// legal & compliance

Privacy Policy

Effective Date: September 21, 2026 Version 2.4 Domain: kallme-relay.lol

01 Scope & Core Commitment

This Privacy Policy governs your use of kallme-relay (kallme-relay.lol), including all API gateway endpoints (/relay/*, /huijv/*) and the browser playground interface.

Our fundamental architectural principle is Zero Data Retention (ZDR). We treat every incoming inference request as ephemeral transit. We do not inspect prompt contents for monetization, we do not train or fine-tune machine learning models on your inputs or outputs, we do not construct behavioral consumer profiles, and we never sell, lease, or broker your personal information.

02 Information We Process

Because kallme-relay acts strictly as a stateless protocol translator and quota multiplexer, our data processing footprint is categorized as follows:

A. User & Inference Content (In-Memory Transit Only)
  • Prompt Inputs: Text prompts, system instructions, role arrays, and multivariable messages sent to /v1/chat/completions or /v1/messages.
  • Generated Responses: Completion tokens, tool calls, and <thought_process> reasoning scratchpads.
  • Lifecycle: These payloads reside exclusively in volatile server memory (RAM) for the duration of the HTTP/SSE streaming connection. Upon connection termination or SSE stream close, memory references are freed. No payload data is written to disk databases, query logs, or long-term caching stores.
B. Client-Side Browser Storage (Zero Server Retention)

The web playground stores state entirely client-side inside your browser’s localStorage. The server neither stores nor has automated read access to:

  • Past session histories and active chat message lists (kallme_sessions).
  • UI customizations, system prompt presets, thinking budgets, and temperature sliders (kallme_settings).
  • Last selected model preference (kallme_selected_model).
C. Operational & Transport Telemetry
  • Socket Metadata: Originating IP address, timestamp, HTTP request method, requested path, response status code, and bytes transferred.
  • Purpose: Used by Nginx strictly for real-time TCP connection routing, DDoS mitigation, and transient rate-limit enforcement.
  • Retention: Rotated and overwritten on short system log lifecycles. Not linked to user identities or query contents.

03 Data Processing Pipeline

Stage Action Storage Mechanism
Ingress TLS 1.3 handshake & route parsing Ephemeral kernel network buffer
Translation Convert Anthropic Messages $\leftrightarrow$ OpenAI Chat Stateless Python memory buffer
Quota Routing Bank key selection & pre-auth token capping In-memory key rotation index
Egress Stream SSE chunk streaming to client Volatile socket stream (Zero disk write)

04 Upstream Model Providers

To fulfill inference requests, kallme-relay dispatches sanitized payloads to the selected model infrastructure. Depending on the model identifier requested, data is proxied to:

  • Anthropic PBC: Claude 3.5, 3.7, 4.x family endpoints.
  • OpenAI LLC: GPT-4o, o1, o3, GPT-5, and GPT-6 frontier endpoints.
  • xAI: Grok series endpoints.
  • Google LLC: Gemini and Gemma model endpoints.
  • Huijv Infrastructure: Specialized research checkpoints (Astra, Luna, Sol).

Upstream processing is subject to each provider’s respective API terms and data protection policies. We route through enterprise API tiers where providers agree not to train foundational models on API payloads.

05 Cookies & Tracking Technologies

kallme-relay does not employ tracking cookies, marketing pixels, canvas fingerprinting, or third-party behavioral analytics tools (e.g., Google Analytics, Meta Pixel).

Session continuity is achieved solely via standard HTTP authorization headers and browser local storage.

06 Security Measures

We implement industry-standard administrative, physical, and technical safeguards:

  • Transport Layer Encryption: All traffic in transit is secured using modern TLS (TLS 1.2/1.3) with strong cipher suites and HSTS enforcement.
  • Memory Sandboxing: Worker processes execute in isolated runtime environments with limited process privileges and automatic process recycling.
  • Least-Privilege Routing: Upstream API credentials are encrypted at rest and isolated from public surface exposure.

07 Regulatory Rights (GDPR, UK GDPR, CCPA/CPRA)

Depending on your jurisdiction, you are entitled to specific legal rights regarding your personal data:

  • Right to Erasure / Deletion (GDPR Art. 17, CCPA § 1798.105): You can wipe all client-side data immediately via the data deletion tool. Because we do not retain prompt records on servers, there are no persistent server-side transcripts to purge.
  • Right of Access (GDPR Art. 15): You may inspect your full stored history directly in your browser’s Developer Tools under Local Storage.
  • Right to Opt-Out of Sale / Sharing: We do not sell or share personal data; therefore, no opt-out request is necessary.
  • Non-Discrimination: We will never discriminate against you for exercising your privacy rights.

08 Children’s Privacy

kallme-relay is an engineering and developer platform. It is not intended for or directed to individuals under the age of 18. We do not knowingly collect or solicit information from minors. If you believe a minor has accessed the gateway, contact us immediately.

09 Amendments & Policy Updates

We may update this Privacy Policy periodically to reflect infrastructure modifications, new model catalog additions, or statutory regulatory revisions. Any updates will be published directly to this page with an updated Effective Date. Continued usage of the relay following changes constitutes acknowledgment of the revised terms.

10 Contact & Data Protection Officer

For technical questions regarding our zero-retention architecture, legal compliance inquiries, or formal data protection requests, contact:

chantelwertz6450+dfemz8vk9j@outlook.com