Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.qredence.ai/llms.txt

Use this file to discover all available pages before exploring further.

fleet-rlm is a Daytona-backed recursive runtime wrapped by a thin transport shell and a narrow hosted-policy layer.

​
Design principles

Three intentional choices drive the shape of the codebase:
  1. The backend is intentionally thin. The Python layer is a transport + orchestration shell over dspy.ReAct, dspy.RLM, and Daytona sandboxes. “Intelligence” lives in DSPy (upstream) and in how recursive child sandboxes are scheduled (this repo). Plumbing belongs in src/fleet_rlm/api/; policy belongs in src/fleet_rlm/runtime/.
  2. The UI is treated as core, not peripheral. The runtime emits streaming events, code-execution results, and artifacts that only make sense in an interactive surface. That is why src/frontend/ is comparable in line count to src/fleet_rlm/.
  3. Two agent layers, both dspy.*, both real.
    • Chat surface: dspy.ReAct at src/fleet_rlm/runtime/agent/agent.py.
    • Recursive engine: dspy.RLM at src/fleet_rlm/runtime/models/builders.py with delegation at src/fleet_rlm/runtime/tools/rlm_delegate.py.

​
Layering

​
Layer 1 — Transport

Thin FastAPI/WebSocket shell. Primary files:
  • src/fleet_rlm/api/main.py
  • src/fleet_rlm/api/routers/ws/endpoint.py
  • src/fleet_rlm/api/runtime_services/chat_runtime.py
  • src/fleet_rlm/api/runtime_services/chat_persistence.py
  • src/fleet_rlm/api/runtime_services/diagnostics.py
  • src/fleet_rlm/api/runtime_services/settings.py
  • src/fleet_rlm/api/runtime_services/volumes.py
Responsibilities:
  • App factory, lifespan, route mounting, SPA asset serving.
  • Auth-derived HTTP and WebSocket identity.
  • Session lookup, runtime preparation, service orchestration.
  • WebSocket lifecycle and execution-event envelope delivery.
  • Runtime settings, diagnostics, and Daytona volume browsing.

​
Layer 2 — Runtime core

Primary files:
  • src/fleet_rlm/api/routers/ws/stream.py
  • src/fleet_rlm/runtime/factory.py
  • src/fleet_rlm/runtime/agent/agent.py
  • src/fleet_rlm/runtime/agent/runtime.py
  • src/fleet_rlm/runtime/execution/*
  • src/fleet_rlm/runtime/models/*
Responsibilities:
  • Shared chat/runtime execution.
  • Recursive delegation and tool execution.
  • Execution-event assembly and workbench hydration.
  • Runtime model assembly and registry management.

​
Layer 3 — Daytona substrate

Primary files:
  • src/fleet_rlm/integrations/daytona/interpreter.py
  • src/fleet_rlm/integrations/daytona/runtime.py
  • src/fleet_rlm/integrations/daytona/filesystem.py
  • src/fleet_rlm/integrations/daytona/diagnostics.py
Responsibilities:
  • Sandbox and interpreter lifecycle.
  • Repo checkout, workspace path staging, durable mounted volumes.
  • Provider-specific diagnostics and volume normalization.

​
Layer 4 — Offline quality

Primary files: src/fleet_rlm/runtime/quality/* Responsibilities:
  • DSPy evaluation.
  • GEPA optimization.
  • Offline scoring, datasets, and module registry management.

​
Stateful restore

Session manifests on durable storage are the authoritative local restart-restore source. The manifest state payload restores:
  • dspy.History conversation turns.
  • AgentRuntime core memory (default core memory plus persisted keys).
  • Session-local loaded document paths.
  • Daytona interpreter state — sandbox ID, workspace path, repo/ref/context paths, volume name, volume subpath.
Importing a session replaces session-local memory and document state instead of merging into the active runtime. Empty or missing state resets history, core memory, loaded documents, and sandbox buffers so switching sessions cannot leak stale agent context.

​
Reading order

When you need the current backend story, read in this order:
  1. src/fleet_rlm/api/main.py
  2. src/fleet_rlm/api/routers/ws/endpoint.py
  3. src/fleet_rlm/api/routers/ws/stream.py
  4. src/fleet_rlm/runtime/factory.py
  5. src/fleet_rlm/runtime/agent/agent.py
  6. src/fleet_rlm/integrations/daytona/interpreter.py
  7. src/fleet_rlm/integrations/daytona/runtime.py
Older transition notes may still mention orchestration_app/ or api/orchestration/. Those labels are historical only and are intentionally absent from the current tree.