Hacktoberfest 2026: los issues que los mantenedores marcaron para octubre, abiertos y aptos para principiantes. Explorar issues de Hacktoberfest

paper_flow() fails on its own README example: dict-typed nodes field breaks strict structured output, UUID id fields reject the model's natural output

Abierto
  • ¿Algo no está bien?
#91 2 comentarios 0 reacciones 0 asignados Ver en GitHub

Nadie ha tomado este issue todavía.

  • #92 de @isorobo — cerrado sin fusionar

Evaluación

Dificultad
4/5
Tiempo estimado
3-5 días
Aptitud para principiantes
58/100
Tipo de issue
Error
Claridad
Bien especificado
Estado de actividad
Tranquilo
Stack tecnológico
python
Área
ai, backend

Línea de trabajo

Empieza por quantmind/flows/paper.py y quantmind/knowledge/_tree.py; después, inspecciona los modelos Paper y los modelos de knowledge relacionados mencionados en el issue. Ejecuta el ejemplo de un solo paper del README con openai-agents 0.18.1 y la suite completa de pruebas. Se considera terminado cuando el ejemplo se completa sin errores de validación de schema o UUID y las pruebas existentes siguen pasando, salvo el fallo no relacionado indicado.

Escrito por el modelo de indexación a partir del texto del issue.

Descripción

area: flows stale type: bug

Summary

The single-paper example in the README does not run against the currently resolved dependency versions (openai-agents>=0.14, no ceiling, resolves to 0.18.1):

import asyncio
from quantmind.configs import PaperFlowCfg
from quantmind.configs.paper import ArxivIdentifier
from quantmind.flows import paper_flow

async def main() -> None:
    paper = await paper_flow(
        ArxivIdentifier(id="2401.12345"),
        cfg=PaperFlowCfg(model="gpt-4o-mini"),
    )

asyncio.run(main())

Two separate bugs stack on top of each other.

Bug 1 — Agent(output_type=Paper) raises before any LLM call

TreeKnowledge.nodes is dict[UUID, TreeNode] (quantmind/knowledge/_tree.py). OpenAI's strict-mode structured output — the default in current openai-agents releases — cannot represent dict-typed fields at all (arbitrary keys are incompatible with a fixed JSON Schema). paper_flow (quantmind/flows/paper.py) passes the bare Paper type straight into Agent(output_type=out_type), with no strict_json_schema=False escape hatch, so the run fails immediately:

agents.exceptions.UserError: Strict JSON schema is enabled, but the output type is not valid.
Either make the output type strict, or wrap your type with AgentOutputSchema(YourType, strict_json_schema=False)

Bug 2 — once strict mode is off, UUID-typed id fields reject the model's output

Wrapping the output type in AgentOutputSchema(Paper, strict_json_schema=False) gets past bug 1, but a non-strict schema drops the UUID format constraint. The extraction agent then does exactly what you'd expect an LLM to do when free-form: it fills TreeNode.node_id / parent_id / children_ids with readable slugs ("root", "introduction", "methodology") and BaseKnowledge.id with the paper's arXiv id ("2404.11584"), rather than leaving these at their UUID4 defaults. Validation then fails with ~80+ uuid_parsing errors.

Repro

Ran the unmodified README example against a real paper (arXiv 2404.11584, "The landscape of emerging AI agent architectures for reasoning, planning, and tool calling: A survey", model gpt-4o-mini) on the current main (8e21888). Fails both ways described above. Full traceback available on request.

Fix

Opened # — changes TreeNode.node_id/parent_id/children_ids, TreeKnowledge.root_node_id/nodes, BaseKnowledge.id, and PaperKnowledgeCard.paper_id from UUID to str (accepts both UUIDs and readable slugs), and has paper_flow pass AgentOutputSchema(out_type, strict_json_schema=False) instead of a bare type. Verified: the unmodified README example now completes end-to-end with zero validation errors against the same real paper, and the full test suite passes (232/233 — the one failure is pre-existing and unrelated, a Windows path-separator assertion in a local-file test).

Environment

  • openai-agents==0.18.1 (resolved from the unpinned >=0.14 floor)
  • Windows 11, Python 3.12.12
Lenguaje dominante
Python
Estrellas
3.1k
Forks
492
Sin PR fusionados en 30 d

Preparar el entorno

Primeros pasos

  1. Lee el issue completo y luego la guía de contribución del proyecto.
  2. Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
  3. Haz un fork del repositorio y trabaja en una rama.
  4. Abre un pull request que haga referencia al número del issue.

Más de LLMQuant/quant-mind

Todos los issues de LLMQuant/quant-mind

Issues similares

Más issues de Python

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.