Centralized Access to
Mani AI Core Infrastructure
Generate cryptographically secure API keys, manage daily quotas, track request latencies, and route developer queries directly to Project Mani's intelligence router.
RSMK Technologies: Engineering the Foundations
At RSMK Technologies, we reject superficial AI wrappers. We construct the core infrastructure, database systems, and low-latency API layers necessary to drive state-of-the-art cognitive products.
Our engineering values prioritize data safety, secure authentication, and architectural speed. From optimizing local SQLite engines via modern WASM client drivers, to routing high-throughput queries, RSMK builds for absolute developer trust and scale.
Engineering Spec
Gateway Processing Pipeline
Authentication Hashing
Verify incoming headers, hash the bearer API key using cryptographically secure SHA-256, and validate identity.
Rate-Limit Validation
Check daily execution budgets and concurrent usage spikes against key-specific quotas to prevent service abuse.
Context Assembly
Re-construct the user context, history payloads, and dynamic system state for the model router.
Inference & Logging
Generate completions from the underlying LLM engine and asynchronously log response latencies to the audit database.
Project Mani: The Semantic Intelligence Layer
Project Mani is not a static assistant. It serves as a unified routing node, processing user commands, combining them with rich application states (like `siteContext`), and maintaining a coherent conversational thread (`history`) to output high-fidelity responses.
When developers invoke the gateway, the system maps their OpenAI-compatible messages structure directly to the Mani Core endpoint payload. It securely handles the transaction, evaluates performance metrics, and logs the query for auditable analytics.
Designed for Secure Operations
We provide the tools developers need to scale Mani AI securely across their systems.
Cryptographic Keys
Tokens are verified by matching secure SHA-256 hashes in our database, defending your credentials against exposure or database leaks.
Adaptive Limits
Assign daily limits (ranging from 100 to 50,000 requests/day) to individual keys to protect downstream resources.
Full Audits & Logs
Every request logs user-agents, network IPs, status codes, and exact gateway latency in milliseconds.
Standard Integration Snippet
Mani AI Gateway is a drop-in replacement for standard model completion clients. Authorize your requests using the standard Bearer header convention:
Accepts and returns standard choices, message roles, and token counts.
The gateway translates the query, merges contexts, and queries llama-3.3-70b-versatile dynamically.
const response = await fetch('https://api.rsmk.dev/v1/chat', {
method: 'POST',
headers: {
'Authorization': 'Bearer mani_live_f3a7c...',
'Content-Type': 'application/json'
},
body: JSON.stringify({
model: 'mani-reasoning-v1',
messages: [
{ role: 'user', content: 'What is BudgetBuddy?' }
]
})
});
const result = await response.json();
console.log(result.choices[0].message.content);Frequently Asked Questions
Quick answers regarding credentialing, system design, and platform availability.
What is Project Mani?
Project Mani is the core intelligence layer built by RSMK Technologies. It functions as a centralized gateway routing semantic queries to advanced models like llama-3.3-70b-versatile, incorporating user context and chat history dynamically.
Who is RSMK Technologies?
RSMK Technologies is a systems engineering enterprise focusing on foundational AI systems, low-latency API routers, secure cryptographic identity, and high-fidelity developer tools.
How secure is the API Key management system?
Highly secure. We treat API keys like passwords. They are hashed using SHA-256 prior to database storage. The raw key token (e.g. mani_live_...) is shown to the user exactly once during creation and can never be retrieved again.
What endpoints are currently active?
The production chat router is active at /api/v1/chat. It accepts OpenAI-compatible payloads (model, messages, temperature) and proxies them to the private core model routing cluster after key and rate-limit verification.