Klingai Reference Architecture

What it does

When your agent uses it

• Designing scalable systems
• With phrases like klingai architecture
• Kling ai system design
• Video platform architecture

Example prompts

• “klingai architecture”
• “kling ai system design”
• “video platform architecture”
• “/klingai-reference-architecture”

Requirements

• Python 3
• Docker
• A credential in KLING_ACCESS_KEY
• A credential in KLING_SECRET_KEY
• Compatibility (from SKILL.md): Designed for Claude Code
• Pre-approved tools (allowed-tools): Read, Write, Edit, Bash(npm:*), Grep

Workflow steps

6 steps, taken from the first numbered list in SKILL.md.

1. 1Keep the API gateway responsible for authentication, authorization, prompt and provenance validation, content-policy checks, destination…
2. 2Put only opaque job references and approved parameters on the queue. Workers obtain short-lived credentials from the secret manager…
3. 3Start each release with a watermarked sandbox canary. Verify policy, source rights, suppression/destination rules, output integrity…
4. 4Store generated media under encrypted, access-controlled paths with a retention deadline. Keep logs and events redacted; never copy…
5. 5Promote by immutable digest and record the approval. On policy, quality, budget, storage, or provider failure, stop the queue, quarantine…
6. 6Test rollback and deletion in staging, then retain a receipt containing only opaque IDs, hashes, aggregate metrics, approval state…

What it can do on your machine

Read from SKILL.md and the folder at commit cfae287. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves these tools, so the agent can use them without asking each time:

  • Read
  • Write
  • Edit
  • Bash(npm:*)
  • Grep

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  No scripts in the folder and no shell commands in SKILL.md (its code samples are python and yaml).

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  Hosts in commands or code, which the agent is likely to contact:

  • api.klingai.com

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names these keys or tokens, usually read from environment variables:

  • KLING_ACCESS_KEY
  • KLING_SECRET_KEY

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

• Compatibility

  Designed for Claude Code

  From compatibility in the SKILL.md frontmatter.

Context cost

Klingai Reference Architecture loads about 2.2k tokens when it runs, and up to ~5.8k if it reads all its reference files. Until then it costs about 69 tokens; SKILL.md has 518 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~69

When it runs · the whole SKILL.md, loaded when a task matches

~2.2k

With references · SKILL.md plus every file in references/, read only if the agent opens them

~5.8k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from jeremylongshore/tons-of-skills-marketplace at commit cfae287, republished under its MIT licence (© jeremylongshore). 518 words, ~2,163 tokens.

name

klingai-reference-architecture

description

Production reference architecture for Kling AI video generation platforms. Use when designing scalable systems. Trigger with phrases like 'klingai architecture', 'kling ai system design', 'video platform architecture', 'klingai production setup'.

allowed-tools

Read, Write, Edit, Bash(npm:*), Grep

compatibility

Designed for Claude Code

version

1.18.0

license

MIT

author

Jeremy Longshore <jeremy@intentsolutions.io>

tags

saas, kling-ai, architecture, scaling

User Request
    |
[API Gateway / Load Balancer]
    |
[Application Server]
    |--- validate prompt & estimate cost
    |--- enqueue job to Redis/SQS
    |
[Job Queue (Redis / SQS / Pub/Sub)]
    |
[Worker Pool (N workers)]
    |--- generate JWT token
    |--- POST https://api.klingai.com/v1/videos/text2video
    |--- receive task_id
    |--- register callback_url OR poll
    |
[Webhook Receiver / Poller]
    |--- receive completion callback
    |--- download video from Kling CDN
    |--- upload to S3/GCS
    |--- update job status in DB
    |--- notify user
    |
[Object Storage (S3 / GCS)]
    |
[CDN (CloudFront / Cloud CDN)]
    |
User views video

from fastapi import FastAPI, HTTPException
from pydantic import BaseModel

app = FastAPI()

class VideoRequest(BaseModel):
    prompt: str
    model: str = "kling-v2-master"
    duration: int = 5
    mode: str = "standard"

@app.post("/api/videos")
async def create_video(req: VideoRequest):
    # 1. Validate
    if len(req.prompt) > 2500:
        raise HTTPException(400, "Prompt exceeds 2500 chars")

    # 2. Estimate cost
    credits = estimate_credits(req.duration, req.mode)
    if not budget_guard.check(credits):
        raise HTTPException(402, "Budget exceeded")

    # 3. Enqueue
    job_id = await queue.enqueue({
        "prompt": req.prompt,
        "model": req.model,
        "duration": str(req.duration),
        "mode": req.mode,
    })

    return {"job_id": job_id, "status": "queued", "estimated_credits": credits}

import redis
import json

class VideoWorker:
    def __init__(self, kling_client, storage_client, redis_url="redis://localhost"):
        self.kling = kling_client
        self.storage = storage_client
        self.redis = redis.Redis.from_url(redis_url)

    def process_loop(self):
        while True:
            raw = self.redis.brpop("kling:jobs:pending", timeout=5)
            if not raw:
                continue

            job = json.loads(raw[1])
            try:
                # Submit to Kling API
                result = self.kling.text_to_video(
                    job["prompt"],
                    model=job["model"],
                    duration=int(job["duration"]),
                    mode=job["mode"],
                    callback_url=os.environ.get("WEBHOOK_URL"),
                )

                # If using polling (no callback)
                if isinstance(result, dict) and "videos" in result:
                    video_url = result["videos"][0]["url"]
                    stored_url = self.storage.download_and_upload(video_url, job["id"])
                    self.redis.publish("kling:events", json.dumps({
                        "type": "completed",
                        "job_id": job["id"],
                        "video_url": stored_url,
                    }))

            except Exception as e:
                self.redis.lpush("kling:jobs:failed", json.dumps({
                    **job, "error": str(e)
                }))

# docker-compose.yml
services:
  api:
    build: ./api
    ports: ["8000:8000"]
    environment:
      - REDIS_URL=redis://redis:6379
      - KLING_ACCESS_KEY=${KLING_ACCESS_KEY}
      - KLING_SECRET_KEY=${KLING_SECRET_KEY}

  worker:
    build: ./worker
    deploy:
      replicas: 2
    environment:
      - REDIS_URL=redis://redis:6379
      - KLING_ACCESS_KEY=${KLING_ACCESS_KEY}
      - KLING_SECRET_KEY=${KLING_SECRET_KEY}
      - S3_BUCKET=${S3_BUCKET}

  webhook:
    build: ./webhook
    ports: ["8001:8001"]
    environment:
      - REDIS_URL=redis://redis:6379

  redis:
    image: redis:7-alpine
    volumes: ["redis-data:/data"]

volumes:
  redis-data:

artifact=sha256:opaque; environment=staging; fixture=synthetic-v4;
rights=cleared; policy=pass; canary=watermarked-private;
budget=within-limit; output_digest=sha256:opaque; approval=recorded;
retention=24h; deletion=verified; rollback=release-r31

© jeremylongshore, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 6 other files (references) in skills/.curated/klingai-reference-architecture of jeremylongshore/tons-of-skills-marketplace.

• SKILL.md
• references/architecture-patterns.md
• references/complete-reference-implementation.md
• references/docker-compose-setup.md
• references/errors.md
• references/examples.md
• references/kubernetes-deployment.md

Open the folder on GitHubat commit cfae287