Foundational Concepts

Before integrating Naagmani into your codebase, familiarize yourself with these core concepts:

graph TD
    Org["Organization (Acme Corp)"] --> P1["Project A (E-Commerce)"]
    Org --> P2["Project B (Internal HR)"]
    P1 --> E1["Development Environment"]
    P1 --> E2["Production Environment"]
    E2 --> ST["Project Service Tokens"]
    E2 --> AK["API Keys"]
    E2 --> Agents["Autonomous Agents"]

Core Vocabulary #

1. Organization #

The top-level root account representing your company or team. Governs billing, enterprise membership, global guardrails, and compliance logs.

2. Project #

A logical workspace dedicated to a specific application or product line. Projects isolate credentials, routing policies, tools, and agents from other teams.

3. Environment #

An execution boundary inside a project (Test and Production). Environments allow teams to safely test experimental model weights without affecting production workloads.

4. Project Service Token (nst_...) #

A specialized machine-to-machine credential bound to a project and environment. Supports fine-grained capability matrices (e.g., inference:chat, tools:execute), TTL expiration, and human member attribution.

5. Smart Model Routing #

An automated policy engine that dynamically routes inference requests across foundational model providers based on cost optimization, latency thresholds, or fallback priority chains.

6. Provider Attempt #

A discrete unit of upstream execution telemetry. If a request experiences 2 failovers before succeeding, Naagmani records 3 discrete attempts, detailing latency, tokens, cost, and upstream error reasons for each hop.

7. Model Context Protocol (MCP) #

An open protocol enabling AI models to safely access tools, filesystems, and databases. Naagmani acts as a secure MCP gateway, enforcing permissions and audit logging on every tool invocation.

Progressive Learning Path #