1. Introduction - The AstraDesk Vision¶
AstraDesk Enterprise AI Framework v1.0 - Technical Guide (EN)
AstraDesk is a secure, modular framework for building enterprise AI agents that reason, plan, and act through tools. It operationalizes AgentOps (observability + evaluation), governance, and DevSecOps specifically for agentic systems.
Why AstraDesk now¶
-
From deterministic code → probabilistic reasoning loops.
-
From static workflows → adaptive, goal-driven agents.
-
From code-first → evaluation-first (ship with evidence, not intuition).
What’s inside v1.0¶
-
Single-agent runtime (SupportAgent / OpsAgent) with human-in-the-loop.
-
MCP-first integrations (tools/resources/prompts) via AstraDesk Gateway.
-
AstraOps for traces, metrics, evals; AstraCatalog for ownership, risk, versions.
-
Hybrid deployments: AWS/Kubernetes/OpenShift/on-prem; PostgreSQL 18; OpenTelemetry.
Document map¶
- Architecture Overview → big picture
- (3.-7.) ADLC → plan, build, test/optimize, deploy, operate
- Security & Governance
- MCP Gateway & Domain Packs
- Future Roadmap and Glossary
High-level architecture sketch (for context)¶
```mermaid
flowchart LR subgraph GW["AstraDesk Gateway"] G1[MCP Ingress] G2[LLM Gateway] end subgraph RT["Agents Runtime"] A1[SupportAgent] A2[OpsAgent] end subgraph OPS["AstraOps"] O1[Traces] O2[Evals] end subgraph CAT["AstraCatalog"] C1[Registry] C2[Policies] end subgraph DATA["Data/Tools"] D1[(PostgreSQL 18)] D2[(Vector/Graph DB)] D3[(External APIs)] end G1 --> A1 G1 --> A2 A1 <-->|signals| OPS A2 <-->|signals| OPS A1 --> DATA A2 --> DATA OPS --> CAT CAT --> G1
````
Next: 2. Architecture Overview