@openmeter/sdk 1.0.0-beta.9 → 1.0.0-beta.91

package/dist/next.js

@@ -1,46 +0,0 @@
-let encodingForModel;
-export async function createOpenAIStreamCallback({ model, prompts, }, openAIStreamCallbacks) {
-    // Tiktoken is an optional dependency, so we import it conditionally
-    if (!encodingForModel) {
-        const { encodingForModel: encodingForModel_ } = await import('js-tiktoken');
-        encodingForModel = encodingForModel_;
-    }
-    const enc = encodingForModel(model);
-    let promptTokens = 0;
-    let completionTokens = 0;
-    const streamCallbacks = {
-        ...openAIStreamCallbacks,
-        async onStart() {
-            for (const content of prompts) {
-                const tokens = enc.encode(content);
-                promptTokens += tokens.length;
-            }
-            if (typeof openAIStreamCallbacks?.onStart === 'function') {
-                return openAIStreamCallbacks.onStart();
-            }
-        },
-        async onToken(content) {
-            // To test tokenizaton see: https://platform.openai.com/tokenizer
-            const tokens = enc.encode(content);
-            completionTokens += tokens.length;
-            if (typeof openAIStreamCallbacks?.onToken === 'function') {
-                return openAIStreamCallbacks.onToken(content);
-            }
-        },
-        async onFinal(completion) {
-            // Mimicking OpenAI usage metadata API
-            const usage = {
-                total_tokens: promptTokens + completionTokens,
-                prompt_tokens: promptTokens,
-                completion_tokens: completionTokens,
-            };
-            if (typeof openAIStreamCallbacks?.onUsage === 'function') {
-                await openAIStreamCallbacks.onUsage(usage);
-            }
-            if (typeof openAIStreamCallbacks?.onFinal === 'function') {
-                return openAIStreamCallbacks.onFinal(completion);
-            }
-        },
-    };
-    return streamCallbacks;
-}

package/index.ts

@@ -1,24 +0,0 @@
-import { OpenMeterConfig } from './clients/client.js'
-import { EventsClient } from './clients/event.js'
-import { MetersClient } from './clients/meter.js'
-
-export { OpenMeterConfig, RequestOptions } from './clients/client.js'
-export { Event } from './clients/event.js'
-export {
-  Meter,
-  MeterValue,
-  MeterAggregation,
-  WindowSize,
-} from './clients/meter.js'
-
-export { createOpenAIStreamCallback } from './next.js'
-
-export class OpenMeter {
-  public events: EventsClient
-  public meters: MetersClient
-
-  constructor(config: OpenMeterConfig) {
-    this.events = new EventsClient(config)
-    this.meters = new MetersClient(config)
-  }
-}

package/next.ts

@@ -1,76 +0,0 @@
-import type { OpenAIStreamCallbacks } from 'ai'
-import type { TiktokenModel } from 'js-tiktoken'
-
-let encodingForModel: (model: TiktokenModel) => any | undefined
-
-type OpenAIUsage = {
-  total_tokens: number
-  prompt_tokens: number
-  completion_tokens: number
-}
-
-type OpenAIStreamCallbacksWithUsage = OpenAIStreamCallbacks & {
-  onUsage?: (usage: OpenAIUsage) => Promise<void> | void
-}
-
-export async function createOpenAIStreamCallback(
-  {
-    model,
-    prompts,
-  }: {
-    model: TiktokenModel
-    prompts: string[]
-  },
-  openAIStreamCallbacks: OpenAIStreamCallbacksWithUsage
-) {
-  // Tiktoken is an optional dependency, so we import it conditionally
-  if (!encodingForModel) {
-    const { encodingForModel: encodingForModel_ } = await import('js-tiktoken')
-    encodingForModel = encodingForModel_
-  }
-
-  const enc = encodingForModel(model)
-  let promptTokens = 0
-  let completionTokens = 0
-
-  const streamCallbacks: OpenAIStreamCallbacks = {
-    ...openAIStreamCallbacks,
-
-    async onStart() {
-      for (const content of prompts) {
-        const tokens = enc.encode(content)
-        promptTokens += tokens.length
-      }
-
-      if (typeof openAIStreamCallbacks?.onStart === 'function') {
-        return openAIStreamCallbacks.onStart()
-      }
-    },
-    async onToken(content) {
-      // To test tokenizaton see: https://platform.openai.com/tokenizer
-      const tokens = enc.encode(content)
-      completionTokens += tokens.length
-
-      if (typeof openAIStreamCallbacks?.onToken === 'function') {
-        return openAIStreamCallbacks.onToken(content)
-      }
-    },
-    async onFinal(completion: string) {
-      // Mimicking OpenAI usage metadata API
-      const usage: OpenAIUsage = {
-        total_tokens: promptTokens + completionTokens,
-        prompt_tokens: promptTokens,
-        completion_tokens: completionTokens,
-      }
-
-      if (typeof openAIStreamCallbacks?.onUsage === 'function') {
-        await openAIStreamCallbacks.onUsage(usage)
-      }
-
-      if (typeof openAIStreamCallbacks?.onFinal === 'function') {
-        return openAIStreamCallbacks.onFinal(completion)
-      }
-    },
-  }
-  return streamCallbacks
-}