[fal.ai] frame_processor: Pipeline setup fails with GraphConfig validation error — edges use 'source'/'target' but schema expects 'from'/'to_node'

[livepeer-tessa]

Summary

scope.server.frame_processor logs ERROR Pipeline setup failed: 8 validation errors for GraphConfig when a client sends edge definitions using old-style source/target keys instead of the current schema fields from, from_port, to_node, to_port.

Observed 4 times in the 12h window (2026-04-09 18:00 – 2026-04-10 06:00 UTC).

Error Details

From Grafana/Loki — 2026-04-09 ~23:09 UTC

scope.server.frame_processor - ERROR - Pipeline setup failed: 8 validation errors for GraphConfig

edges.0.from
  Field required [type=missing, input_value={'source': 'input', 'target': 'pipeline'}, input_type=dict]
edges.0.from_port
  Field required [type=missing, input_value={'source': 'input', 'target': 'pipeline'}, input_type=dict]
edges.0.to_node
  Field required [type=missing, input_value={'source': 'input', 'target': 'pipeline'}, input_type=dict]
edges.0.to_port
  Field required [type=missing, input_value={'source': 'input', 'target': 'pipeline'}, input_type=dict]
edges.1.from
  Field required [type=missing, input_value={'source': 'pipeline', 'target': 'output'}, input_type=dict]
edges.1.from_port
  Field required [type=missing, input_value={'source': 'pipeline', 'target': 'output'}, input_type=dict]
edges.1.to_node
  Field required [type=missing, input_value={'source': 'pipeline', 'target': 'output'}, input_type=dict]
edges.1.to_port
  Field required [type=missing, input_value={'source': 'pipeline', 'target': 'output'}, input_type=dict]

Timestamps: 2026-04-09 23:09:57, 23:12:44, 23:39:26, 23:45:02 UTC
fal_job_id: d35991f0-7869-44ca-88b4-1459ad9c7774
fal_app_name: scope-app--prod

Context

The client is sending edges as [{'source': 'input', 'target': 'pipeline'}, ...] — an older schema where edges were identified by source/target node names without port information.

The server-side GraphConfig pydantic model now requires from, from_port, to_node, to_port on each edge.

Despite the validation failure, the stream appears to start (pipeline loaded successfully just before), but frame_processor cannot set up processing — video output shows 0 frames published for the session duration.

Impact

• Session starts but no frames are processed/output
• 4 occurrences in 12h window — suggests multiple users on an older client version
• Silent failure: user sees stream but no AI output

Possible Causes

1. Client/server version skew — an older Scope desktop client sending the old edge schema to an updated cloud backend
2. Missing schema migration — GraphConfig model doesn't include a backwards-compat source/target alias

Suggested Fix

Either:

• Add pydantic alias/validator to accept both old (source/target) and new (from/to_node) edge schema formats
• Or surface a clearer error to the client asking them to update