Accurate pipeline: every ingest_data task crashes - ObservationsAgent.observe() called without required 'context' argument

[Starosdev]

Summary

With PIPELINE_MODE=accurate, every ingest_data Celery task raises before observations are saved and the task is marked FAILURE. The ingestion task calls observe() with only text and observate_for, but the method signature requires a third positional argument context.

Environment

• Image: ghcr.io/lumen-labs/brainapi:v2.14.2-dev
• Also present on main (af5c778)
• Backends: GRAPH_DB=networkx, DATA_DB=postgresql, VECTOR_DB=postgresql (backend-independent - the crash is in the task body before any graph work)
• MODELS_MODE=remote (OpenAI), AGENTIC_ARCHITECTURE=custom

Reproduction

1. Deploy with PIPELINE_MODE=accurate.
2. POST /ingest/ with any text payload:

   {"data": {"data_type": "text", "text_data": "Some document text."}}

3. Watch the worker log.

Actual result

[ERROR/MainProcess] Task src.workers.tasks.ingestion.ingest_data[...] raised unexpected:
TypeError("ObservationsAgent.observe() missing 1 required positional argument: 'context'")
Traceback (most recent call last):
  File "/app/src/workers/tasks/ingestion.py", line 346, in ingest_data
    observations = observations_agent.observe(
TypeError: ObservationsAgent.observe() missing 1 required positional argument: 'context'

Task meta ends FAILURE; text chunks and data vectors written before the call remain, so the brain is left partially ingested.

Cause

Call site passes two args:

brainapi2/src/workers/tasks/ingestion.py

Lines 346 to 353 in af5c778

	observations = observations_agent.observe(
	text=(
	payload.data.text_data
	if payload.data.data_type == IngestionTaskDataType.TEXT.value
	else json.dumps(payload.data.json_data)
	),
	observate_for=payload.observate_for,
	)

Signature requires three:

brainapi2/src/core/agents/observations_agent.py

Lines 26 to 28 in af5c778

	def observe(
	self, text: str, observate_for: Optional[List[str]], context: Optional[str]
	) -> list[str]:

def observe(
    self, text: str, observate_for: Optional[List[str]], context: Optional[str]
) -> list[str]:

Expected

Either the call site passes context (or context=None), or the signature defaults it: context: Optional[str] = None.

Workaround

PIPELINE_MODE=lightweight (skips the observations stage entirely).