Configuration

All configuration options for lexigram-multimedia-interpolate.

---

Overview

InterpolationConfig extends BaseConfig (stdlib dataclasses, not pydantic) and declares its section via config_section = "multimedia_interpolate". Loading follows the canonical BaseConfig.from_yaml() order — each later source overrides the previous:

1. application.yaml values (base layer)
2. Profile overlay application.{LEX_PROFILE}.yaml (if set)
3. LEX_* environment variables

InterpolationGenerationProvider passes the section into the container (config_key = "multimedia_interpolate"). Under the lexigram-multimedia umbrella, the same keys nest as multimedia: interpolate:.

The config surface is deliberately tiny: frame-pair interpolation needs only a backend, a server URL, and a timeout. There is no factor/fps knob here — those are per-call arguments on VideoInterpolationService.interpolate_video().

Basic Example

application.yaml

multimedia_interpolate:
  backend: "rife"
  rife_base_url: "http://localhost:5500"
  timeout: 30.0

config = InterpolationConfig.from_yaml("application.yaml")  # section used automatically
assert config.backend == "rife"

Programmatic equivalent:

from lexigram.multimedia.interpolate import InterpolationModule
from lexigram.multimedia.interpolate.config import InterpolationConfig

module = InterpolationModule.configure(config=InterpolationConfig(timeout=30.0))

Options

All fields of InterpolationConfig:

Option	Type	Default	Description
backend	Literal["rife"]	"rife"	Interpolation backend to register (only rife is implemented)
rife_base_url	str	"http://localhost:5500"	Base URL of the RIFE reference server
timeout	float	15.0	HTTP request timeout in seconds for /interpolate

Environment Variables

Prefix: LEX_MULTIMEDIA__INTERPOLATE__ (umbrella-layout YAML) or LEX_MULTIMEDIA_INTERPOLATE__ (flat multimedia_interpolate section). Field names map directly:

Variable	Description
LEX_MULTIMEDIA__INTERPOLATE__BACKEND	Backend selection (rife)
LEX_MULTIMEDIA__INTERPOLATE__RIFE_BASE_URL	RIFE server URL
LEX_MULTIMEDIA__INTERPOLATE__TIMEOUT	Request timeout in seconds

Example:

export LEX_MULTIMEDIA__INTERPOLATE__RIFE_BASE_URL=http://rife.internal:5500
export LEX_MULTIMEDIA__INTERPOLATE__TIMEOUT=45

Advanced Configuration

Resilience Wrapping

RifeInterpolationProvider accepts optional RetryPolicyProtocol / CircuitBreakerProtocol instances. Under DI they resolve automatically when registered in the container, wrapping every /interpolate call. Direct construction works too:

from lexigram.multimedia.interpolate.providers import RifeInterpolationProvider

backend = RifeInterpolationProvider(
    base_url="http://localhost:5500",
    timeout=15.0,
    retry=my_retry_policy,
    circuit_breaker=my_circuit_breaker,
)

Composing with a VideoProcessor

The VideoInterpolationService registration is automatic: if the container has a VideoProcessor (from lexigram-multimedia-video), the provider composes VideoInterpolationService(interpolation_provider, video_processor) and registers it. No config keys control this — presence of the protocol in the container is the only switch:

service = await app.container.resolve(VideoInterpolationService)
result = await service.interpolate_video(asset, factor=2, fps=24.0)

Profile-Based Overlays

export LEX_PROFILE=production

with application.production.yaml overriding just the server URL — env vars still win over the profile file.

Best Practices

• Keep config minimal — only rife_base_url and timeout matter in practice.
• Prefer environment variables for per-deployment server URLs (rife runs on a different host in prod).
• Raise timeout for slow GPUs; interpolation of large frames is compute-bound, not network-bound.
• Use InterpolationModule.stub() in tests (pins the real rife backend without a live server).