Getting Started¶

High-level orientation for Astra. Implementation setup, repo layout, and contributor workflows are intentionally not documented on this public wiki — they live with the source tree for people who already have access.

What Astra is¶

Applications use the Astra SDK, which talks to a small kernel: actor runtime, task graph, scheduler, messaging, and durable state coordination. Under that sits Postgres (source of truth), Redis (streams and fast ephemeral state), Memcached (LLM-related caching), and object storage. Agent and product logic stay outside the kernel via documented APIs — see the architecture overview and PRD §3–5 in the Astra repo.

If you contribute¶

1. Read PRD §1–9 for vision, kernel boundaries, and services.
2. Read the Glossary and Architecture overview.
3. Follow the contributor onboarding and tooling in the private Astra repository (local run, tests, migrations).

Key surfaces¶

• Goal-based agents: Submit goals via REST; the planner builds a task DAG, the scheduler dispatches, workers execute in sandboxes, and results persist to Postgres.
• Agent profiles (Phase 9): Agents can have system prompts and attached documents (rules, skills, context docs, references) that propagate through planning to execution. See Services.
• Chat agents (Phase 10): Users interact with chat-capable agents via WebSocket streaming — real-time responses with optional tool invocation and memory context.
• Slack integration (Phase 12, partial): Connect agents to Slack workspaces for chat via DMs and channels. Proactive posting supported.
• 16 canonical microservices plus optional services (slack-adapter, webhook-ingest, cost-tracker). See the full service catalogue.

Public API smoke test¶

If you have a running environment, health and basic flows are described at a high level in PRD §15 and the API endpoints summary. Exact request bodies and port defaults may change — verify against your deployment.

Further reading¶