Graphify-Labs/graphify · error · ValueError

Extraction JSON has {len(errors)} error(s):

Error message

Extraction JSON has {len(errors)} error(s):

What it means

Raised by assert_valid (graphify/validate.py) after validate_extraction collected one or more schema violations in an extraction JSON dict: missing 'nodes'/'edges' keys, non-list values, nodes/edges that are not objects, missing required fields (id, label, file_type, source_file on nodes; source, target, relation, confidence, source_file on edges), invalid file_type/confidence enums, non-hashable ids, or edge endpoints that reference no known node id. The single ValueError aggregates every problem as a bulleted list, so one raise reports all defects at once.

Source

Thrown at graphify/validate.py:95

                try:
                    unmatched = bool(node_ids) and val not in node_ids
                except TypeError:
                    errors.append(
                        f"Edge {i} {endpoint} {val!r} is non-hashable - must be a string"
                    )
                    continue
                if unmatched:
                    errors.append(f"Edge {i} {endpoint} '{val}' does not match any node id")

    return errors


def assert_valid(data: dict) -> None:
    """Raise ValueError with all errors if extraction is invalid."""
    errors = validate_extraction(data)
    if errors:
        msg = f"Extraction JSON has {len(errors)} error(s):\n" + "\n".join(f"  • {e}" for e in errors)
        raise ValueError(msg)

View on GitHub (pinned to 7fe58b0b0f)

Solutions

1. Read the bulleted lines in the message — each names the index and the exact defect; fix them in the JSON before re-validating.
2. For unmatched edge endpoints, either add the missing nodes or drop/repair the dangling edges (most common failure).
3. Constrain your LLM prompt to the enums: file_type in {code, document, paper, image, rationale, concept}, confidence in {EXTRACTED, INFERRED, AMBIGUOUS}.
4. Run validate_extraction(data) (returns a list, no raise) in a pre-commit or pipeline step so drift is caught before assert_valid.

Example fix

# before (edge references a node that doesn't exist)
data = {"nodes": [{"id": "n1", "label": "A", "file_type": "code", "source_file": "a.py"}],
         "edges": [{"source": "n1", "target": "nX", "relation": "calls", "confidence": "EXTRACTED", "source_file": "a.py"}]}
assert_valid(data)  # ValueError: 1 error(s)

# after
data["edges"][0]["target"] = "n1"
assert_valid(data)  # passes

from graphify.validate import validate_extraction

errors = validate_extraction(data)  # returns list[str], never raises
if errors:
    data = repair_extraction(data, errors)  # add missing nodes, coerce enums, etc.
assert not validate_extraction(data)

def is_valid_extraction(data: object) -> bool:
    return isinstance(data, dict) and validate_extraction(data) == []

try:
    assert_valid(data)
except ValueError as e:
    # message aggregates every defect as '  • ...' lines — parse and report all
    defects = [ln.strip("• ") for ln in str(e).splitlines() if ln.strip().startswith("•")]
    report_to_user(defects)  # fix-and-retry loop instead of failing on the first

When it happens

Trigger: Calling assert_valid(data) on LLM-produced or hand-written extraction JSON that violates the schema — typical cases: edge source/target ids that don't match any node id, a node missing 'label', file_type 'snippet' outside {code, document, paper, image, rationale, concept}, or confidence 'high' outside {EXTRACTED, INFERRED, AMBIGUOUS}. Also 'edges' spelled as neither edges nor links (links is accepted as a NetworkX <=3.1 fallback).

Common situations: Prompting an LLM to emit graphify extraction JSON and getting drifted field names or hallucinated edge ids; merging partial extractions where a node was dropped but its edges remained; schema drift between graphify versions; hand-editing graph.json and forgetting required fields.

Related errors

• Graph path must be a .json file, got: {graph_path!r}
• could not read Google Workspace shortcut {path}: {exc}
• Google Workspace shortcut {path} does not include a Drive fi
• ingest: {exc}
• outcome must be one of {OUTCOMES}, got {outcome!r}

AI-assisted analysis of Graphify-Labs/graphify@7fe58b0b0f (2026-08-14). Data as JSON: /api/errors/be899e777aa1a848. Report an issue: GitHub.