Skip to main content ホーム クリエイター jeremylongshore tons-of-skills-marketplace assemblyai-performance-tuning
assemblyai-performance-tuning Optimize AssemblyAI API performance with caching, parallel processing, and model selection.
Use when experiencing slow transcriptions, implementing caching strategies,
or optimizing throughput for batch transcription workloads.
Trigger with phrases like "assemblyai performance", "optimize assemblyai",
"assemblyai latency", "assemblyai caching", "assemblyai slow", "assemblyai batch".
インストールへ移動 Skills Marketplace コミュニティが作成したAIスキルを発見・探索
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/jeremylongshore/tons-of-skills-marketplace --skill assemblyai-performance-tuningコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Zipをダウンロード ダウンロード中... このリポジトリの他の Skills langchain-deploy-integration Deploy a LangChain 1.0 / LangGraph 1.0 app to Cloud Run, Vercel, or LangServe correctly — with timeouts sized for chain length, cold-start mitigation, SSE anti-buffering headers, and Secret Manager over .env. Use when prepping a first production deploy, debugging a stream that hangs behind a proxy, or diagnosing p99 latency spikes. Trigger with "langchain deploy", "langchain cloud run", "langchain vercel python", "langchain langserve", or "langchain docker".
langchain-langgraph-agents Build a correct LangGraph 1.0 ReAct agent with create_react_agent — typed tools, error propagation, recursion caps, and stop conditions that actually stop. Use when writing a first tool-calling agent, migrating from AgentExecutor or initialize_agent, or diagnosing an agent that loops on vague prompts. Trigger with "langgraph agent", "create_react_agent", "langgraph tool calling", "AgentExecutor migration", or "agent loop cost".
langchain-langgraph-human-in-loop Build LangGraph 1.0 human-in-the-loop approval flows with interrupt_before /
interrupt_after and Command(resume=...) — JSON-serializable state, clean
resume semantics, and UI wiring for approval decisions. Use when adding an
approval gate before an expensive tool call, wiring a Slack/web UI for agent
approvals, or debugging a graph that crashes on interrupt.
Trigger with "langgraph human in loop", "langgraph interrupt_before",
"langgraph approval flow", "Command resume", "langgraph HITL".
name assemblyai-performance-tuning description Optimize AssemblyAI API performance with caching, parallel processing, and model selection.
Use when experiencing slow transcriptions, implementing caching strategies,
or optimizing throughput for batch transcription workloads.
Trigger with phrases like "assemblyai performance", "optimize assemblyai",
"assemblyai latency", "assemblyai caching", "assemblyai slow", "assemblyai batch".
allowed-tools Read, Write, Edit version 1.5.0 license MIT author Jeremy Longshore <jeremy@intentsolutions.io> tags ["saas","ai","speech-to-text","assemblyai","transcription","performance"] compatibility Designed for Claude Code
AssemblyAI Performance Tuning
Overview
Optimize AssemblyAI transcription performance through model selection, parallel processing, caching, and webhook-based architectures.
Prerequisites
assemblyai package installedUnderstanding of async patterns
Redis or in-memory cache available (optional)
Latency Benchmarks (Actual)
Async Transcription
Audio Duration Approx. Processing Time Notes 30 seconds ~10-15 seconds Includes queue time 5 minutes ~30-60 seconds Scales sub-linearly 1 hour ~3-5 minutes Depends on queue load 10 hours ~15-30 minutes Max async duration
Streaming
Metric Value First partial transcript ~300ms (P50) Final transcript latency ~500ms (P50) End-of-turn detection Automatic with endpointing
Model Speed vs. Accuracy
Model Speed Accuracy Price/hr nanoFastest Good $0.12 best (Universal-3)Standard Highest $0.37 nova-3 (streaming)Real-time High $0.47 nova-3-pro (streaming)Real-time Highest $0.47
Instructions
Step 1: Choose the Right Model
import { AssemblyAI } from 'assemblyai' ;
const client = new AssemblyAI ({
apiKey
env
ASSEMBLYAI_API_KEY
const
await
transcripts
transcribe
audio
speech_model
'best'
const
await
transcripts
transcribe
audio
speech_model
'nano'
Step 2: Parallel Batch Processing import PQueue from 'p-queue' ;
const queue = new PQueue ({ concurrency : 10 });
async function batchTranscribe (audioUrls : string []const results = await Promise .all (
audioUrls.map (url =>add (() =>
client.transcripts .transcribe ({ audio : url, speech_model : 'nano' })
)
)
);
return results.filter (t =>status === 'completed' );
}
const urls = Array .from ({ length : 100 }, (_, i ) => `https://storage.example.com/audio-${i} .mp3` );
const transcripts = await batchTranscribe (urls);
console .log (`Completed: ${transcripts.length} /${urls.length} ` );
Step 3: Use Webhooks Instead of Polling
const slow = await client.transcripts .transcribe ({ audio : audioUrl });
const fast = await client.transcripts .submit ({
audio : audioUrl,
webhook_url : 'https://your-app.com/webhooks/assemblyai' ,
});
Step 4: Cache Transcript Results import { LRUCache } from 'lru-cache' ;
import type { Transcript } from 'assemblyai' ;
const transcriptCache = new LRUCache <string , Transcript >({
max : 500 ,
ttl : 60 * 60 * 1000 ,
});
async function getCachedTranscript (transcriptId : string Promise <Transcript > {
const cached = transcriptCache.get (transcriptId);
if (cached) return cached;
const transcript = await client.transcripts .get (transcriptId);
if (transcript.status === 'completed' ) {
transcriptCache.set (transcriptId, transcript);
}
return transcript;
}
Step 5: Redis Cache for Distributed Systems import Redis from 'ioredis' ;
const redis = new Redis (process.env .REDIS_URL !);
async function getCachedTranscriptRedis (transcriptId : string Promise <Transcript > {
const cached = await redis.get (`transcript:${transcriptId} ` );
if (cached) return JSON .parse (cached);
const transcript = await client.transcripts .get (transcriptId);
if (transcript.status === 'completed' ) {
await redis.setex (
`transcript:${transcriptId} ` ,
3600 ,
JSON .stringify (transcript)
);
}
return transcript;
}
Step 6: Minimize Feature Overhead
const minimal = await client.transcripts .transcribe ({
audio : audioUrl,
speech_model : 'nano' ,
punctuate : true ,
format_text : true ,
});
const full = await client.transcripts .transcribe ({
audio : audioUrl,
speech_model : 'best' ,
speaker_labels : true ,
sentiment_analysis : true ,
entity_detection : true ,
auto_highlights : true ,
content_safety : true ,
iab_categories : true ,
summarization : true ,
summary_type : 'bullets' ,
});
Step 7: Performance Monitoring async function timedTranscribe (audioUrl : string , options : Record <string , any > = {}const start = Date .now ();
const transcript = await client.transcripts .transcribe ({
audio : audioUrl,
...options,
});
const durationMs = Date .now () - start;
const stats = {
transcriptId : transcript.id ,
status : transcript.status ,
audioDuration : transcript.audio_duration ,
processingTimeMs : durationMs,
ratio : transcript.audio_duration
? (durationMs / 1000 / transcript.audio_duration ).toFixed (2 )
: 'N/A' ,
wordCount : transcript.words ?.length ?? 0 ,
model : options.speech_model ?? 'best' ,
};
console .log ('Transcription stats:' , stats);
return { transcript, stats };
}
Output
Optimal model selection based on speed/accuracy/cost trade-offs
Parallel batch processing with concurrency control
Webhook-based architecture (eliminates polling overhead)
In-memory and Redis caching for transcript retrieval
Performance monitoring with processing time ratios
Error Handling Issue Cause Solution Slow transcription Large file + best model Use nano model or split audio Queue backlog Too many concurrent submissions Limit concurrency with p-queue Cache stale data Transcript re-processed Set appropriate TTL, invalidate on webhook Polling overhead Using transcribe() for many files Switch to submit() + webhooks
Resources
Next Steps For cost optimization, see assemblyai-cost-tuning.
