Chatplayground

🌍 View in other languages

ChatPlayground Provider Integration

OmniRoute supports ChatPlayground (chatplayground, alias cpl), allowing access to diverse model endpoints (Azure, LMSYS, and Perplexity) routed through ChatPlayground with automated Clerk session JWT minting and model discovery.

---

Authentication & Cookie Setup

ChatPlayground requires short-lived Clerk session JWTs (Authorization: Bearer <jwt>) for all API requests. OmniRoute handles the entire token lifecycle automatically via browserless auto-minting.

Required Cookies

To configure a ChatPlayground connection, you only need two cookies, both obtainable directly from web.chatplayground.ai (you do not need to inspect clerk.chatplayground.ai separately):

1. __client: Long-lived Clerk client authorization token.
2. __session: A Clerk JWT (eyJhbGciOi...). OmniRoute automatically decodes its payload to extract your active session ID (sid: "sess_...").

How to Capture Your Cookies

1. Log in to your account at https://web.chatplayground.ai/.
2. Open Browser DevTools (F12 or right-click → Inspect).
3. Use either of the following methods:

Method A (Easiest — Network Tab)

• Switch to the Network tab in DevTools.
• Refresh the page or click any action (e.g. model picker or settings).
• Click on any request sent to web.chatplayground.ai or app.chatplayground.ai (such as models or user).
• In the right-hand panel, go to Headers → Request Headers.
• Copy the entire value of the Cookie: header and paste it into OmniRoute's connection credential field. OmniRoute automatically extracts __client and __session.

Method B (Application Tab)

• In DevTools, go to Application (Chrome/Edge) or Storage (Firefox) → Cookies → https://web.chatplayground.ai.
• Copy the values for __client and __session.
• Paste them into OmniRoute formatted as:

  __client=<value>; __session=<value>
  

---

Token Minting & Rotation Lifecycle

1. Auto-Minting:

   • OmniRoute extracts sid from __session and posts to Clerk's token mint endpoint (https://clerk.chatplayground.ai/v1/client/sessions/{sid}/tokens) using __client.
   • The returned short-lived Bearer JWT is cached in memory.
   • OmniRoute decodes the JWT's exp claim and automatically mints a fresh token ~15 seconds before expiration.

2. Direct Bearer JWT:

   • Alternatively, you can paste an active Clerk Bearer JWT directly.

3. Multi-Account Management (Native OmniRoute):

   • To use multiple ChatPlayground accounts, add them as separate provider connections in the OmniRoute dashboard (e.g. ChatPlayground 1, ChatPlayground 2).
   • OmniRoute's native Combo Router (round-robin, priority, least-used, headroom, reset-aware) automatically distributes queries across accounts.
   • If an account encounters rate limits (429) or quota limits, OmniRoute's automated account fallback system (open-sse/services/accountFallback.ts) switches to the next healthy account transparently.
   • Each connection independently monitors its own usage, plan, and daily/monthly allowances on the dashboard Limits page.

---

Supported Endpoints & Models

ChatPlayground routes models to three distinct backend paths:

1. Azure Endpoint (https://app.chatplayground.ai/api/chat/azure)

• GPT: gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna, gpt-5.5, gpt-5.5-pro, gpt-4.5, gpt-4o
• Claude: claude-sonnet-5, claude-sonnet-4-6, claude-opus-4-8, claude-opus-4-6, claude-haiku-4-5
• Gemini: gemini-3-flash, gemini-3-pro
• DeepSeek: deepseek-v4-pro, deepseek-v4-flash
• Mistral: mistral-large-3

2. LMSYS Endpoint (https://app.chatplayground.ai/api/chat/lmsys)

• Kimi: kimi-k3, kimi-k2.6
• Llama: llama-4-scout, llama-3.3-70b
• Qwen: qwen3.8-max, qwen3.7-plus
• Grok: grok-4.5, grok-4
• Other: command-a, minimax-m3, glm-5

3. Perplexity Endpoint (https://app.chatplayground.ai/api/chat/perplexity)

• perplexity-sonar-pro / sonar-pro
• sonar

Dynamic Model Discovery, Pricing Plans & Credit Weights

Official Subscription Plans (chatplayground.ai/checkout)

1. Pro ($15/mo):
   • 5,000 basic model queries / month
   • 1,500 advanced model queries / month
   • 300 document uploads / month
   • 300 image generations / month
   • Compare 4 models side-by-side
2. Unlimited ($25/mo):
   • Unlimited basic model queries / month
   • Unlimited advanced model queries / month (enforced via a fair-use sliding limit of 300 credits daily)
   • Unlimited document uploads / month
   • 1,000 image generations / month
   • Compare 6 models side-by-side

StackSocial Lifetime Subscription Tiers

ChatPlayground also offers lifetime deals on platforms like StackSocial:

1. Basic Plan (Lifetime):
   • 500 messages / month
   • Geared toward casual prompting and exploration
2. Pro Plan (Lifetime):
   • 2,000 messages / month
   • Geared toward higher-volume workflows
3. Unlimited Plan (Lifetime):
   • Unlimited messages / month (subject to fair use: 300 credits daily)
   • Access to full AI model catalog side-by-side

Upstream API Flags (GET https://app.chatplayground.ai/api/models)

While ChatPlayground's public marketing organizes models into "Basic" vs "Advanced", the internal API exposes:

• premiumOnly: boolean: Upstream subscriber-gate boolean. Models with premiumOnly: true (e.g. gpt-5.6-sol, claude-sonnet-5, claude-opus-4-8, gpt-5.5-pro) require an active paid plan (Pro or Unlimited/Lifetime). When accessed from a free account, the API rejects with HTTP 403: "This model is only available to active subscribers. Please upgrade your plan." Models with premiumOnly: false (e.g. gpt-5.6-terra, gpt-5.6-luna, deepseek-v4-pro, grok-4.5) are accessible across all tiers.
• creditWeight: number: Deductions multiplier per query against your plan allowance:
  • 0.5: Lightweight models (gemini-3.5-flash-lite, deepseek-v4-flash, sonar)
  • 1.0: Standard models (gpt-5.6-terra, gpt-5.6-luna, deepseek-v4-pro, kimi-k2.6, llama-4-scout, qwen3.7-plus, command-a, grok-4.5, mistral-large-3)
  • 1.5 - 2.0: Heavy / flagship reasoning models (gemini-3-pro, gpt-5.5-pro, claude-opus-4-8)
• active: boolean: Upstream availability toggle.
• group: string: Filtered to "chat" for conversational text completions.

---

Quota & Limits Dashboard

OmniRoute automatically connects ChatPlayground to its Limits & Quotas dashboard:

• Endpoint: GET https://app.chatplayground.ai/api/user (authenticated via minted Clerk Bearer JWT).
• Unlimited Plan (Lifetime / Monthly $25): Tracked at 300 credits daily under fair-use. Displays daily used queries (user.dailyQueriesCount), remaining allowance, percentage, and daily midnight UTC resets.
• Pro Plan: Displays 2,000 queries/month (StackSocial Lifetime) or 1,500 advanced queries/month (Monthly Pro), tracking used queries against the monthly limit and billing period end (stripeCurrentPeriodEnd).
• Basic Plan: Displays 500 queries/month (StackSocial Lifetime), tracking used queries and resets.
• Free Accounts: Displays base usage metrics and upgrade notices.

---

Message Character Limit (15,000 Characters)

ChatPlayground enforces a strict server-side boundary of 15,000 characters per message:

• If an individual message's content exceeds 15,000 characters (15,001+ characters), ChatPlayground's upstream API rejects the turn with HTTP 400: "Message cannot exceed 15,000 characters".
• This limit is evaluated per message turn in the conversation array, not as a total request payload size.
• Pre-Validation: OmniRoute's ChatPlaygroundExecutor validates message character lengths before dispatch and surfaces an actionable error if any turn exceeds 15,000 characters.
• Mitigation Strategies:
  • Enable OmniRoute's prompt compression (e.g. RTK or lite mode) to condense large contexts, system instructions, and tool outputs.
  • Chunk or split large codebases, files, or multi-turn histories across separate prompts.

---

Streaming & Protocol Details

• ChatPlayground streams chunks appending or containing sentinel tokens in the form of CHAT_ID:[a-zA-Z0-9_-]+.
• OmniRoute's ChatPlaygroundExecutor strips these sentinel tokens on the fly and translates the output into OpenAI-standard Server-Sent Events (data: {"choices":[{"delta":{"content":"..."}}]}\n\n) terminating with data: [DONE]\n\n.
• For non-streaming requests, a standard chat.completion response envelope is returned.
• Upstream returns a complete response for Perplexity queries rather than incremental SSE chunks; when stream: true is requested on Perplexity queries, OmniRoute synthesizes standard SSE chunks so streaming OpenAI clients remain fully compatible.

---

Tool Functions & Thinking / Reasoning Content

1. Tool Functions (Not Supported Upstream)

• Ignored by Upstream: ChatPlayground's web backend does not support native function calling or tool execution. When an OpenAI-format tools parameter is passed, the upstream API returns HTTP 200, but models ignore the tool definitions and produce standard conversational text without generating structured tool_calls.
• Emulated Support: Classified with toolCalling: "emulated" in src/shared/constants/providers/web-cookie.ts. If tool calling is required by an agentic client, OmniRoute handles it via prompt-based tool emulation.

2. Thinking & Reasoning Content

• Inline <think> Blocks: Reasoning models (such as DeepSeek R1 / DeepSeek V4, Kimi K3, and Sonar Reasoning Pro) include their internal chain-of-thought directly in the generated output, typically enclosed within <think>...</think> blocks.
• Pass-Through: These reasoning messages are preserved and streamed verbatim as part of the assistant response text.

---

Terms of Service, Compliance & Operator Risk

ChatPlayground (operated by Yaseen AI) includes specific restrictions in its Terms of Service:

• Section 11 (Excessive / Exploitative Usage): Strictly prohibits "sharing of accounts among multiple individuals or bots" and "running scripts to exploit the AI on Yaseen AI". Accounts violating these provisions may be suspended immediately without refund.
• Section 12 (Right to Terminate Usage Temporarily): Daily usage that exceeds $10 in internal costs triggers temporary suspension for that day, and usage exceeding $80 in monthly internal costs triggers termination for the remainder of the month.

OmniRoute Posture & Safety Recommendations

• Categorization: ChatPlayground is explicitly classified as a web-session provider with subscriptionRisk: true and riskNoticeVariant: "webCookie". The OmniRoute dashboard presents a risk notice modal upon setup.
• Single-User Workflows: Do not share a single ChatPlayground session cookie across a public proxy or multi-user team. Use it for personal assistant/coding sessions.
• Stay Under Daily Budget: Do not flood the provider with high-volume benchmarks or batch automation. The 300 queries/day fair-use cap helps stay safely below the internal $10/day cost threshold.
• Browser-Identical Headers: OmniRoute automatically attaches standard desktop browser User-Agents, origins, and referrers (https://web.chatplayground.ai/) and removes internal CHAT_ID tokens to avoid unnecessary anomalies.