@probeo/anymodel 0.4.0 → 0.5.1

package/dist/index.d.cts

@@ -77,6 +77,7 @@ interface ChatCompletionRequest {
     tools?: Tool[];
     tool_choice?: ToolChoice;
     user?: string;
+    service_tier?: 'auto' | 'flex';
     models?: string[];
     route?: 'fallback';
     transforms?: string[];
@@ -455,6 +456,7 @@ declare class BatchManager {
     private syncNativeBatchStatus;
     /**
      * Process batch requests concurrently (fallback path).
+     * Streams requests from disk to avoid holding them all in memory.
      */
     private processConcurrentBatch;
 }
@@ -574,6 +576,10 @@ declare class BatchStore {
      * List all batch IDs.
      */
     listBatches(): Promise<string[]>;
+    /**
+     * Stream requests from JSONL one line at a time (memory-efficient).
+     */
+    streamRequests(id: string): AsyncGenerator<unknown>;
     /**
      * Check if a batch exists.
      */
@@ -629,4 +635,21 @@ declare function createOpenAIBatchAdapter(apiKey: string): BatchAdapter;
 
 declare function createAnthropicBatchAdapter(apiKey: string): BatchAdapter;
 
-export { AnyModel, type AnyModelConfig, AnyModelError, type AnyModelErrorMetadata, type BatchAdapter, type BatchCreateRequest, BatchManager, type BatchMode, type BatchObject, type BatchPollOptions, type BatchRequestItem, type BatchResultItem, type BatchResults, type BatchStatus, BatchStore, type BatchUsageSummary, type ChatCompletion, type ChatCompletionChoice, type ChatCompletionChunk, type ChatCompletionRequest, type ChunkChoice, type ChunkDelta, type ContentPart, type CustomProviderConfig, type FinishReason, type GenerationStats, GenerationStatsStore, type Message, type ModelArchitecture, type ModelInfo, type ModelPricing, type ModelTopProvider, type NativeBatchStatus, type ProviderAdapter, type ProviderConfig, type ProviderPreferences, type ResponseFormat, type Role, type ServerOptions, type Tool, type ToolCall, type ToolChoice, type Usage, appendFileQueued, configureFsIO, createAnthropicBatchAdapter, createAnyModelServer, createOpenAIBatchAdapter, ensureDir, getFsQueueStatus, joinPath, readFileQueued, resolveConfig, startServer, waitForFsQueuesIdle, writeFileFlushedQueued, writeFileQueued };
+declare function createGoogleBatchAdapter(apiKey: string): BatchAdapter;
+
+/**
+ * Rough token estimation and model-limit lookups for automatic max_tokens calculation.
+ */
+/** Estimate the number of tokens in a string (~4 chars per token). */
+declare function estimateTokenCount(text: string): number;
+/**
+ * Determine the best max_tokens value for a request.
+ *
+ * - If the caller already supplied a value, return it unchanged.
+ * - Otherwise estimate input tokens, apply a 5 % safety margin, then
+ *   return min(maxCompletionTokens, contextLength - estimatedInput).
+ * - The result is clamped to at least 1.
+ */
+declare function resolveMaxTokens(model: string, messages: unknown[], userMaxTokens?: number): number;
+
+export { AnyModel, type AnyModelConfig, AnyModelError, type AnyModelErrorMetadata, type BatchAdapter, type BatchCreateRequest, BatchManager, type BatchMode, type BatchObject, type BatchPollOptions, type BatchRequestItem, type BatchResultItem, type BatchResults, type BatchStatus, BatchStore, type BatchUsageSummary, type ChatCompletion, type ChatCompletionChoice, type ChatCompletionChunk, type ChatCompletionRequest, type ChunkChoice, type ChunkDelta, type ContentPart, type CustomProviderConfig, type FinishReason, type GenerationStats, GenerationStatsStore, type Message, type ModelArchitecture, type ModelInfo, type ModelPricing, type ModelTopProvider, type NativeBatchStatus, type ProviderAdapter, type ProviderConfig, type ProviderPreferences, type ResponseFormat, type Role, type ServerOptions, type Tool, type ToolCall, type ToolChoice, type Usage, appendFileQueued, configureFsIO, createAnthropicBatchAdapter, createAnyModelServer, createGoogleBatchAdapter, createOpenAIBatchAdapter, ensureDir, estimateTokenCount, getFsQueueStatus, joinPath, readFileQueued, resolveConfig, resolveMaxTokens, startServer, waitForFsQueuesIdle, writeFileFlushedQueued, writeFileQueued };

package/dist/index.d.ts

@@ -77,6 +77,7 @@ interface ChatCompletionRequest {
     tools?: Tool[];
     tool_choice?: ToolChoice;
     user?: string;
+    service_tier?: 'auto' | 'flex';
     models?: string[];
     route?: 'fallback';
     transforms?: string[];
@@ -455,6 +456,7 @@ declare class BatchManager {
     private syncNativeBatchStatus;
     /**
      * Process batch requests concurrently (fallback path).
+     * Streams requests from disk to avoid holding them all in memory.
      */
     private processConcurrentBatch;
 }
@@ -574,6 +576,10 @@ declare class BatchStore {
      * List all batch IDs.
      */
     listBatches(): Promise<string[]>;
+    /**
+     * Stream requests from JSONL one line at a time (memory-efficient).
+     */
+    streamRequests(id: string): AsyncGenerator<unknown>;
     /**
      * Check if a batch exists.
      */
@@ -629,4 +635,21 @@ declare function createOpenAIBatchAdapter(apiKey: string): BatchAdapter;
 
 declare function createAnthropicBatchAdapter(apiKey: string): BatchAdapter;
 
-export { AnyModel, type AnyModelConfig, AnyModelError, type AnyModelErrorMetadata, type BatchAdapter, type BatchCreateRequest, BatchManager, type BatchMode, type BatchObject, type BatchPollOptions, type BatchRequestItem, type BatchResultItem, type BatchResults, type BatchStatus, BatchStore, type BatchUsageSummary, type ChatCompletion, type ChatCompletionChoice, type ChatCompletionChunk, type ChatCompletionRequest, type ChunkChoice, type ChunkDelta, type ContentPart, type CustomProviderConfig, type FinishReason, type GenerationStats, GenerationStatsStore, type Message, type ModelArchitecture, type ModelInfo, type ModelPricing, type ModelTopProvider, type NativeBatchStatus, type ProviderAdapter, type ProviderConfig, type ProviderPreferences, type ResponseFormat, type Role, type ServerOptions, type Tool, type ToolCall, type ToolChoice, type Usage, appendFileQueued, configureFsIO, createAnthropicBatchAdapter, createAnyModelServer, createOpenAIBatchAdapter, ensureDir, getFsQueueStatus, joinPath, readFileQueued, resolveConfig, startServer, waitForFsQueuesIdle, writeFileFlushedQueued, writeFileQueued };
+declare function createGoogleBatchAdapter(apiKey: string): BatchAdapter;
+
+/**
+ * Rough token estimation and model-limit lookups for automatic max_tokens calculation.
+ */
+/** Estimate the number of tokens in a string (~4 chars per token). */
+declare function estimateTokenCount(text: string): number;
+/**
+ * Determine the best max_tokens value for a request.
+ *
+ * - If the caller already supplied a value, return it unchanged.
+ * - Otherwise estimate input tokens, apply a 5 % safety margin, then
+ *   return min(maxCompletionTokens, contextLength - estimatedInput).
+ * - The result is clamped to at least 1.
+ */
+declare function resolveMaxTokens(model: string, messages: unknown[], userMaxTokens?: number): number;
+
+export { AnyModel, type AnyModelConfig, AnyModelError, type AnyModelErrorMetadata, type BatchAdapter, type BatchCreateRequest, BatchManager, type BatchMode, type BatchObject, type BatchPollOptions, type BatchRequestItem, type BatchResultItem, type BatchResults, type BatchStatus, BatchStore, type BatchUsageSummary, type ChatCompletion, type ChatCompletionChoice, type ChatCompletionChunk, type ChatCompletionRequest, type ChunkChoice, type ChunkDelta, type ContentPart, type CustomProviderConfig, type FinishReason, type GenerationStats, GenerationStatsStore, type Message, type ModelArchitecture, type ModelInfo, type ModelPricing, type ModelTopProvider, type NativeBatchStatus, type ProviderAdapter, type ProviderConfig, type ProviderPreferences, type ResponseFormat, type Role, type ServerOptions, type Tool, type ToolCall, type ToolChoice, type Usage, appendFileQueued, configureFsIO, createAnthropicBatchAdapter, createAnyModelServer, createGoogleBatchAdapter, createOpenAIBatchAdapter, ensureDir, estimateTokenCount, getFsQueueStatus, joinPath, readFileQueued, resolveConfig, resolveMaxTokens, startServer, waitForFsQueuesIdle, writeFileFlushedQueued, writeFileQueued };