during footer rendering.
+- Footer colors continue to come from Pi's active theme; the extension does not embed ANSI color codes.