@oh-my-pi/pi-coding-agent 16.0.4 → 16.0.5

package/src/modes/utils/context-usage.ts

@@ -94,7 +94,7 @@ export function computeNonMessageTokens(session: AgentSession): number {
  * the status-line fast path intentionally uses the equivalent collapsed total
  * in `computeNonMessageTokens`.
  */
-function computeNonMessageBreakdown(session: AgentSession): {
+export function computeNonMessageBreakdown(session: AgentSession): {
 	skillsTokens: number;
 	toolsTokens: number;
 	systemContextTokens: number;
@@ -119,22 +119,37 @@ export function computeContextBreakdown(
 	const model = session.model;
 	const contextWindow = model?.contextWindow ?? 0;
 
+	const breakdown = typeof session.getContextBreakdown === "function" ? session.getContextBreakdown() : undefined;
+
 	let messagesTokens = 0;
-	const convo = session.messages;
-	if (convo) {
-		for (const message of convo) {
-			messagesTokens += estimateTokens(message);
+	let skillsTokens = 0;
+	let toolsTokens = 0;
+	let systemContextTokens = 0;
+	let systemPromptTokens = 0;
+	let usedTokens = 0;
+
+	if (breakdown) {
+		messagesTokens = breakdown.messagesTokens;
+		skillsTokens = breakdown.skillsTokens;
+		toolsTokens = breakdown.systemToolsTokens;
+		systemContextTokens = breakdown.systemContextTokens;
+		systemPromptTokens = breakdown.systemPromptTokens;
+		usedTokens = breakdown.usedTokens;
+	} else {
+		const convo = session.messages;
+		if (convo) {
+			for (const message of convo) {
+				messagesTokens += estimateTokens(message);
+			}
 		}
+		const nonMessage = computeNonMessageBreakdown(session);
+		skillsTokens = nonMessage.skillsTokens;
+		toolsTokens = nonMessage.toolsTokens;
+		systemContextTokens = nonMessage.systemContextTokens;
+		systemPromptTokens = nonMessage.systemPromptTokens;
+		usedTokens = skillsTokens + toolsTokens + systemContextTokens + systemPromptTokens + messagesTokens;
 	}
 
-	// The rendered system prompt already contains the skill descriptions and the
-	// markdown tool descriptions. To present a non-overlapping breakdown:
-	//   System prompt = total system prompt text - skills section (tool descriptions stay)
-	//   Tools         = JSON tool schema sent separately on the wire
-	//   Skills        = the skill list embedded in the system prompt
-	//   Messages      = conversation messages
-	const { skillsTokens, toolsTokens, systemContextTokens, systemPromptTokens } = computeNonMessageBreakdown(session);
-
 	const categories: CategoryInfo[] = [
 		{ id: "systemPrompt", label: "System prompt", tokens: systemPromptTokens, color: "accent", glyph: CELL_FILLED },
 		{ id: "systemTools", label: "System tools", tokens: toolsTokens, color: "warning", glyph: CELL_FILLED },
@@ -155,8 +170,6 @@ export function computeContextBreakdown(
 		},
 	];
 
-	const usedTokens = categories.reduce((sum, c) => sum + c.tokens, 0);
-
 	let autoCompactBufferTokens = 0;
 	if (contextWindow > 0) {
 		const compactionSettings = session.settings.getGroup("compaction") as CompactionSettings;

package/src/prompts/tools/image-attachment-describe-system.md

@@ -0,0 +1,8 @@
+You are an image-analysis assistant. The user attached an image to a model that cannot see images, so your description is injected into that model's context in place of the image. The downstream model relies entirely on your text — it never sees the pixels.
+
+Core behavior:
+- Be faithful and evidence-first: distinguish direct observations from inferences.
+- Transcribe ALL visible text verbatim, preserving casing, punctuation, and layout order. Mark unreadable segments explicitly rather than guessing.
+- NEVER fabricate occluded, blurry, or uncertain details — say what is uncertain.
+- Be thorough but compact: prefer dense, information-rich prose over filler.
+- Do not add meta commentary, preambles ("This image shows…"), or closing remarks. Output only the description.

package/src/prompts/tools/image-attachment-describe.md

@@ -0,0 +1,10 @@
+Describe this image in enough detail that a model which cannot see it can reason about its content.
+
+Cover, where present:
+- The overall scene, subject, and what is happening.
+- People, objects, and their relationships, positions, colors, and counts.
+- All visible text, transcribed verbatim (OCR).
+- UI/screenshot elements: labels, buttons, inputs, states, errors, highlighted or disabled controls.
+- Diagrams, charts, tables: structure, axes, series, and the values they encode.
+
+Flag anything ambiguous or unreadable. Output the description as plain prose only.

package/src/sdk.ts

@@ -21,6 +21,7 @@ import {
 	getOpenAICodexTransportDetails,
 	prewarmOpenAICodexResponses,
 } from "@oh-my-pi/pi-ai/providers/openai-codex-responses";
+import { FALLBACK_DIALECT, preferredDialect } from "@oh-my-pi/pi-catalog/identity";
 import type { Component } from "@oh-my-pi/pi-tui";
 import {
 	$env,
@@ -40,7 +41,6 @@ import { AutoLearnController, buildAutoLearnInstructions } from "./autolearn/con
 import { loadCapability } from "./capability";
 import { type Rule, ruleCapability, setActiveRules } from "./capability/rule";
 import { bucketRules } from "./capability/rule-buckets";
-import { createApiKeyResolver } from "./config/api-key-resolver";
 import { shouldEnableAppendOnlyContext } from "./config/append-only-context-mode";
 import { ModelRegistry } from "./config/model-registry";
 import {
@@ -415,6 +415,8 @@ export interface CreateAgentSessionOptions {
 	providerSessionId?: string;
 	/** Optional provider-facing prompt cache key, distinct from request lineage. */
 	providerPromptCacheKey?: string;
+	/** Absolute wall-clock deadline in Unix epoch milliseconds. */
+	deadline?: number;
 
 	/** Custom tools to register (in addition to built-in tools). Accepts both CustomTool and ToolDefinition. */
 	customTools?: (CustomTool | ToolDefinition)[];
@@ -567,10 +569,15 @@ export type DialectFormat = "auto" | "native" | Dialect;
 
 export function resolveDialect(
 	format: DialectFormat,
-	model: Pick<Model, "supportsTools"> | undefined,
+	model: (Pick<Model, "supportsTools"> & Partial<Pick<Model, "id">>) | undefined,
 ): Dialect | undefined {
 	if (format === "native") return undefined;
-	if (format === "auto") return model?.supportsTools === false ? "glm" : undefined;
+	if (format === "auto") {
+		if (model?.supportsTools !== false) return undefined;
+		if (!model.id) return "glm";
+		const preferred = preferredDialect(model.id);
+		return preferred === FALLBACK_DIALECT ? "glm" : preferred;
+	}
 	return format;
 }
 
@@ -2458,6 +2465,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			onResponse,
 			sessionId: providerSessionId,
 			promptCacheKey: options.providerPromptCacheKey,
+			deadline: options.deadline,
 			transformContext,
 			transformProviderContext,
 			steeringMode: settings.get("steeringMode") ?? "one-at-a-time",
@@ -2475,28 +2483,16 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			kimiApiFormat: settings.get("providers.kimiApiFormat") ?? "anthropic",
 			preferWebsockets: preferOpenAICodexWebsockets,
 			getToolContext: tc => toolContextStore.getContext(tc),
-			getApiKey: async (provider, ctx) => {
-				// Read agent.sessionId at call time so credential selection stays aligned
-				// with metadataResolver after /new, fork, resume, or branch switches.
-				// Retry steps (ctx carries an auth error) drive the central a/b/c
-				// policy — force-refresh the same account, then rotate to a sibling —
-				// and may legitimately yield no key when every account is exhausted.
-				if (ctx?.error !== undefined) {
-					return createApiKeyResolver(modelRegistry, provider, { sessionId: agent.sessionId })(ctx);
-				}
-				const key = await modelRegistry.getApiKeyForProvider(provider, agent.sessionId);
-				if (!key) {
-					throw new Error(`No API key found for provider "${provider}"`);
-				}
-				return key;
-			},
+			getApiKey: requestModel => modelRegistry.resolver(requestModel, agent.sessionId),
 			streamFn: (streamModel, context, streamOptions) => {
 				const openrouterRoutingPreset = settings.get("providers.openrouterVariant");
 				const openrouterVariant =
 					openrouterRoutingPreset && openrouterRoutingPreset !== "default" ? openrouterRoutingPreset : undefined;
+				const antigravityEndpointMode = settings.get("providers.antigravityEndpoint");
 				return streamSimple(streamModel, context, {
 					...streamOptions,
 					openrouterVariant: streamOptions?.openrouterVariant ?? openrouterVariant,
+					antigravityEndpointMode: streamOptions?.antigravityEndpointMode ?? antigravityEndpointMode,
 				});
 			},
 			cursorExecHandlers,