datawhalechina/hello-agents · error · Exception

Unknown relation type '{relation}'.

Error message

Unknown relation type '{relation}'.

What it means

find_fact(relation) validates its argument against self.predicate_to_fact_instances, a dict grouping currently-known facts by predicate name (e.g. 'Parallel', 'Perpendicular', 'Eq'). Passing a relation string that is not a key raises this Exception. Note the logic bug: the emptiness check 'if len(self.predicate_to_fact_instances) == 0' comes after the membership check, so an unknown relation raises even when the fact table is empty instead of returning the friendly 'list is empty' message.

Source

Thrown at Co-creation-projects/BitSecret-GPSAgent/src/gps/symbolic_solver.py:1409

        all_goal_ids = set()
        for father_id in father_ids:  # add sub_goals
            operation_id = self._add_operation(('Decompose', theorem_name, theorem_paras))
            goal_ids = self._add_goals(sub_goals, father_id, operation_id)
            if goal_ids is not None:
                all_goal_ids.update(goal_ids)

        if len(all_goal_ids) == 0:
            return f"使用定理'{theorem}'分解目标{goal}失败，新分解的子目标不能是原目标的父目标。"

        self._check_goals(all_goal_ids)

        result = f"使用定理'{theorem}'分解目标{goal}成功，以下为问题的状态更新。\n"
        return result + self._get_update(old_fact_id, old_goal_id, old_goal_status)

    def find_fact(self, relation):
        if relation not in self.predicate_to_fact_instances:
            msg = f"Unknown relation type '{relation}'."
            raise Exception(msg)

        if len(self.predicate_to_fact_instances) == 0:
            return relation + f"类型的关系列表为空，当前问题暂时未推导出{relation}关系。"

        if relation == 'Eq':
            result = []
            if len(self.equations) > 0:
                result.append("按照方程变量是否相交来分组，得到的代数方程组（所有方程省略'=0'、组序号可能不连续）：")
                for group_id in self.equations:
                    eqs = [str(eq).replace(' ', '') for eq in self.equations[group_id][0]]
                    result.append(f'Group {group_id}: ' + ', '.join(eqs))
            if len(self.sym_to_value) > 0:
                result.append("以下是所有已经求解出值的变量：")
                result.append(str(self.sym_to_value))
            return '\n'.join(result)
        else:
            instances = []
            for instance in self.predicate_to_fact_instances[relation]:

View on GitHub (pinned to 606a07d341)

Solutions

1. Enumerate valid relations first: valid = sorted(solver.predicate_to_fact_instances.keys()) and pass one of those exactly.
2. Match the casing used by the GDL predicates (typically capitalized like 'Parallel').
3. If the dict is empty, add/derive facts before querying — the 'empty' friendly path is unreachable for unknown names due to the check order.

Example fix

# before
solver.find_fact('parallel')  # wrong casing -> raises

# after
valid = sorted(solver.predicate_to_fact_instances.keys())
rel = next((r for r in valid if r.lower() == 'parallel'), None)
print(solver.find_fact(rel) if rel else f'no such relation, have {valid}')

def safe_find_fact(solver, relation):
    keys = solver.predicate_to_fact_instances.keys()
    match = next((k for k in keys if k.lower() == relation.lower()), None)
    if match is None:
        return f'unknown relation; valid: {sorted(keys)}'
    return solver.find_fact(match)

try:
    print(solver.find_fact(relation))
except Exception as e:
    if str(e).startswith("Unknown relation type"):
        print('valid relations:', sorted(solver.predicate_to_fact_instances))

When it happens

Trigger: Calling solver.find_fact('paralle') (typo), find_fact('parallel') with wrong casing, or a predicate that simply has no facts recorded yet in a fresh problem state — the membership check fires before any content check.

Common situations: Agents guessing predicate names not present in the current problem's fact table; casing/naming mismatches between GDL predicates and the relation string; querying before any facts were added or derived.

Related errors

• Unknown theorem name: '{theorem_name}'.
• Theorem parameters must be uppercase letters and , only. The
• '{theorem}' has wrong number of parameters (expected {len(se
• When using the 'apply' tool with theorem '{theorem_name}', t
• When the theorem name contains 'perimeter', 'area', 'similar

AI-assisted analysis of datawhalechina/hello-agents@606a07d341 (2026-08-14). Data as JSON: /api/errors/0ce9fbf24ee7423c. Report an issue: GitHub.