indusagi 0.13.9 → 0.13.11

package/dist/types/facade/ml/adapters/openai-responses.d.ts

@@ -1,7 +1,7 @@
 import type { ResponseCreateParamsStreaming } from "openai/resources/responses/responses.js";
 import type { SimpleStreamOptions, StreamFunction, StreamOptions } from "../types.js";
 export interface OpenAIResponsesOptions extends StreamOptions {
-    reasoningEffort?: "minimal" | "low" | "medium" | "high" | "xhigh";
+    reasoningEffort?: "minimal" | "low" | "medium" | "high" | "xhigh" | "max";
     reasoningSummary?: "auto" | "detailed" | "concise" | null;
     serviceTier?: ResponseCreateParamsStreaming["service_tier"];
 }

package/dist/types/facade/ml/adapters/openai-responses.test.d.ts

@@ -0,0 +1 @@
+export {};

package/dist/types/facade/ml/adapters/simple-options.d.ts

@@ -31,8 +31,8 @@ export declare class SimpleOptionsProviderError extends Error {
 export declare function normalizeProviderError(error: unknown): SimpleOptionsProviderError;
 export declare function executeWithRetry<T>(operation: () => Promise<T>, policy: RetryPolicy): Promise<T>;
 export declare function buildBaseOptions(model: Model<Api>, options?: SimpleStreamOptions, apiKey?: string): StreamOptions;
-export declare function clampReasoning(effort: ThinkingLevel | undefined): Exclude<ThinkingLevel, "xhigh"> | undefined;
-export declare function mapThinkingLevel(level: ThinkingLevel | undefined, provider?: "supports-xhigh" | "clamp-xhigh"): ThinkingLevel | Exclude<ThinkingLevel, "xhigh"> | undefined;
+export declare function clampReasoning(effort: ThinkingLevel | undefined): Exclude<ThinkingLevel, "xhigh" | "max"> | undefined;
+export declare function mapThinkingLevel(level: ThinkingLevel | undefined, provider?: "supports-xhigh" | "clamp-xhigh"): Exclude<ThinkingLevel, "max"> | undefined;
 export declare function adjustMaxTokensForThinking(baseMaxTokens: number, modelMaxTokens: number, reasoningLevel: ThinkingLevel, customBudgets?: ThinkingBudgets): {
     maxTokens: number;
     thinkingBudget: number;

package/dist/types/facade/ml/models.generated.d.ts

@@ -4637,6 +4637,33 @@ export declare const MODELS: {
             contextWindow: number;
             maxTokens: number;
         };
+        readonly "gpt-6-astra": {
+            id: string;
+            name: string;
+            api: "openai-responses";
+            provider: string;
+            baseUrl: string;
+            reasoning: true;
+            reasoningEfforts: ("low" | "medium" | "high" | "xhigh" | "max")[];
+            input: ("text" | "image")[];
+            cost: {
+                input: number;
+                output: number;
+                cacheRead: number;
+                cacheWrite: number;
+            };
+            longContextPricing: {
+                threshold: number;
+                multipliers: {
+                    input: number;
+                    output: number;
+                    cacheRead: number;
+                    cacheWrite: number;
+                };
+            };
+            contextWindow: number;
+            maxTokens: number;
+        };
         readonly "gpt-5.6-sol": {
             id: string;
             name: string;
@@ -5098,6 +5125,33 @@ export declare const MODELS: {
             contextWindow: number;
             maxTokens: number;
         };
+        readonly "gpt-6-astra": {
+            id: string;
+            name: string;
+            api: "openai-codex-responses";
+            provider: string;
+            baseUrl: string;
+            reasoning: true;
+            reasoningEfforts: ("low" | "medium" | "high" | "xhigh" | "max")[];
+            input: ("text" | "image")[];
+            cost: {
+                input: number;
+                output: number;
+                cacheRead: number;
+                cacheWrite: number;
+            };
+            longContextPricing: {
+                threshold: number;
+                multipliers: {
+                    input: number;
+                    output: number;
+                    cacheRead: number;
+                    cacheWrite: number;
+                };
+            };
+            contextWindow: number;
+            maxTokens: number;
+        };
     };
     readonly opencode: {
         readonly "big-pickle": {

package/dist/types/facade/ml/types.d.ts

@@ -4,13 +4,14 @@ export type KnownApi = "mock-ai" | "openai-completions" | "openai-responses" | "
 export type Api = KnownApi | (string & {});
 export type KnownProvider = "mock" | "amazon-bedrock" | "anthropic" | "google" | "google-vertex" | "openai" | "azure-openai-responses" | "openai-codex" | "github-copilot" | "xai" | "groq" | "cerebras" | "openrouter" | "vercel-ai-gateway" | "zai" | "mistral" | "minimax" | "minimax-cn" | "opencode" | "kimi" | "kimi-coding" | "sarvam" | "krutrim" | "nvidia";
 export type Provider = KnownProvider | string;
-export type ThinkingLevel = "minimal" | "low" | "medium" | "high" | "xhigh";
+export type ThinkingLevel = "minimal" | "low" | "medium" | "high" | "xhigh" | "max";
 /** Per-level token allowances for reasoning, used only by token-budget providers. */
 export interface ThinkingBudgets {
     minimal?: number;
     low?: number;
     medium?: number;
     high?: number;
+    max?: number;
 }
 export interface StreamOptions {
     temperature?: number;
@@ -228,6 +229,8 @@ export interface Model<TApi extends Api> {
     provider: Provider;
     baseUrl: string;
     reasoning: boolean;
+    /** Explicit reasoning efforts accepted by this model; omitted when provider defaults apply. */
+    reasoningEfforts?: readonly Exclude<ThinkingLevel, "minimal">[];
     input: ("text" | "image")[];
     cost: {
         input: number;
@@ -235,6 +238,16 @@ export interface Model<TApi extends Api> {
         cacheRead: number;
         cacheWrite: number;
     };
+    /** Optional pricing tier applied by providers after an input-token threshold. */
+    longContextPricing?: {
+        threshold: number;
+        multipliers: {
+            input: number;
+            output: number;
+            cacheRead: number;
+            cacheWrite: number;
+        };
+    };
     contextWindow: number;
     maxTokens: number;
     headers?: Record<string, string>;

package/dist/types/llmgateway/connectors/openai-responses.test.d.ts

@@ -0,0 +1 @@
+export {};

package/dist/types/llmgateway/contract/index.d.ts

@@ -3,11 +3,11 @@
  *
  * Every downstream gateway module imports its vocabulary from this barrel.
  */
-export type { ProviderId, ApiKind, Modality, CostSheet, ModelCard, } from "./model-card";
+export type { ProviderId, ApiKind, Modality, CostSheet, LongContextPricing, ModelCard, } from "./model-card";
 export type { JsonSchema, TextBlock, ThinkingBlock, ToolCallBlock, ToolResultBlock, ImageBlock, CommandBlock, Block, UserTurn, AssistantTurn, ToolTurn, Turn, ToolDescriptor, Conversation, } from "./conversation";
 export type { Usage, StopReason, Reply } from "./reply";
 export type { TextEmission, ThinkingEmission, ToolCallStartEmission, ToolCallDeltaEmission, UsageEmission, StopEmission, DoneEmission, ErrorEmission, Emission, Channel, } from "./emission";
-export type { ThinkingLevel, ToolChoice, StreamOptions, } from "./options";
+export type { ThinkingLevel, ReasoningEffort, ToolChoice, StreamOptions, } from "./options";
 export type { GatewayErrorKind, GatewayErrorExtra, } from "./errors";
 export { GatewayError, gatewayError } from "./errors";
 export type { Connector, CredentialResolver } from "./connector";

package/dist/types/llmgateway/contract/model-card.d.ts

@@ -5,6 +5,7 @@
  * behavior; connectors read it to shape their wire requests, and the pricing
  * helpers read it to turn token counts into money.
  */
+import type { ReasoningEffort } from "./options";
 /**
  * The upstream organization that owns a model. This is the *vendor* identity,
  * independent of which HTTP dialect we speak to reach it (see {@link ApiKind}).
@@ -25,6 +26,16 @@ export interface CostSheet {
     readonly cacheReadPerMTok?: number;
     readonly cacheWritePerMTok?: number;
 }
+/** Optional higher-price tier triggered by a long input context. */
+export interface LongContextPricing {
+    readonly threshold: number;
+    readonly multipliers: {
+        readonly input: number;
+        readonly output: number;
+        readonly cacheRead: number;
+        readonly cacheWrite: number;
+    };
+}
 /**
  * Static descriptor for one model. Identified by `id` within a `provider`.
  */
@@ -38,5 +49,8 @@ export interface ModelCard {
     readonly maxOutputTokens: number;
     readonly modalities: readonly Modality[];
     readonly reasoning: boolean;
+    /** Explicit wire efforts for models whose accepted set differs from provider defaults. */
+    readonly reasoningEfforts?: readonly ReasoningEffort[];
+    readonly longContextPricing?: LongContextPricing;
     readonly cost: CostSheet;
 }

package/dist/types/llmgateway/contract/options.d.ts

@@ -9,7 +9,9 @@
  * How much reasoning budget to grant a reasoning-capable model.
  * `off` disables thinking entirely; the remaining levels scale the budget up.
  */
-export type ThinkingLevel = "off" | "low" | "medium" | "high" | "max";
+export type ThinkingLevel = "off" | "low" | "medium" | "high" | "xhigh" | "max";
+/** Reasoning efforts that can be sent on the wire (excluding disabled reasoning). */
+export type ReasoningEffort = Exclude<ThinkingLevel, "off">;
 /** Constrains how the model may use the declared tools on this turn. */
 export type ToolChoice = "auto" | "required" | "none";
 /** Options for a single streaming or one-shot invocation. */

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "indusagi",
-  "version": "0.13.9",
+  "version": "0.13.11",
   "description": "Indusagi — a terminal-first AI coding agent framework. Clean-room implementation.",
   "author": "Varun Israni",
   "license": "AGPL-3.0",
@@ -157,6 +157,7 @@
     "marked": "^18.0.4",
     "openai": "^6.41.0",
     "partial-json": "^0.1.7",
+    "proxy-agent": "^8.0.2",
     "react": "^18.3.1",
     "strip-ansi": "^7.2.0",
     "ulid": "^3.0.2",