HKUDS/DeepTutor · error · ValueError

IMA accepts at most {MAX_DETAIL_IDS} knowledge base IDs per

Error message

IMA accepts at most {MAX_DETAIL_IDS} knowledge base IDs per request.

What it means

ValueError from IMA client get_knowledge_bases: after normalization/dedup, more than MAX_DETAIL_IDS (20) knowledge base IDs were passed to the detail lookup, exceeding the endpoint's per-request ID cap.

Source

Thrown at deeptutor/services/rag/pipelines/ima/client.py:310

                }
            )
        return {
            "knowledge_bases": knowledge_bases,
            "next_cursor": cursor_state.next_cursor,
            "is_end": cursor_state.is_end,
        }

    async def get_knowledge_bases(self, ids: list[str]) -> dict[str, dict[str, Any]]:
        """Return details for at most :data:`MAX_DETAIL_IDS` knowledge base ids."""
        normalized: list[str] = []
        for item in ids:
            kb_id = str(item or "").strip()
            if kb_id and kb_id not in normalized:
                normalized.append(kb_id)
        if not normalized:
            return {}
        if len(normalized) > MAX_DETAIL_IDS:
            raise ValueError(
                f"IMA accepts at most {MAX_DETAIL_IDS} knowledge base IDs per request."
            )

        data = await self._wire.post("get_knowledge_base", {"ids": normalized})
        infos = data.get("infos")
        if not isinstance(infos, dict):
            return {}
        return {str(kb_id): info for kb_id, info in infos.items() if isinstance(info, dict)}

    async def get_knowledge_base(self) -> dict[str, Any]:
        """Return the bound knowledge base's info, or ``{}`` when unknown.

        Doubles as the credential check: bad credentials raise
        :class:`ImaAuthError`, while a well-formed but unknown id simply yields
        no entry for it.
        """
        kb_id = self._config.knowledge_base_id
        return (await self.get_knowledge_bases([kb_id])).get(kb_id, {})

View on GitHub (pinned to 3e82f13042)

Solutions

1. Batch IDs into groups of ≤20 and merge the returned info dicts.
2. Request only the specific KBs you need to display.
3. Cache detail lookups to avoid re-fetching wide ID sets.

Example fix

# before
infos = await client.get_knowledge_bases(all_ids)  # 50 ids
# after
infos = {}
for chunk in batched(all_ids, 20):
    infos.update(await client.get_knowledge_bases(list(chunk)))

normalized = dedupe(kb_ids)
if len(normalized) > MAX_DETAIL_IDS:
    normalized = normalized[:MAX_DETAIL_IDS]  # or batch

When it happens

Trigger: Calling get_knowledge_bases with >20 IDs directly, or via search_knowledge_bases / get_knowledge_base when the search result set resolves to more than 20 IDs.

Common situations: Listing a large tenant with many KBs and then fetching details for all of them; pagination logic that assumes unbounded ID lookups.

Related errors

• IMA accepts at most {MAX_IMPORT_URLS} URLs per call.
• At least one URL is required.
• Client ID and API key are required.
• OpenAI-compatible embedding model '{model}' does not support
• openai_sdk adapter does not support multimodal `contents`. P

AI-assisted analysis of HKUDS/DeepTutor@3e82f13042 (2026-08-27). Data as JSON: /api/errors/0ac0ed73c770cfbc. Report an issue: GitHub.