HKUDS/DeepTutor · error · ValueError

IMA accepts at most {MAX_IMPORT_URLS} URLs per call.

Error message

IMA accepts at most {MAX_IMPORT_URLS} URLs per call.

What it means

ValueError from IMA client import_urls: after cleaning/dedup, more than MAX_IMPORT_URLS URLs were supplied in a single call; the IMA endpoint enforces a per-request batch cap.

Source

Thrown at deeptutor/services/rag/pipelines/ima/client.py:224

        )

    # ----- writing --------------------------------------------------------

    async def import_urls(self, urls: list[str], *, folder_id: str = "") -> list[ImaImportedUrl]:
        """Add up to :data:`MAX_IMPORT_URLS` web pages to the bound library.

        IMA reports a per-URL verdict rather than failing the batch, so partial
        success is normal and is returned as-is for the caller to report.
        """
        cleaned: list[str] = []
        for raw in urls:
            url = str(raw or "").strip()
            if url and url not in cleaned:
                cleaned.append(url)
        if not cleaned:
            raise ValueError("At least one URL is required.")
        if len(cleaned) > MAX_IMPORT_URLS:
            raise ValueError(f"IMA accepts at most {MAX_IMPORT_URLS} URLs per call.")

        # ``folder_id`` is required here, and the root folder's id is the
        # knowledge base id itself.
        target = str(folder_id or "").strip() or self._config.knowledge_base_id
        data = await self._wire.post(
            "import_urls",
            {
                "urls": cleaned,
                "knowledge_base_id": self._config.knowledge_base_id,
                "folder_id": target,
            },
        )
        results = parse_imported_urls(data)
        if results:
            return results
        # Some responses acknowledge the batch without echoing per-URL rows.
        return [ImaImportedUrl(url=url, ok=True) for url in cleaned]

View on GitHub (pinned to 3e82f13042)

Solutions

1. Chunk the URL list into batches of MAX_IMPORT_URLS and issue one import_urls call per chunk.
2. Add a client-side count check against the limit before submitting.
3. Trim unnecessary URLs from the batch.

Example fix

# before
await client.import_urls(folder_id, all_500_urls)
# after
for chunk in batched(urls, MAX_IMPORT_URLS):
    await client.import_urls(folder_id, list(chunk))

cleaned = dedupe(urls)
assert 0 < len(cleaned) <= MAX_IMPORT_URLS, "chunk the URL list"

When it happens

Trigger: Calling import_urls with a list longer than MAX_IMPORT_URLS (deduplicated count is what's checked, so removing duplicates alone won't help past the cap).

Common situations: Bulk-import scripts feeding dozens/hundreds of URLs at once; UI allowing arbitrary multi-paste without chunking.

Related errors

• IMA accepts at most {MAX_DETAIL_IDS} knowledge base IDs per
• At least one URL is required.
• Client ID and API key are required.
• 'every' interval must be at least 30 seconds
• OpenAI-compatible embedding model '{model}' does not support

AI-assisted analysis of HKUDS/DeepTutor@3e82f13042 (2026-08-27). Data as JSON: /api/errors/04c48d5fec7c2bb1. Report an issue: GitHub.