Skip to content

PLATFORM_RUNTIME_MAP.md

Latest commit

 

History

History
69 lines (45 loc) · 2.51 KB

PLATFORM_RUNTIME_MAP.md

File metadata and controls

69 lines (45 loc) · 2.51 KB

Platform Runtime Map (API Apps and Temporal Apps)

This page is the canonical map of where features belong in Fred.

Use it before adding endpoints, workflows, or policies.

1) API Applications

Fred has three API applications:

  1. Agentic Backend API (agentic-backend)

    • Main role: chat/session runtime and agent orchestration.
    • Typical concerns: sessions, agent execution, chat interactions.

  2. Knowledge Flow Backend API (knowledge-flow-backend)

    • Main role: ingestion, documents, tags/libraries, retrieval-facing operations.
    • Typical concerns: content lifecycle, metadata, vectors, document pipelines.

  3. Control Plane API (control-plane-backend)

    • Main role: teams/users operations and policy-driven lifecycle control.
    • Typical concerns: team membership changes, policy evaluation, purge orchestration.

2) Temporal Applications (Workers)

Fred also has Temporal workers separated by concern:

  1. Knowledge Flow Temporal Worker

    • Runs ingestion/processing workflows.
    • Focus: batch conversion, extraction, indexing pipelines.

  2. Agentic Temporal Worker

    • Runs long-running/scheduled agentic workloads (to be progressively consolidated there).
    • Focus: durable agent executions outside synchronous API request lifecycle.

  3. Control Plane Temporal Worker

    • Runs lifecycle/policy jobs.
    • Focus: policy-based purge/archive workflows (for example member-removal cleanup).

3) Placement Rules

When adding new behavior, decide with these rules:

  1. User/team/admin API? Put it in Control Plane API.
  2. Document ingestion/indexing pipeline? Put it in Knowledge Flow (API + Temporal if async/batch).
  3. Chat/session runtime behavior? Put it in Agentic API.
  4. Policy-driven scheduled lifecycle action? Put it in Control Plane Temporal.
  5. Cross-backend shared primitive? Put it in fred-core (only if truly shared, stable, and minimal).

4) Startup Model (Same Pattern Across Apps)

All Python backends follow the same startup convention:

  • ENV_FILE for secrets/env variables.
  • CONFIG_FILE for YAML configuration.

Standard commands:

  • make run for API process.
  • make run-worker for Temporal worker process.

See:

5) Related Docs